Obese patients are particularly sensitive to the sedative and respiratory depressive effects of long-acting opioids. Many obese patients also have obstructive sleep apnoea syndrome (OSAS) and will be prone to airway obstruction and desaturation in the postoperative period, especially if opioids have been given.1 2 Given this background, multimodal analgesia is advocated for bariatric surgery with the aim of reducing opioid use.3 4 At the time of writing, no studies were able to demonstrate a technique that can consistently remove the need for any postoperative opioid analgesia. In this study, we report the use of an anaesthesia protocol that allowed a significant proportion of our patients undergoing laparoscopic sleeve gastrectomy to be completely free from any long-acting potent opioids in the intra-operative and postoperative period.

This was a prospective observational study. The study was conducted in a private hospital in Hong Kong that has been accredited as a Centre of Excellence for Bariatric Surgery. All patients scheduled for laparoscopic sleeve gastrectomy for management of obesity or type 2 diabetes from 1 January 2015 onwards were anaesthetised using the same protocol. We analysed 30 consecutive cases between 1 January 2015 and 31 March 2015 to investigate the postoperative opioid requirements using this anaesthesia protocol. Patients were excluded from the case series if they had contra-indications or allergy to any of the anaesthetic or analgesic drugs, or if anaesthesia deviated from the standard protocol for any reason. Three patients were excluded—one was taking serotonin-specific reuptake inhibitor antidepressants and pethidine was avoided to prevent serotonin syndrome (morphine given instead); one was allergic to non-steroidal anti-inflammatory drugs (NSAIDs), so intravenous parecoxib and oral etoricoxib were not given; one accidentally had a larger dose of ketamine given intra-operatively than allowed by the protocol. Concomitant laparoscopic cholecystectomy was performed with laparoscopic sleeve gastrectomy in three patients who were included in the study.

All patients were fasted from midnight on the night before surgery. All operations were scheduled in the morning. Patients were premedicated with oral pantoprazole 40 mg on the night before surgery, and 2 g of oral paracetamol and 150 mg or 300 mg of oral pregabalin (for patients of body mass index <35 kg/m² or ≥35 kg/m², respectively) 2 hours before surgery.

Upon arrival in the operating theatre, intravenous access was established and 1 to 2 mg of intravenous midazolam was administered followed by an infusion of dexmedetomidine. The dose of dexmedetomidine was titrated according to the calculated lean body weight (LBW) using the Hume formula.5 The starting dose of the dexmedetomidine was 0.2 µg/kg/h using LBW.6 No loading dose was given.

Standard monitoring was applied to the patient together with a bispectral index (BIS) monitor and peripheral nerve stimulation monitor. Graduated compression stockings and sequential compression devices were used for all patients. Induction of anaesthesia was accomplished with fentanyl 100 µg, a titrated dose of propofol, and either suxamethonium or rocuronium as appropriate. The trachea was intubated and patients were ventilated with a mixture of air, oxygen, and desflurane.

Intra-operatively, desflurane was titrated to maintain BIS value between 40 and 60. Muscle relaxation was maintained with a rocuronium infusion to keep a train-of-four count of 1. Dexmedetomidine infusion continued at 0.2 µg/kg/h or higher if necessary. Shortly after induction, the various supplementary analgesic drugs were given. A loading dose of ketamine 0.3 mg/kg LBW was given followed by intermittent boluses roughly equivalent to 0.2 to 0.3 mg/kg/h of LBW. Intravenous parecoxib 40 mg and tramadol 100 mg were given. Dexamethasone 8 mg and tropisetron 5 mg were given intravenously for prophylaxis of postoperative nausea and vomiting (PONV).

For intravenous fluids, patients were given 10 mL/kg actual body weight of either lactated Ringer’s solution or normal saline, then more were given as appropriate. Hypotension was treated with either ephedrine or phenylephrine.

When the surgeon started to close the wounds, rocuronium infusion was stopped and dexmedetomidine infusion rate was reduced to 0.1 µg/kg/h. Wounds were infiltrated with 20 mL of 0.5% levobupivacaine. When all wounds were closed, dexmedetomidine infusion was stopped and desflurane switched off, muscle relaxation reversed by neostigmine and atropine. Patients were extubated when awake and able to obey command.

After extubation, patients were transferred to the post-anaesthetic care unit (PACU) for observation for 30 minutes, or longer if appropriate. If a patient required rescue analgesia, intravenous pethidine 20 mg with intravenous ketamine 5 mg was given, and the dose repeated if necessary. When 10 mg of intravenous ketamine had been given, further rescue analgesia was intravenous pethidine 20 mg without any more ketamine. This avoided administration of too much ketamine in an awake patient causing dizziness or hallucinations. When patients had good pain control and stable vital signs, they were transferred back to the ward. The standard postoperative protocol was initiated: if patients requested analgesics, an intramuscular injection of pethidine 50 mg was given, and repeated after 4 hours if necessary. By early evening, when vital signs were stable, patients were allowed sips of water followed by a fluid diet of 60 mL/h. Regular oral paracetamol and etoricoxib were given, and oral pregabalin was added to the protocol the next day. Opioid requirements were reviewed for 24 hours after surgery.

As part of the standard postoperative protocol, patients were asked to get off the bed and walk around the ward with the assistance of nursing or physiotherapy staff by the evening of the day of surgery. Provided there were no complications, patients were discharged on the second postoperative day. The anaesthesia protocol is summarised in Table 1.

Patient characteristics are shown in Table 2, and postoperative opioid requirements are listed in Table 3.

Of the 30 patients, no opioid rescue analgesia was required in 14 (46.7%) throughout the postoperative period; six (20%) required intravenous pethidine for rescue analgesia in the PACU, but not after their return to the ward. The remaining 10 (33.3%) patients were given intramuscular pethidine injections in the ward on request.

The mean postoperative opioid requirement per patient in the whole case series was 32 mg of pethidine. Among the 16 patients who required rescue analgesia in the ward or in the PACU, their mean opioid requirement was 60 mg of pethidine, with a range of 20 to 150 mg.

This anaesthetic protocol included a dexmedetomidine infusion that might cause hypotension and bradycardia due to its alpha-2 adrenoceptor blocking action. In our case series, 11 (36.7%) patients developed transient hypotension despite intravenous fluid loading and required either intravenous ephedrine or phenylephrine. One patient had transient intra-operative bradycardia requiring atropine, probably due to preoperative use of a beta blocker and low resting heart rate.

Opioids are among the world’s oldest known drugs. They have been used in anaesthesia traditionally as part of a balanced anaesthesia, to provide hypnosis and analgesia, to blunt the sympathetic response to surgery, and are the mainstay of postoperative analgesia in many situations. Morbidly obese patients, however, are particularly sensitive to the respiratory depressant effects of opioids. Taylor et al2 found that the use of opioids per se is a risk factor for respiratory events in the first 24 hours after surgery. Ahmad et al1 demonstrated in their study of 40 morbidly obese patients who presented for laparoscopic bariatric surgery, that in using desflurane and remifentanil-morphine–based anaesthesia, hypoxaemic episodes in the first 24 hours were common, and 14 of their 40 patients had more than five hypoxic episodes per hour despite supplementary oxygen.

Another concern with use of opioids in bariatric patients is the high incidence (>70%) of OSAS.7 In our study, 30% (n=9) of patients had OSAS confirmed by an overnight sleep study. The remaining patients were not tested although many had varying symptoms of OSAS. These untested patients were assumed to have OSAS unless proven otherwise. The American Society of Anesthesiologists recommends that in patients with OSAS, methods should be used to reduce or eliminate the requirement for systemic opioids.8 Hence, reducing perioperative opioid use by these obese patients can potentially reduce morbidity.

How can the anaesthetist avoid or reduce the use of perioperative opioids, and yet still provide balanced anaesthesia with hypnosis, analgesia, haemodynamic stability, and satisfactory postoperative analgesia? The first method is to combine general anaesthesia with regional analgesia techniques, such that anaesthetic agents will provide hypnosis while the regional blocks will provide analgesia and block sympathetic responses to surgery. Any form of major regional block in a morbidly obese patient can be technically challenging, however. Furthermore, with respect to bariatric surgery, most procedures are now performed laparoscopically, so that thoracic epidural analgesia techniques have become largely unnecessary.

Putting aside the use of regional analgesia, the second method to reduce perioperative opioid use is to use a combination of non-opioid agents with volatile agents or propofol to achieve analgesia and haemodynamic control.3 A point to note here is that as acute tolerance to the analgesic effects of opioids can rapidly develop (such as after 90 minutes of remifentanil infusion),9 any attempts to reduce postoperative opioid requirement must include an effort to either eliminate or reduce the use of intra-operative opioids. These techniques are now often described as opioid-free anaesthesia or non-opioid techniques.

Paracetamol, NSAID, or COX-2 inhibitors, gabapentinoids, ketamine and alpha-2 agonists, when used individually, have all been shown to reduce postoperative opioid requirement and improve pain relief.10 11 12 13 14 Different combinations of these agents, together with local anaesthetic infiltration of the wounds, have been reported for bariatric surgery, as discussed below.

In 2003, Feld et al15 described a technique of using sevoflurane combined with ketorolac, clonidine, ketamine, lignocaine, and magnesium for patients undergoing open gastric bypass. Compared with the control group where sevoflurane was used with fentanyl, they found the non-opioid group to be less sedated, with less morphine use in PACU although the total morphine use at 16 hours was not significantly different to the opioid group.

In 2006 Feld et al16 again described using desflurane combined with dexmedetomidine infusion, and compared it with a control group using desflurane and fentanyl, for patients undergoing open gastric bypass. In the dexmedetomidine group, there were lower pain scores and less morphine use in the PACU.

In 2005, Hofer et al17 described a case report of a super-obese patient weighing 433 kg who underwent open gastric bypass. No opioids were used but instead replaced with a high-dose dexmedetomidine infusion together with isoflurane.

As laparoscopic techniques have become more common in bariatric surgery, more studies have been carried out of non-opioid anaesthetic techniques for laparoscopic bariatric surgery. Tufanogullari et al18 described a technique in which either fentanyl or varying doses of dexmedetomidine were used with desflurane for laparoscopic bariatric surgery. All patients were also given celecoxib. Postoperatively, patients were given fentanyl boluses in PACU, then intravenous morphine via a patient-controlled analgesia system. The only statistical difference was decreased PACU fentanyl use in the dexmedetomidine groups.

Ziemann-Gimmel et al19 looked at 181 patients undergoing laparoscopic gastric bypass. In the treatment group, volatile anaesthetics were used together with intravenous paracetamol and ketorolac. Postoperatively patients were given regular paracetamol and ketorolac. If there was breakthrough pain, intermittent oral oxycodone or intravenous hydromorphone was given. A small number of patients in this treatment group (3/89) were able to remain opioid-free throughout, and 15 patients did not require opioid medications when they were back to the ward.

In another study where the primary outcome was the incidence of PONV, Ziemann-Gimmel et al20 evaluated 119 patients undergoing laparoscopic bariatric surgery. The treatment group was managed with propofol infusion, dexmedetomidine infusion, paracetamol, ketorolac, and ketamine. The other group was managed with volatile anaesthetic and opioids. Postoperative analgesia regimen was the same as the previous study.19 They reported a large reduction in PONV in their treatment group.

While most studies reported decreased requirement of opioids for postoperative analgesia in their non-opioid groups, very few studies could achieve zero postoperative opioid use. Only Ziemann-Gimmel et al19 could achieve total opioid sparing in a small proportion (3 out of 92 patients) of the treatment group by using intra-operative and postoperative intravenous paracetamol and ketorolac.

Most of these earlier studies used a combination of only a few of the available non-opioid adjuncts. Dexmedetomidine remains a mainstay of non-opioid adjunct in most of these studies. We hence proposed the use of a wider mix of non-opioid adjuncts, using a combination of paracetamol, COX-2 inhibitor, pregabalin, ketamine, dexmedetomidine, and local anaesthesia infiltration. In contrast to the earlier studies, in our study we were able to achieve zero postoperative opioid use in a significant percentage of patients (46.7%).

In our protocol, the only opioid given during anaesthesia was fentanyl 100 µg for intubation, and tramadol 100 mg, a weak opioid, shortly after induction. All other opioid analgesics, if required, were given after the patient was awake. This avoided having to blindly give intra-operative long-acting opioids during anaesthesia, and allowed better titration of the drug by giving small boluses each time with the patient awake.

Dexmedetomidine was a useful agent in our protocol. Before the addition of this agent to our protocol, total opioid sparing was very difficult to achieve. Dexmedetomidine is a highly selective alpha-2 adrenoceptor blocker, with analgesic and sedative properties.21 Previous study of its use in bariatric anaesthesia has failed to show any reduction in opioid requirements.18 In our protocol, we used more non-opioid adjuncts, and since we calculated the infusion dose using LBW instead of total body weight (TBW), overall we administered a much lower dose of dexmedetomidine.

Infusion of dexmedetomidine may cause initial hypertension and tachycardia (especially during a loading dose infusion), followed by hypotension and bradycardia. In our study, no loading dose was given. Of the 30 patients, 11 (36.7%) developed transient hypotension despite intravenous fluid loading and required either intravenous ephedrine or phenylephrine. This transient hypotension was also aggravated by putting the patient in a steep reverse trendelenburg position to facilitate surgical exposure, which decreases the venous return. When using dexmedetomidine in bariatric surgery, care must be taken to ensure the patient is euvolaemic.

Ketamine was another useful adjunct in our protocol. Ketamine is an N-methyl-D-aspartate receptor antagonist with strong analgesic properties when given at subanaesthetic doses.22 The use of ketamine has advantages in morbidly obese patients as it causes little respiratory depression compared with opioids. In our protocol, we used LBW to calculate the ketamine dose, and used relatively low ketamine doses (0.3 mg/kg bolus followed by 0.2-0.3 mg/kg/h with intermittent boluses). This resulted in a low total ketamine dose, with a mean of 31 mg ketamine per patient (range, 25-50 mg). Midazolam 1 to 2 mg was also given at induction to prevent any psychomimetic reactions caused by ketamine. No patient developed any hallucinations or dysphoria and there was no delay in emergence noted in our patients.

In our protocol, we chose to use LBW to calculate the dose for dexmedetomidine and ketamine. The classic teaching is that for obese patients, anaesthetic drugs can be dosed according to the TBW versus ideal body weight or LBW according to lipid solubility. Lipophilic drugs are better dosed according to actual body weight due to an increase in volume of distribution, whereas hydrophilic drugs are better dosed according to LBW or ideal body weight.23 Lean body weight is significantly correlated with cardiac output, and drug clearance increases proportionately with LBW.6

There is insufficient information regarding the pharmacokinetics and pharmacodynamics of dexmedetomidine and ketamine in the morbidly obese patient. In the few previous studies regarding dexmedetomidine and bariatric anaesthesia, TBW was used as dosing scalars. For example, Feld et al16 used 0.5 µg/kg TBW loading dose followed by 0.4 µg/kg/h infusion in their series of 10 patients with open gastric bypass. Ziemann-Gimmel et al20 used 0.5 µg/kg TBW loading dose followed by 0.1 to 0.3 µg/kg/h infusion for their group of 60 patients undergoing a variety of bariatric procedures. Tufanogullari et al18 gave no loading dose and infused from 0 to 0.8 µg/kg/h in their series of 80 patients undergoing laparoscopic banding or bypass. There were little data regarding ketamine dose in bariatric surgery. We chose to dose these two drugs using LBW to see how our results would differ from the other published studies.

Our study has several limitations. It was a prospective observational study with a relatively small number of cases. We do not have data to compare this protocol with our previous protocols, nor do we have data in the form of a randomised controlled trial to look at the isolated effect of any of the drugs used.

The opioid that we used for rescue analgesia was pethidine, given intravenously in the recovery room by the anaesthetist, or given intramuscularly on the ward by the nurses upon standing order. One can argue that the mean opioid dose per patient was not accurate as some were given small intravenous boluses and others were given intramuscular injections of fixed dose. To accurately assess the postoperative parenteral opioid requirements in theory, all patients should be given a patient-controlled analgesia system to deliver boluses of parenteral opioids as required. This, however, is not practical and not necessary for the patient, given that two thirds of our patients did not require any opioids at all. This would also represent a lot of drug wastage when the whole cassette of drugs was unused.

We were able to demonstrate that a significant proportion of patients did not require any opioids, but we do not have data to demonstrate a reduction in respiratory complications or an improvement in time to ambulation or discharge. This could be the basis for further studies.

All authors have disclosed no conflicts of interest.

1. Ahmad S, Nagle A, McCarthy RJ, et al. Postoperative hypoxaemia in morbidly obese patients with and without obstructive sleep apnea undergoing laparoscopic bariatric surgery. Anesth Analg 2008;107:138-43. Crossref

2. Taylor S, Kirton OC, Staff I, Kozol RA. Postoperative day one: a high risk period for respiratory events. Am J Surg 2005;190:752-6. Crossref

3. Mulier JP. Perioperative opioids aggravate obstructive breathing in sleep apnea syndrome: mechanisms and alternative anaesthesia strategies. Curr Opin Anaesthesiol 2016;29:129-33. Crossref

4. Alvarez A, Singh PM, Sinha AC. Postoperative analgesia in morbid obesity. Obes Surg 2014;24:652-9. Crossref

5. Hume R. Prediction of lean body mass from height and weight. J Clin Pathol 1966;19:389-91. Crossref

6. Ingrande J, Lemmens HJ. Dose adjustment of anaesthetics in the morbidly obese. Br J Anaesth 2010;105 Suppl 1:i16-23. Crossref

7. Lopez PP, Stefan B, Schulman CI, Byers PM. Prevalence of sleep apnea in morbidly obese patients who presented for weight loss surgery evaluation: more evidence for routine screening for obstructive sleep apnea before weight loss surgery. Am Surg 2008;74:834-8.

8. American Society of Anesthesiologists Task Force on Perioperative Management of patients with obstructive sleep apnea. Practice guidelines for the perioperative management of patients with obstructive sleep apnea: an updated report by the American Society of Anesthesiologists Task Force on Perioperative Management of patients with obstructive sleep apnea. Anesthesiology 2014;120:268-86. Crossref

9. Vinki HR, Kissin I. Rapid development of tolerance to analgesia during remifentanil infusion in humans. Anesth Analg 1998;86:1307-11. Crossref

10. Dahl JB, Nielsen RV, Wetterslev J, et al. Post-operative analgesic effects of paracetamol, NSAIDs, glucocorticoids, gabapentinoids and their combinations: a topical review. Acta Anaesthesiol Scand 2014;58:1165-81. Crossref

11. Blaudszun G, Lysakowski C, Elia N, Tramèr MR. Effect of perioperative systemic α2 agonists on postoperative morphine consumption and pain intensity: systematic review and meta-analysis of randomized controlled trials. Anesthesiology 2012;116:1312-22. Crossref

12. Cabrera Schulmeyer MC, de la Maza J, Ovalle C, Farias C, Vives I. Analgesic effects of a single preoperative dose of pregabalin after laparoscopic sleeve gastrectomy. Obes Surg 2010;20:1678-81. Crossref

13. Weinbroum AA. Non-opioid IV adjuvants in the perioperative period: pharmacological and clinical aspects of ketamine and gabapentinoids. Pharmocol Res 2012;65:411-29. Crossref

14. Alimian M, Imani F, Faiz SH, Pournajafian A, Navadegi SF, Safari S. Effect of oral pregabalin premedication on post-operative pain in laparoscopic gastric bypass surgery. Anesth Pain Med 2012;2:12-6. Crossref

15. Feld JM, Laurito CE, Beckerman M, Vincent J, Hoffman WE. Non-opioid analgesia improves pain relief and decreases sedation after gastric bypass surgery. Can J Anaesth 2003;50:336-41. Crossref

16. Feld JM, Hoffam WE, Stechert MM, Hoffman IW, Anada RC. Fentanyl or dexmedetomidine combined with desflurane for bariatric surgery. J Clin Anesth 2006;18:24-8. Crossref

17. Hofer RE, Sprung J, Sarr MG, Wedel DJ. Anesthesia for a patient with morbid obesity using dexmedetomidine without narcotics. Can J Anaesth 2005;52:176-80. Crossref

18. Tufanogullari B, White PF, Peixoto MP, et al. Dexmedetomidine infusion during laparoscopic bariatric surgery: the effect on recovery outcome variables. Anesth Analg 2008;106:1741-8. Crossref

19. Ziemann-Gimmel P, Hensel P, Koppman J, Marema R. Multimodal analgesia reduces narcotic requirements and antiemetic rescue medication in laparoscopic Roux-en-Y gastric bypass surgery. Surg Obes Relat Dis 2013;9:975-80. Crossref

20. Ziemann-Gimmel P, Goldfarb AA, Koppman J, Marema RT. Opioid-free total intravenous anaesthesia reduces postoperative nausea and vomiting in bariatric surgery beyond triple prophylaxis. Br J Anaesth 2014;112:906-11. Crossref

21. Carollo DS, Nossaman BD, Ramadhyani U. Dexmedetomidine: a review of clinical applications. Curr Opin Anaesthesiol 2008;21:457-61. Crossref

22. Gammon D, Bankhead B. Perioperative pain adjuncts. In: Johnson KB, editor. Clinical pharmacology for anesthesiology. McGraw-Hill Education; 2014: 157-78.

23. Sinha AC, Eckmann DM. Anesthesia for bariatric surgery. In: Miller RD, Eriksson LI, Fleisher LA, Wiener-Kronish JP, Young WL, editors. Miller’s anesthesia. 7th ed. Philadelphia: Churchill Livingston; 2015: 2089-104.

Ventriculoperitoneal (VP) shunting is one of the most common neurosurgical procedures performed to treat patients with hydrocephalus, which is a disorder related to an abnormal accumulation of cerebrospinal fluid (CSF) in the brain. The operation involves diverting CSF from the ventricles of the brain to the peritoneal cavity of the abdomen by catheter implantation. Despite being a well-established procedure, shunt failure can be as high as 70% in the first year with an annual occurrence rate of 5% thereafter.1 One of the main causes for failure is shunt infection, a potentially debilitating complication that more than doubles the risk of death and exposes affected patients to 3 times as many neurosurgical procedures as non-infected patients.2 Shunt infection varies and occurs in 3% to 17% of patients. Standard management involves intravenous antibiotic therapy, shunt removal, insertion of an external ventricular drain, and replacement with a new shunt once the patient’s CSF is free of microbial infection.1 3 4 5 6 The economic impact of VP shunt infection can be considerable. In the US, the median cost per episode per patient has been reported as US$23 500, accountable for US$2.0 billion in annual hospital charges.7 8 Evidence suggests that the adoption of a strict institutional implantation protocol can significantly reduce the risk of this most challenging shunt complications.9 10 11 12 This retrospective study aimed to determine the frequency of primary VP shunt reinsertions and infection among patients treated in Hong Kong’s public health system and to identify risk factors for shunt infection.

This was a multicentre retrospective study of patients who underwent VP shunt implantation at all seven Hong Kong Hospital Authority neurosurgical units. The Hospital Authority is a public service highly subsidised by the Hong Kong Special Administrative Region Government, and responsible for delivering health care for 90% of inpatient bed days in the city.13 Clinical research ethics committee approval was obtained from the participating centres. Patients who underwent primary VP shunting from 1 January 2009 to 31 December 2011 were included in this study. Those who underwent alternative CSF diversion procedures or those with a history of VP shunt implantation were excluded from this review. Data from clinical records, operation notes, medication-dispensing records, CSF biochemistry, cell counts, and microbiological cultures were collected. The primary endpoint for this study was primary VP shunt infection. The criteria for shunt infection were: (1) CSF or shunt hardware culture that yielded the pathogenic micro-organism with associated compatible symptoms and signs of central nervous system (CNS) infection or shunt malfunction5 14 15; or (2) surgical incision site infection, as defined by the National Nosocomial Infection Surveillance System, requiring shunt reinsertion (even in the absence of a positive culture)16; or (3) intraperitoneal pseudocyst formation (even in the absence of a positive culture). Secondary endpoints were shunt malfunction, defined as unsatisfactory CSF drainage that required shunt reinsertion, and 30-day mortality. Potential risk factors for shunt infection were classified as patient-, disease-, or surgical-related factors. All subjects were followed up for at least 30 days from the operation date or until death.

Statistical analysis was carried out using Pearson’s Chi squared test, Fisher’s exact test, and binary logistic regression to identify risk factors for shunt infection. The Kaplan-Meier (log-rank) and Cox proportional hazards models were employed for survival analysis. Patient, disease, and surgical factors were used as covariates and a stepwise regression strategy was adopted (Table 1). P values of <0.05 were considered statistically significant. All tests were performed using the Statistical Package for the Social Sciences (Windows version 16.0.1; SPSS Inc, Chicago [IL], US).

During the 3-year period, 538 patients underwent primary VP shunt implantation and 87% (n=470) had complete clinical follow-up with a median duration of 37 months (range, 3 days to 76 months). Seven (1%) patients were transferred to other hospitals within 30 days of the procedure. The median duration of hospitalisation was 42 days (range, 3 days to 36 months) and the median length of time from admission to shunting was 18 days (range, <1 day to 21 months). The clinical features and surgical variables are presented in Table 1. The mean (± standard deviation) age of patients was 48 ± 13 years (range, 13-88 years) and the male-to-female ratio was 1:1. In the study group, 80 (15%) were paediatric patients and 48 (9%) were infants. Overall, primary VP shunting was performed for post-aneurysmal subarachnoid haemorrhage communicating hydrocephalus in 169 (31%) patients, for CNS neoplasms in 164 (30%) patients, and for spontaneous intracerebral or intraventricular haemorrhage in 64 (12%) patients. For patients who had preoperative CSF sampling performed, the mean red blood cell count was 1900/µL, white cell count was 17/µL, total protein level was 0.78 g/L, and glucose level was 3.6 mmol/L.

Over one quarter of patients (n=155, 29%) had never had prior cranial neurosurgery and approaching half had undergone either one (n=141, 26%) or two (n=115, 21%) previous procedures. Antiseptic skin preparation was povidone-iodine 10% combined with another antiseptic in 422 (78%) patients and with povidone iodine alone in the remainder. The mean operating time for VP shunting was 75 ± 29 minutes. All patients had antibiotic prophylaxis of whom 328 (61%) were prescribed a third-generation cephalosporin and 40 (7%) had vancomycin. Twelve (2%) patients had a rifampicin-clindamycin antibiotic-impregnated ventricular catheter as part of the shunt system. The majority of operations were performed in an emergency setting (n=312, 58%) and shunt implantation was the sole procedure performed (n=514, 96%). The burr hole was most frequently positioned at the parietal location in 320 (59%) patients and 135 (25%) had a frontal burr hole. New burr holes were fashioned for shunt placement 95% of the time. The median number of surgeons was two, with a third of shunts performed by higher neurosurgical trainees (n=174, 32%) and the remaining performed by a neurosurgical specialist. Almost three quarters of VP shunts had a fixed-pressure valve (n=390, 72%) and the predominant design utilised was the Integra Pudenz flushing valve (Integra LifeSciences Corporation, Plainsboro [NJ], US) in 324 (60%) patients.

The rate of VP shunt reinsertion was 16% (n=87) and infection was 7% (n=36). The main causes for reinsertion were malfunction (9%) followed by infection. The annual proportion of shunts that required reoperation or were infected was comparable (P=0.87) [Fig 1]. The median time from shunt implantation to shunt removal for infection was 64 days (range, 2 days to 10 months). A cumulative risk for infection was noted affecting 3% of shunts in the first 30 days, 6% in 6 months, and 7% in 1 year. Although 68 (13%) patients were lost to follow-up, attrition analysis revealed that this did not affect infection rates. The mean follow-up duration in this subgroup between those with infection and those without was comparable at 526 days and 554 days, respectively (P=0.43). In addition, the incidence of shunt infection in patients with incomplete follow-up (5%) was similar to those with complete follow-up (7%) [P=0.42].

Most infections manifested as meningitis or ventriculitis (n=19, 53%), followed by wound breakdown (n=15, 42%) and peritonitis (n=2, 6%). The most common causative bacteria were coagulase-negative staphylococci (CoNS) [n=25, 69%] of which methicillin resistance was detected in 19 (76%) patients (Table 2). All CoNS species were sensitive to vancomycin with a quarter of methicillin-resistant (MR) species susceptible to aminoglycosides such as gentamicin or amikacin. The second most common infective agent affecting four (11%) patients was MR Staphylococcus aureus (MRSA). Polymicrobial infection was evident in six (17%) patients. One patient with peritonitis had mixed Gram-positive and -negative micro-organisms from CSF cultures.

The only patient risk factor for shunt infection was sex (Table 1). Male patients had a greater than two-fold increased odds of infection (odds ratio [OR]=2.2; 95% confidence interval [CI], 1.1-4.5). Traumatic brain injury (TBI) was the only disease risk factor (OR=7.8; 95% CI, 2.9-18.1). Surgical factors included the use of vancomycin as the prophylactic antibiotic (OR=3.7; 95% CI, 1.3-10.5) and shunts implanted as an emergency procedure (OR=2.2; 95% CI, 1.0-4.7). After adjusting for confounding factors, the independent risk factors for primary VP shunt infection were TBI (adjusted OR=6.2; 95% CI, 2.3-16.8), the use of vancomycin (adjusted OR=3.4; 95% CI, 1.1-11.0), and emergency shunting (adjusted OR=2.3; 95% CI, 1.0-5.1) [Table 3].

With respect to shunt infection, there was a difference in duration of shunt implantation for patients with the aforementioned risk factors as demonstrated by the significant separation of Kaplan-Meier survival curves (Fig 2). This was ascertained by Cox regression analysis (Table 3). Median shunt survival for trauma patients was 35 days (vs 154 days in non-trauma patients), 32 days for patients who received vancomycin (vs 124 days for alternative antibiotics), and 65 days for emergency operations (vs 208 days for elective operations).

In this study, 30-day all-cause mortality was 6% (n=32), but none was directly procedure-related. Almost half of these patients (n=15, 47%) had an underlying malignant CNS tumour; the majority being brain metastases (n=12, 80%). After accounting for patient age, sex, disease aetiology, shunt reinsertion and infection, a diagnosis of malignant brain tumour was the only significant independent predictor for 30-day mortality with an adjusted OR of 5.6 (95% CI, 2.6-11.7).

Ventriculoperitoneal CSF shunting has considerably reduced the morbidity and mortality of patients with hydrocephalus since its first description in 1908.17 More than a century later the operation remains the mainstay treatment for this condition. Despite the introduction of antibiotics, improvements in shunt materials and surgical techniques, VP shunt complications are common. Long-term epidemiological studies have indicated that more than half of all patients with CSF shunts will require a surgical revision in their lifetime.4 18 Shunt infection is a serious complication with potentially devastating consequences. Observational studies have recorded infection rates ranging between 3% and 17%, but more consistent estimates from larger patient cohorts cite rates of 6% to 8%.1 3 4 5 Our local shunt infection rate of 7% is relatively lower than other previously published findings and is in keeping with the results from other developed countries.

The wide range of shunt infection rates quoted in the literature is due in part to the diverse definitions adopted for shunt infection and the patient populations studied. Many studies have defined infection as a positive CSF microbial culture or the presence of CSF pleocytosis or low CSF glucose levels with clinical features of CNS infection.3 11 19 Due to the study design, a more pragmatic definition was adopted whereby infection was determined retrospectively by either a positive CSF or shunt hardware microbial culture in the presence of shunt malfunction.5 14 15 Nonetheless, it is acknowledged that the true incidence of shunt infection may be overestimated by false-positive cultures from skin flora. The reasons for selecting this interpretation for shunt infection were three-fold. First, it allowed for micro-organism identification and consequent epidemiological analysis; second, infection may not be clinically apparent with malfunctioning shunts; and third, CSF cultures alone cannot exclude infection in cases of shunt malfunction.15

The only disease risk factor independently associated with VP shunt infection was TBI. This may be due to two reasons. Delayed post-traumatic hydrocephalus invariably occurs in severe TBI patients and develops in over a third of those subject to decompressive craniectomies.20 Such patients often have a prolonged hospital stay and undergo multiple operations before a shunt is eventually implanted. In this cohort, TBI patients had a mean duration of hospitalisation of 89 days, which was 2 weeks more than the mean stay of 74 days for hydrocephalic patients with alternative neurosurgical conditions. Protracted hospitalisation may lead to skin colonisation with drug-resistant organisms that can evade single-agent conventional antibiotic prophylaxis.21 This is supported by evidence from this study where causative bacteria of shunt infection were resistant to the prescribed prophylactic antibiotic in 80% of TBI patients. These patients also had a mean number of three prior cranial procedures before shunting compared with two operations in patients with non-traumatic hydrocephalus aetiology. Previous surgery is well known to be a main cause of CSF leak in shunted patients and contributes to an increased risk of infection.5 22 Although in this patient series, the number of prior cranial procedures per se did not impart greater risk, detailed clinical data regarding CSF leak in TBI patients were not collected and therefore the influence of TBI on infection can only be inferred.

An unexpected finding was that patient age was not a risk factor for shunt infection. This is in contrast to several larger studies that identified paediatric patients, especially infants (younger than a year), to be particularly at risk.11 23 Infants have less-developed humoral and cellular immune systems with immature skin growth rendering them more vulnerable to shunt infection. A likely reason for this observation is the small number of paediatric patients (n=80, 15%) in this cohort with only 9% (n=48) being infants. Larger sample size may delineate more distinctive differences among age-groups.

There is little doubt that systemic antibiotic prophylaxis can prevent shunt infection.24 We, however, interestingly identified the sole use of vancomycin as a risk factor for shunt infection, a novel observation that has not been previously reported in the literature. The antibiotic was regularly reserved for patients allergic to penicillin-group antibiotics or for those with documented penicillin-resistant microbial infection or colonisation. This finding apparently seems counter-intuitive especially when all causative CoNS identified in this cohort were sensitive to vancomycin. The issue may lie with the timing of its administration before the procedure and its dosage. With regard to timing, systemic vancomycin requires slow intravenous infusion to reduce the risk of a hypersensitivity reaction that manifests as either red man syndrome or anaphylaxis and occurs in 3.7% to 47% of patients.25 To further illustrate the incidence of these symptoms, the first randomised controlled trial investigating its efficacy in shunt procedures was prematurely discontinued due to these adverse effects.26 Most hospital protocols require infusion rates over an hour as a minimum, but clinical trials have demonstrated that even lengthier 2-hour infusions can further reduce the frequency and severity of these reactions.25 27 Furthermore, the efficacy of vancomycin to treat CoNS and MRSA infections has been questioned due to observations of slower bactericidal activity, compared with nafcillin, than was previously recognised.28 To address this issue we suggest that rigid guidelines should be adhered to with respect to the adequate timing of vancomycin infusion before skin incision. Should more rapid infusions be required, for example, in the emergency setting, pre-administration of intravenous diphenhydramine before vancomycin infusion can prevent the development of red man syndrome.25

Limited data are available about the pharmacokinetics and CSF concentrations of vancomycin in neurosurgical patients. In a study reviewing intra-operative serum and CSF vancomycin concentrations of paediatric patients undergoing shunt implantation, the authors noted that CSF penetration was negligible in patients with non-inflamed meninges despite presumed adequate loading doses.29 This was echoed in a later study determining that among non-meningitic patients, vancomycin CNS penetration was poor with a CSF-to-serum ratio of only 18%.30 Its increasing use over the course of decades has also led to a corresponding rise in minimum bactericidal concentrations of CoNS.28 These unique findings should prompt an extensive review of prophylactic vancomycin use as there is currently no consensus on a recommended loading dose for neurosurgical procedures. In the meantime, researchers have attempted to improve the bactericidal activity in patients treated with vancomycin with varying measures of success. Vancomycin in combination with gentamicin results in more rapid bactericidal rates in animal models28 and has been proven to be as effective as third-generation cephalosporins in preventing surgical site infection for neurosurgical procedures in a randomised trial.31 Others have proposed intra-operative combined vancomycin-aminoglycoside administration either intraventricularly, for shunt hardware antibiotic bath immersion prior to implantation or applied in powder form within the subgaleal space of the wound with tenable positive results.11 32 33 34

Antibiotic-impregnated (by rifampicin with either clindamycin or minocycline) and silver-coated ventricular catheters offer the greatest promise in preventing shunt infection.35 36 There exists a growing body of evidence in support of antibiotic-impregnated ventricular catheters and they are gradually replacing conventional plain silicone catheters in daily practice with considerable cost savings.37 38 39 40 There are also accompanying concerns about the development of antibiotic-resistant micro-organisms and a recent meta-analysis has elucidated the higher risk of Gram-negative and MRSA shunt infections.38 In this series, only 12 patients received antibiotic-impregnated catheters during the study period so it is difficult to draw any conclusion about their effectiveness.

Emergency VP shunting is another surgical-related risk factor for infection and was performed in more than half of patients who underwent the procedure. Its significance is the greatest among the three independent factors identified and is possibly the most amenable to change in current practice. The clinical condition of most patients with hydrocephalus who require primary VP shunting does not warrant emergency surgery although a few indications exist, for example, obstructive pineal region or cerebellar tumours that may present with acute symptoms. More than two thirds of patients (70%) in this study had conditions that necessitated delayed shunting when the primary disease had been treated and the patients stabilised. It is most likely because of limited availability of operating theatre among other related resources and the general practice that VP shunting is delegated to more junior members of the surgical team that this phenomenon prevails. Several reasons support why ‘emergency’ primary shunting should be discouraged. It has long been established by several protocol-driven trials that shunting should be performed as the first procedure of the operative day to minimise the risk of contamination.9 10 11 To illustrate, a surgical incision time after 10 am was observed to be a predictor for infection.5 In elective procedures the neurosurgeon in-charge and other responsible operating theatre staff are likely to be more experienced in comparison with personnel involved in emergencies. In particular, it has been shown that individual surgical experience is an important factor for infection with researchers stating a higher incidence among neurosurgical trainees or in surgeons who performed fewer than 147 shunts within a decade.4 41 Nonetheless, using the former stratification of trainee versus specialist, this was not evident in our cohort. Another argument against ‘emergency’ VP shunting could be the location where the procedure is performed. For a variety of resource allocation reasons, shunting scheduled as an emergency procedure is often not performed in neurosurgery-designated operating theatres. A study investigating the distribution of bacteria in the operating room environment and its relationship with ventricular shunt infections concluded that positive environmental cultures were more likely to occur in a theatre not devoted to neurosurgery.42 Although procedure timing and location were not explored in this audit, it is believed that they were the principal explanations why shunts implanted as an emergency were more likely to become infected.

The time interval from shunt implantation to revision for infection in our study is longer than most published data with a median shunt survival of 64 days.5 34 Our data show that 92% (n=33) of shunt infections occurred within 6 months and is compatible with the commonly held belief that the infection begins intra-operatively with the inoculation of skin flora, either from the patient or surgeon, into the surgical wound.42 43 44 This is further substantiated by the predominance of CoNS and S aureus in 81% of bacterial cultures in this patient series and similarly in several previous reports.4 5 9 10 11 12 32 35 36 43 Coupled with positive research findings that theatre discipline during surgery reduces infection risk, it seems reasonable to conclude that institutional shunt implantation protocols should be established.9 10 11 12 32

Even though the performance of our neurosurgical community with regard to primary VP shunt infection meets international standards, there is room for improvement. The implementation of a standardised shunt surgery protocol that covers preoperative preparation as well as intra-operative and postoperative management has consistently been proven to be effective in reducing infection.9 10 11 12 32 The landmark study by Choux at al10 first demonstrated that meticulous measures—such as adopting a no-touch technique for shunt handling, limiting the length of shunt exposure time and the number of people in the operating room—have dramatically decreased shunt infection rates from 16% to less than 1%. It is our belief that a similarly comprehensive protocol should be developed and based on the findings of this preliminary study.

This study has several limitations. Data collection was retrospective so key clinical information such as the presence of CSF leak and patient co-morbidities were missing. This may have led to inadequate control for confounding factors. An additional limitation inherent in studies of this nature is the potential presence of observational bias where data were collected without blinding after outcomes were known. Follow-up was incomplete with 68 (13%) patients defaulting from clinical review over the course of 3 years. Inadequate follow-up duration was also noted; seven (1%) patients were transferred to other hospitals within 30 days of the procedure and this might have influenced 30-day all-cause mortality findings. Finally, our definition of shunt infection did not include abnormal CSF biochemistry criteria that could have confirmed or refuted positive culture results of specimens that might have been contaminated during collection.

This is the first territory-wide review of infection in primary VP shunts conducted in Hong Kong’s public health care setting. This study is also one of the largest in the literature examining shunt infection complications among a predominantly Chinese population. Shunt infection was the second most common cause for reinsertion occurring in 7% of patients. Significant independent predictors for shunt infection were TBI, vancomycin administration for prophylaxis, and procedures performed in an emergency setting. Although VP shunt infection rates meet international standards, there are areas of improvement that can be readily addressed such as the timing or dosage of vancomycin and the avoidance of performing the procedure as an emergency. The best approach to reducing shunt infection may be the design and adoption of a standardised shunt surgery protocol customised to local practice.

We would like to thank members of the Hospital Authority Head Office (HAHO) Co-ordinating Committee (Neurosurgery), the Clinical Effectiveness & Technology Management Department, the Division of Quality and Safety, and the Clinical Data Analysis and Reporting System Team, HAHO IT Service for their administrative advice and data collection. We also wish to thank Drs Chris YW Liu, Alphon HY Ip, and Claudia Law for their contributions in data collection and entry. This study was supported by the Tung Wah Group of Hospitals Neuroscience Research Fund.

All authors have disclosed no conflicts of interest.

1. Wong JM, Ziewacz JE, Ho AL, et al. Patterns in neurosurgical adverse events: cerebrospinal fluid shunt surgery. Neurosurg Focus 2012;33:E13. Crossref

2. Schoenbaum SC, Gardner P, Shillito J. Infections of cerebrospinal fluid shunts: epidemiology, clinical manifestations, and therapy. J Infect Dis 1975;131:543-52. Crossref

3. Birjandi A, Zare E, Hushmandi F. Ventriculoperitoneal shunt infection: a review of treatment. Neurosurg Q 2012;22:145-8. Crossref

4. Borgbjerg BM, Gjerris F, Albeck MJ, Børgesen SE. Risk of infection after cerebrospinal fluid shunt: an analysis of 884 first-time shunts. Acta Neurochir (Wien) 1995;136:1-7. Crossref

5. Korinek AM, Fulla-Oller L, Boch AL, Golmard JL, Hadiji B, Puybasset L. Morbidity of ventricular cerebrospinal fluid shunt surgery in adults: an 8-year study. Neurosurgery 2011;68:985-94; discussion 994-5. Crossref

6. Patwardhan RV, Nanda A. Implanted ventricular shunts in the United States: the billion-dollar-a-year cost of hydrocephalus treatment. Neurosurgery 2005;56:139-44; discussion 144-5. Crossref

7. Simon TD, Riva-Cambrin J, Srivastava R, et al. Hospital care for children with hydrocephalus in the United States: utilization, charges, comorbidities, and deaths. J Neurosurg Pediatr 2008;1:131-7. Crossref

8. Shannon CN, Simon TD, Reed GT, et al. The economic impact of ventriculoperitoneal shunt failure. J Neurosurg Pediatr 2011;8:593-9. Crossref

9. Faillace WJ. A no-touch technique protocol to diminish cerebrospinal fluid shunt infection. Surg Neurol 1995;43:344-50. Crossref

10. Choux M, Genitori L, Lang D, Lena G. Shunt implantation: reducing the incidence of shunt infection. J Neurosurg 1992;77:875-80. Crossref

11. Kestle JR, Riva-Cambrin J, Wellons JC 3rd, et al. A standardized protocol to reduce cerebrospinal fluid shunt infection: the Hydrocephalus Clinical Research Network Quality Improvement Initiative. J Neurosurg Pediatr 2011;8:22-9. Crossref

12. Pirotte BJ, Lubansu A, Bruneau M, Loqa C, Van Cutsem N, Brotchi J. Sterile surgical technique for shunt placement reduces the shunt infection rate in children: preliminary analysis of a prospective protocol in 115 consecutive procedures. Childs Nerv Syst 2007;23:1251-61. Crossref

13. World Health Organization and Department of Health, Hong Kong. Hong Kong (China) health service delivery profile, 2012. Available from: http://www.wpro.who.int/health_services/service_delivery_profile_hong_kong_(china).pdf. Accessed Mar 2016.

14. Overturf GD. Defining bacterial meningitis and other infections of the central nervous system. Pediatr Crit Care Med 2005;6 Suppl:S14-8. Crossref

15. Vanaclocha V, Sáiz-Sapena N, Leiva J. Shunt malfunction in relation to shunt infection. Acta Neurochir (Wien) 1996;138:829-34. Crossref

16. Horan TC, Gaynes RP, Martone WJ, Jarvis WR, Emori TG. CDC definitions of nosocomial surgical site infections, 1992: a modification of CDC definitions of surgical wound infections. Infect Control Hosp Epidemiol 1992;13:606-8. Crossref

17. Kausch W. Die behandlung des hydrocephalus der kleinen kinder. Arch Klin Chir 1908;87:709-96.

18. Tuli S, Drake J, Lawless J, Wigg M, Lamberti-Pasculli M. Risk factors for repeated cerebrospinal shunt failures in pediatric patients with hydrocephalus. J Neurosurg 2000;92:31-8. Crossref

19. Odio C, McCracken GH Jr, Nelson JD. CSF shunt infections in pediatrics. A seven-year experience. Am J Dis Child 1984;138:1103-8. Crossref

20. Honeybul S, Ho KM. Incidence and risk factors for post-traumatic hydrocephalus following decompressive craniectomy for intractable intracranial hypertension and evacuation of mass lesions. J Neurotrauma 2012;29:1872-8. Crossref

21. Wang KW, Chang WN, Shih TY, et al. Infection of cerebrospinal fluid shunts: causative pathogens, clinical features, and outcomes. Jpn J Infect Dis 2004;57:44-8.

22. Jeelani NU, Kulkarni AV, Desilva P, Thompson DN, Hayward RD. Postoperative cerebrospinal fluid wound leakage as a predictor of shunt infection: a prospective analysis of 205 cases. Clinical article. J Neurosurg Pediatr 2009;4:166-9. Crossref

23. Davis SE, Levy ML, McComb JG, Masri-Lavine L. Does age or other factors influence the incidence of ventriculoperitoneal shunt infections? Pediatr Neurosurg 1999;30:253-7. Crossref

24. Ratilal B, Costa J, Sampaio C. Antibiotic prophylaxis for surgical introduction of intracranial ventricular shunts. Cochrane Database Syst Rev 2006;(3):CD005365. Crossref

25. Sivagnanam S, Deleu D. Red man syndrome. Crit Care 2003;7:119-20. Crossref

26. Odio C, Mohs E, Sklar FH, Nelson JD, McCracken GH Jr. Adverse reactions to vancomycin used as prophylaxis for CSF shunt procedures. Am J Dis Child 1984;138:17-9. Crossref

27. Healy DP, Sahai JV, Fuller SH, Polk RE. Vancomycin-induced histamine release and “red man syndrome”: comparison of 1- and 2-hour infusions. Antimicrob Agents Chemother 1990;34:550-4. Crossref

28. Stevens DL. The role of vancomycin in the treatment paradigm. Clin Infect Dis 2006;42 Suppl 1:S51-7. Crossref

29. Fan-Havard P, Nahata MC, Bartkowski MH, Barson WJ, Kosnik EJ. Pharmacokinetics and cerebrospinal fluid (CSF) concentrations of vancomycin in pediatric patients undergoing CSF shunt placement. Chemotherapy 1990;36:103-8. Crossref

30. Albanèse J, Léone M, Bruguerolle B, Ayem ML, Lacarelle B, Martin C. Cerebrospinal fluid penetration and pharmacokinetics of vancomycin administered by continuous infusion to mechanically ventilated patients in an intensive care unit. Antimicrob Agents Chemother 2000;44:1356-8. Crossref

31. Pons VG, Denlinger SL, Guglielmo BJ, et al. Ceftizoxime versus vancomycin and gentamicin in neurosurgical prophylaxis: a randomized, prospective, blinded clinical study. Neurosurgery 1993;33:416-22; discussion 422-3. Crossref

32. Choksey MS, Malik IA. Zero tolerance to shunt infections: can it be achieved? J Neurol Neurosurg Psychiatry 2004;75:87-91.

33. Abdullah KG, Attiah MA, Olsen AS, Richardson A, Lucas TH. Reducing surgical site infections following craniotomy: examination of the use of topical vancomycin. J Neurosurg 2015;123:1600-4. Crossref

34. Ragel BT, Browd SR, Schmidt RH. Surgical shunt infection: significant reduction when using intraventricular and systemic antibiotic agents. J Neurosurg 2006;105:242-7. Crossref

35. Keong NC, Bulters DO, Richards HK, et al. The SILVER (Silver Impregnated Line Versus EVD Randomized trial): a double-blind, prospective, randomized, controlled trial of an intervention to reduce the rate of external ventricular drain infection. Neurosurgery 2012;71:394-403; discussion 403-4. Crossref

36. Sciubba DM, Stuart RM, McGirt MJ, et al. Effect of antibiotic-impregnated shunt catheters in decreasing the incidence of shunt infection in the treatment of hydrocephalus. J Neurosurg 2005;103 Suppl:131-6. Crossref

37. Thomas R, Lee S, Patole S, Rao S. Antibiotic-impregnated catheters for the prevention of CSF shunt infections: a systematic review and meta-analysis. Br J Neurosurg 2012;26:175-84. Crossref

38. Konstantelias AA, Vardakas KZ, Polyzos KA, Tansarli GS, Falagas ME. Antimicrobial-impregnated and -coated shunt catheters for prevention of infections in patients with hydrocephalus: a systematic review and meta-analysis. J Neurosurg 2015;122:1096-112. Crossref

39. Parker SL, Farber SH, Adogwa O, Rigamonti D, McGirt MJ. Comparison of hospital cost and resource use associated with antibiotic-impregnated versus standard shunt catheters. Clin Neurosurg 2011;58:122-5. Crossref

40. Parker SL, McGirt MJ, Murphy JA, Megerian JT, Stout M, Engelhart L. Cost savings associated with antibiotic-impregnated shunt catheters in the treatment of adult and pediatric hydrocephalus. World Neurosurg 2015;83:382-6. Crossref

41. Cochrane DD, Kestle JR. The influence of surgical operative experience on the duration of first ventriculoperitoneal shunt function and infection. Pediatr Neurosurg 2003;38:295-301. Crossref

42. Duhaime AC, Bonner K, McGowan KL, Schut L, Sutton LN, Plotkin S. Distribution of bacteria in the operating room environment and its relation to ventricular shunt infections: a prospective study. Childs Nerv Syst 1991;7:211-4. Crossref

43. Bayston R, Lari J. A study of the sources of infection in colonised shunts. Dev Med Child Neurol 1974;16 Suppl 32:16-22. Crossref

44. Tulipan N, Cleves MA. Effect of an intraoperative double-gloving strategy on the incidence of cerebrospinal fluid shunt infection. J Neurosurg 2006;104 Suppl:5-8. Crossref

Type 2 diabetes mellitus (T2DM) is one of the most common chronic conditions encountered in primary care, affecting up to 10% of the Hong Kong population.1 It is also a leading cause of morbidity and mortality due to diabetic complications.2 Optimal control of cardiovascular risk factors can decrease the risk of developing diabetes-related complications.3 4 5

Hyperlipidaemia is one of the most important modifiable risk factors for cardiovascular disease (CVD) prevention. Studies have shown that optimal lipid control is associated with an improved cardiovascular outcome.6 7 8 9 Low-density lipoprotein (LDL) particles are considered more atherogenic than other cholesterol components and therefore stringent control of LDL is particularly important for the prevention of CVD in high-risk patients.10

Despite this evidence, lipid control among diabetic patients in the primary care setting, both locally and internationally, has been inadequate.11 The most recent study performed in Hong Kong found that 88.4% of diabetic patients had a suboptimal lipid level.12 Studies in Europe and the US found that the LDL control rate ranged from 30% to 55%.13 14 15 16 17 Similarly, a study of dyslipidaemia management in South Asia including China, South Korea, Malaysia, and Singapore revealed that only 48% of patients attained pre-defined low-density lipoprotein–cholesterol goals.18

Similar to other chronic conditions, the reasons for poor (lipid) control are multifactorial and may include patient, physician, and health care delivery factors. Among them, suboptimal medication augmentation has been identified as an important physician factor. This is known as therapeutic inertia (TI) and is said to exist whenever the health care provider does not initiate or intensify therapy appropriately when therapeutic goals are not reached: “recognition of the problem, but failure to act”.19 20 Such TI has become increasingly acknowledged as a major impediment to CVD risk factor control. Studies have suggested that TI is related to the management of diabetes and hypertension (HT) and may contribute to up to 80% of heart attacks and strokes.21 22

The prevalence of TI in chronic disease management has not been explored in Hong Kong. In this study, we specifically looked at the prevalence of TI in hyperlipidaemia management among diabetic patients. Internal statistical data (internal data from Hospital Authority [HA] Head Office) revealed that lipid control has been relatively poor in this cluster when compared with blood pressure and glycaemic control. Our study aimed to explore the prevalence of TI in the management of hyperlipidaemia among T2DM patients and to explore the underlying factors. By overcoming the barriers to adequate and appropriate treatment, it was expected that the long-term cardiovascular outcome of T2DM patients could be improved.

In this cross-sectional study, all T2DM patients with International Classification of Primary Care code T90 (Non-insulin Dependent Diabetes Mellitus), who had been regularly followed up in all General Outpatient Clinics (GOPCs) of Kowloon Central Cluster (KCC) from 1 October 2011 to 30 September 2013, and had blood lipid levels checked at least once during this period were recruited. In our clinics, blood and urine check-ups are usually carried out in patients with T2DM every 12 to 18 months. This 2-year retrieval period was therefore likely to cover all such patients regularly followed up in our cluster. The diagnosis of diabetes was based on the “Definition and description of diabetes mellitus” from American Diabetes Association (ADA) in 2013.23

The following patients were excluded: patients who had been incorrectly diagnosed with diabetes, type 1 diabetic patients, diabetic patients who had no regular blood or urine check-up during the study period, diabetic patients followed up in a specialist clinic, and patients who died during the study period.

Various studies and guidelines have recommended targets in the treatment of hyperlipidaemia. In the HA of Hong Kong, National Cholesterol Education Program Adult Treatment Panel III Guidelines (NCEP ATP III) and ADA guidelines were used to set up the manual for the risk assessment and management programme. In this study, we used the same set of guidelines to define the level of lipid control in T2DM patients. We focused on the control of LDL as it is the most important risk factor of the lipid profile.

According to NCEP ATP III 200224 and ADA 2013 Guidelines on Diabetes and Lipids,23 target LDL should be <2.6 mmol/L in diabetic patients without overt CVD and <1.8 mmol/L in diabetic patients with overt CVD. In this study, CVD is defined as established ischaemic heart disease (IHD), cerebrovascular accident (CVA), or peripheral vascular disease (PVD).

In this study, lipid control was defined as poor and escalation of treatment indicated if the last LDL level was ≥2.6 mmol/L in diabetic patients without CVD and ≥1.8 mmol/L in diabetic patients with established CVD. Consultation notes of the follow-up immediately after the last available lipid profile test were reviewed through the HA Clinical Management System (CMS). Therapeutic inertia was considered to be present when the attending doctor failed to initiate or intensify treatment if target LDL level was not achieved. If medical notes indicated a valid reason for non-escalation of treatment despite a clinical indication, it was not considered TI. Common justifications included:

(1) Diet and lifestyle modification advice was given to patients newly diagnosed with hyperlipidaemia.

(2) Statin was started following the previous visit and LDL level was improving.

(3) Patient was non-compliant with the existing statin regimen and advice on regular drug compliance was given.

(4) Patient refused to take a statin.

(5) Patient was unable to tolerate side-effects of statin.

(6) Statin was contra-indicated, eg in patients with deranged liver function.

According to the data drawn from Clinical Data Analysis and Reporting System of the HA, a total of 19 662 T2DM patients were attending GOPCs of KCC for regular follow-up with checking of blood lipid profile during the study period. Based on the definitions mentioned above, 9647 of them had suboptimal or poor LDL control. Using the internet sample size calculator (Survey Software from Creative Research System, http://www.surveysystem.com), a sample size of 369 would provide 95% confidence level and 5% margin of error. Thus, 400 patients were sampled to ensure adequate statistical power and allow room for case exclusion. A list of random numbers was then generated from the research randomiser (http://www.randomizer.org/form.htm), from which 400 patients were selected. Details of the visit with latest lipid profile result seen were recorded. Data were derived from the consultation notes in the CMS record of selected patients and recorded on a standard data collection form (Appendix). Data were collected by the principal investigator and counter-checked by another experienced doctor in the research team.

Age and gender of all patients as well as smoking status, body mass index (BMI), latest blood pressure, haemoglobin A1c (HbA1c) level, serum creatinine level, lipid profile, and urine albumin-to-creatinine ratio were retrieved from the CMS. The most recent blood or urine test was used for analysis if more than one test had been performed during the study period. The BMI was calculated as body weight/body height² (kg/m²). The patient was considered a smoker if he/she currently smoked or had stopped in the last 6 months.25 The abbreviated Modification of Diet in Renal Disease formula was used to calculate the estimated glomerular filtration rate.26

The working profile of the attending doctors was retrieved from the Central Office of Department of Family Medicine (FM) and GOPC, KCC. Duration of clinical practice was calculated as the number of years from registration with the Medical Council of Hong Kong. The training status of FM of doctors was documented and categorised according to the following criteria:

All data were entered and analysed using computer software (Windows version 21.0; SPSS Inc, Chicago [IL], US). Student’s t test and analysis of variance were used to analyse continuous variables and the Chi squared test for categorical data. Fisher’s exact test was used if the sample size was less than five. Multivariate stepwise logistic regression was used to determine the association between TI and the significant different variables from patient characteristics and doctor characteristics. All statistical tests were two-sided, and a P value of <0.05 was considered statistically significant.

The study protocol was reviewed and approved by the Research Ethics Committee of HA (Kowloon Central/Kowloon East Cluster) [Reference number: KC/KE-13-0247/ER-1].

A total of 21 960 T2DM patients were identified from the KCC GOPC Diabetes Mellitus registry from 1 October 2011 to 30 September 2013. Among them, 19 662 (89.5%) patients had their lipid profile checked at least once during the study period; 9647 (49.1%) cases had suboptimal lipid control based on the defined criteria above, including 1733 cases with co-existing CVD and 7914 cases without CVD.

Among 400 randomly sampled diabetic patients with suboptimal lipid control, 31 were excluded including 21 who were being followed up in other clinics for diabetic control, nine who died during the study period, and one who was wrongly diagnosed with diabetes. The remaining 369 cases were recruited for data analysis (Fig).

Table 1 summarises the demographic characteristics of the recruited patients. The mean (± standard deviation) age of the study population was 65.5 ± 11.9 years and 186 (50.4%) were female. The mean duration of diabetes was 9.1 ± 7.9 years. With regard to their co-morbidities, 306 (82.9%) patients had concomitant HT, 24 (6.5%) had IHD, 40 (10.8%) had CVA, and two (0.5%) had PVD. The mean LDL level was 3.12 ± 0.61 mmol/L and only 101 (27.4%) patients were prescribed a statin.

Table 2a summarises the demographic characteristics of the attending doctors. A total of 56 doctors, among whom 19 (33.9%) were female, attended the 369 diabetic patients. The mean duration of clinical practice was 13.6 ± 9.6 years. With regard to FM training status, 13 (23.2%) doctors had received no FM training, 18 (32.1%) received basic training or studied DFM, 13 (23.2%) were intermediate FM fellows, and 12 (21.4%) were FM specialists.

Subanalysis of attending doctors’ profile according to their duration of clinical practice and FM training status is shown in Table 2b. Training status of FM varied significantly with duration of clinical practice (P<0.001). Among 13 doctors who had worked for ≤5 years, all had been a basic FM trainee or had obtained a DFM. On the other hand, among 12 doctors who had worked for over 20 years, most (n=9, 75%) had not received any formal FM training.

Among the 369 recruited T2DM patients, treatment was escalated in 47 (12.7%). Justification for not intensifying treatment was provided in 78 (21.1%) cases. Justification was as follows: 19 patients were given dietary advice on lifestyle modifications as they were newly diagnosed with hyperlipidaemia; in 13 patients, a statin had been newly commenced at the previous visit and lipid level was lower compared with pretreatment; five patients were non-compliant with the existing treatment regimen and advice on compliance was given; 28 patients refused to start a statin despite medical advice; six patients had been unable to tolerate side-effects of statin. Statin therapy was contra-indicated in seven patients with impaired liver function. In the remaining 244 cases, TI was present with a prevalence rate of 66.1%.

Table 3 shows the characteristics of physicians in TI-positive and TI-negative patients. The duration of clinical practice of attending doctors was significantly longer in the TI group compared with the non-TI group (P=0.001), with doctors working for over 20 years having a particularly higher rate of TI (82.4%). Doctors without any FM training also had a higher rate of TI (77.7%; P=0.006).

Table 4 summarises the characteristics of T2DM patients in TI-positive and TI-negative groups. Patients in the TI-positive group had a longer duration of diabetes (9.8 ± 8.1 years in TI-positive group vs 7.8 ± 7.4 years in TI-negative group; P=0.024) and lower total cholesterol level and LDL level (both P<0.001). The co-existence of CVD (IHD, CVA, PVD) was more common in the TI-positive group (P=0.003). Other characteristics including patient gender, age, BMI, smoking status, blood pressure, HbA1c level, and type and dose of current statin use were comparable for both groups (all P>0.05).

Based on the results from Tables 3 and 4, multivariate stepwise logistic regression analysis was performed to identify any factors that contributed to TI (Table 5). Only variables that were significantly different in the univariate analysis were included in the regression model. As the FM training status varied significantly with the duration of clinical practice (Table 2b, P<0.001) and these two factors were interrelated, only one of these two variables was included in the logistic regression analysis. As the P value of FM training status (P=0.006) was smaller than that for years of clinical practice (P=0.007) in the univariate analysis (Table 3), FM training status was entered into the logistic regression analysis. Lack of FM training was positively associated with TI (odds ratio [OR]=2.170; P=0.008), whereas patient’s LDL level was inversely associated (OR=0.320; P=0.001).

This was the first clinical analysis of TI in lipid management among T2DM patients managed locally in the primary care setting. It has provided important background information about the prevalence of TI in this group of patients. It also explored possible underlying factors from both the doctor’s and patient’s perspective.

Our study found that lipid control among T2DM patients was far from satisfactory, with 49.1% suboptimally controlled. This is consistent with reports that a high proportion of patients with hyperlipidaemia do not achieve their LDL goal.27 28 It is important to note that TI was present in 66.1% of these cases, meaning that in over 60% of diabetic patients with dyslipidaemia, appropriate management including dietary advice or drug treatment was not provided. This relatively high TI rate should alert primary care physicians to the importance of lipid control among T2DM patients as greater TI leads to poorer clinical outcomes. A similar study carried out by Whitford et al29 has shown that TI was present in 80% of consultations when lipid control was addressed among diabetic patients managed in the primary care setting in Middle East countries. This rate was much higher than the TI in blood pressure control (68%) and glycaemic control (29%). A similar study of lipid management in high-risk patients at a large academic primary care practice in the US has shown that statin dose was augmented at only 16% of over 2000 patient visits where the patient was suboptimally controlled.30 Among the sampled 369 poorly controlled T2DM patients in this study, only 27.4% (n=101) were treated with simvastatin, which is the only statin available in Hong Kong GOPCs. In addition, most (74.3%, 75/101) were treated with a lower dose (5-10 mg daily) that is considered inadequate according to ATP-IV guidelines in which a moderate dose of statin, such as simvastatin 20-40 mg daily, is recommended for T2DM patients in order to achieve target LDL level.31 Thus, the low statin prescription rate and the inadequate dose of statin may together contribute to the suboptimal lipid control among T2DM patients in primary care.

A possible explanation for the TI in dose augmentation of simvastatin is the potential drug-drug interaction with calcium channel blockers (CCB) such as amlodipine.32 The maximum recommended dose for simvastatin in conjunction with amlodipine use is 20 mg/day. Since 306 (82.9%) sampled diabetic patients were found to have concomitant HT and among them 122 (40%) were prescribed amlodipine for blood pressure control, doctors might have hesitated to increase the dose of simvastatin. In our study, 10 diabetic patients in the TI-positive group were prescribed amlodipine and simvastatin 20 mg daily. In this scenario, either changing simvastatin to an alternative statin such as atorvastatin or changing amlodipine to an alternative CCB such as nifedipine is recommended if lipid control remains suboptimal. Failing to switch to another statin or CCB when clinically indicated is also considered to be TI. A more proactive approach to prescribing different drug combinations is required in order to achieve the target LDL in a timely manner.

Further studies of the physician profile relative to the presence of TI have revealed that doctors with longer duration of clinical practice have a higher rate of TI that is even more prominent in those with over 20 years’ clinical practice. These findings are contrary to an overseas study where more experienced doctors had a lower rate of TI33; nonetheless, this study was performed in a secondary care setting and involved cardiologists who managed hyperlipidaemia in patients with IHD. In our study, most doctors who had worked for over 20 years had no formal FM training (9 [75%] of 12 doctors; Table 2b). In addition, when training status was compared, doctors with no FM training had a higher rate of TI than those who had completed FM training (77.7% vs 60.0%; P=0.006). We postulate that doctors who have worked for over 20 years may be less familiar with the latest guidelines on lipid management, possibly due to a lack of FM or related training. If physicians lack appropriate training, there will be gaps in their knowledge of latest clinical management guidelines. This has been confirmed by review articles which showed that TI could be attributed to insufficient knowledge of guidelines.34

When patient’s profile was compared, surprisingly, TI was present in 51 of 62 diabetic patients with overt CVD, and only 11 cases were properly managed (Table 4). This is a considerable concern since lipid control is particularly important and as a secondary prevention strategy in this group of patients. The target for LDL control is much more stringent at <1.8 mmol/L in this group of patients, and more difficult to achieve clinically. Some doctors may not have been aware of this stricter/lower LDL target and have been satisfied with LDL level of 1.8 to 2.6 mmol/L. This is supported by our finding that among diabetic patients with overt CVD whose lipid profile was inadequately controlled (n=62), more than half (n=33, 53.2%) had LDL controlled at 1.8 to 2.6 mmol/L. Physicians should take a more proactive approach particularly in this high-risk group of patients and adhere closely to the prevailing management guidelines in CVD risk factor control.

Multiple variable logistic regression analysis revealed that patients with lower LDL or LDL level closer to normal was associated with TI (Table 5). This could be explained by the threshold effect, that is, the closer the LDL level is to target level, the less likely is the doctor to intensify treatment. This threshold effect has been commonly observed in other similar studies.30 35 Other factors that contribute to the threshold effect could be ‘overestimation of current care’ or ‘complacency with borderline values’, leading to the physician’s subjective misperception that the care provided is sufficient.34

Our study found that TI was common in lipid management among diabetic patients managed in the GOPCs of KCC, with a prevalence of 66.1%. Doctors with a longer duration of clinical practice and who had not received formal FM training had a higher rate of TI. Patients with a closer-to-target LDL were more common in the TI group. Considering that a large volume of diabetic patients are managed in the primary care setting, comprehensive strategies with a more proactive approach should be devised to combat TI so that the cardiovascular outcome of diabetic patients can be improved.

This is the first clinical analysis of TI in lipid management among diabetic patients managed locally in the primary care setting. It has provided important background information about the prevalence of TI in lipid management among diabetic patients and explored the possible underlying factors from both the doctor’s and patient’s perspective. These findings will help improve strategies to overcome TI in lipid control for these patients.

There are some limitations in this study. First, the study was carried out in one single cluster of HA and therefore selection bias might exist. These results from the public primary health care sector might not be applicable to the private sector or secondary care. In addition, the number of doctors with or without FM training was quite discrepant in this study (43 vs 13) and may affect the generalisation of findings. Nevertheless, the present results may lay the groundwork for similar studies in the future, both locally and internationally. Second, patients with diabetes who had not had any blood testing performed during the study period were excluded (n=2298, 10.5% of all diabetic cases). The lipid control status of this group of diabetic patients remained unknown. This might bias the accurate measurement of TI among our target population. Third, only TI in LDL management was explored in this study. Management of hypertriglyceridaemia was not addressed in view of its less-important role as a risk factor for CVD. Future studies exploring the TI in hypertriglyceridaemia management are needed to comprehensively assess lipid control among diabetic patients. Lastly, this study relied heavily on review of consultation notes to identify justification for submaximal therapy and determine presence of TI. Insufficient justification for a certain treatment may have resulted in an overestimation of the prevalence of TI.

This study found that TI was common in the lipid management of diabetic patients managed in GOPCs of KCC, with a prevalence rate of 66.1%. Doctors without FM training and a closer-to-target LDL level among T2DM patients were associated with the presence of TI. Comprehensive strategies should be devised to overcome TI so that the cardiovascular outcome of diabetic patients can be improved.

We are indebted to Ms Katherine Chan, statistical officer of Queen Elizabeth Hospital, for her expert statistical support in the data analysis.

Additional material related to this article can be found on the HKMJ website. Please go to <http://www.hkmj.org>, and search for the article.

All authors have disclosed no conflicts of interest.

1. Chan JC, Malik V, Jia W, et al. Diabetes in Asia: epidemiology, risk factors, and pathophysiology. JAMA 2009;301:2129-40. Crossref

2. Leung GM, Lam KS. Diabetic complications and their implications on health care in Asia. Hong Kong Med J 2000;6:61-8.

3. Colhoun HM, Betteridge DJ, Durrington PN, et al. Primary prevention of cardiovascular disease with atorvastatin in type 2 diabetes in the Collaborative Atorvastatin Diabetes Study (CARDS): multicentre randomised placebo-controlled trial. Lancet 2004;364:685-96. Crossref

4. Collins R, Armitage J, Parish S, Sleigh P, Peto R; Heart Protection Study Collaborative Group. MRC/BHF Heart Protection Study of cholesterol-lowering with simvastatin in 5963 people with diabetes: a randomised placebo-controlled trial. Lancet 2003;361:2005-16. Crossref

5. Gaede P, Vedel P, Larsen N, Jensen GV, Parving HH, Pedersen O. Multifactorial intervention and cardiovascular disease in patients with type 2 diabetes. N Engl J Med 2003;348:383-93. Crossref

6. Waters DD, Ku I. Early statin therapy in acute coronary syndromes: the successful cycle of evidence, guidelines, and implementation. J Am Coll Cardiol 2009;54:1434-7. Crossref

7. Ward S, Lloyd Jones M, Pandor A, et al. A systematic review and economic evaluation of statins for the prevention of coronary events. Health Technol Assess 2007;11:1-160,iii-iv. Crossref

8. O’Regan C, Wu P, Arora P, Perri D, Mills EJ. Statin therapy in stroke prevention: a meta-analysis involving 121,000 patients. Am J Med 2008;121:24-33. Crossref

9. Vrecer M, Turk S, Drinovec J, Mrhar A. Use of statins in primary and secondary prevention of coronary heart disease and ischemic stroke. Meta-analysis of randomized trials. Int J Clin Pharmacol Ther 2003;41:567-77. Crossref

10. Law MR, Wald NJ, Rudnicka AR. Quantifying effect of statins on low density lipoprotein cholesterol, ischaemic heart disease, and stroke: systematic review and meta-analysis. BMJ 2003;326:1423. Crossref

11. Fung CS, Chin WY, Dai DS, et al. Evaluation of the quality of care of a multi-disciplinary risk factor assessment and management programme (RAMP) for diabetic patients. BMC Fam Pract 2012;13:116. Crossref

12. Kung K, Chow KM, Hui EM, et al. Prevalence of complications among Chinese diabetic patients in urban primary care clinics: a cross-sectional study. BMC Fam Pract 2014;15:8. Crossref

13. Kotseva K, Wood D, De Backer G, De Bacquer D, Pyörälä K, Keil U; EUROASPIRE Study Group. Cardiovascular prevention guidelines in daily practice: a comparison of EUROASPIRE I, II, and III surveys in eight European countries. Lancet 2009;373:929-40. Crossref

14. Toth PP, Zarotsky V, Sullivan JM, Laitinen D. Dyslipidemia treatment of patients with diabetes mellitus in a US managed care plan: a retrospective database analysis. Cardiovasc Diabetol 2009;8:26. Crossref

15. Ferrières J, Gousse ET, Fabry C, Hermans MP; French CEPHEUS Investigators. Assessment of lipid-lowering treatment in France—the CEPHEUS study. Arch Cardiovasc Dis 2008;101:557-63. Crossref

16. Kotseva K, Stagmo M, De Bacquer D, De Backer G, Wood D; EUROASPIRE II Study Group. Treatment potential for cholesterol management in patients with coronary heart disease in 15 European countries: findings from the EUROASPIRE II survey. Atherosclerosis 2008;197:710-7. Crossref

17. Santos RD, Waters DD, Tarasenko L, et al. Low- and high-density lipoprotein cholesterol goal attainment in dyslipidemic women: The Lipid Treatment Assessment Project (L-TAP) 2. Am Heart J 2009;158:860-6. Crossref

18. Kim HS, Wu Y, Lin SJ, et al. Current status of cholesterol goal attainment after statin therapy among patients with hypercholesterolemia in Asian countries and region: the Return on Expenditure Achieved for Lipid Therapy in Asia (REALITY-Asia) study. Curr Med Res Opin 2008;24:1951-63. Crossref

19. Phillips LS, Branch WT, Cook CB, et al. Clinical inertia. Ann Intern Med 2001;135:825-34. Crossref

20. Okonofua EC, Simpson KN, Jesri A, Rehman SU, Durkalski VL, Egan BM. Therapeutic inertia is an impediment to achieving the Healthy People 2010 blood pressure control goals. Hypertension 2006;47:345-51. Crossref

21. Andrade SE, Gurwitz JH, Field TS, et al. Hypertension management: the care gap between clinical guidelines and clinical practice. Am J Manag Care 2004;10:481-6.

22. O’Connor PJ, Sperl-Hillen JM, Johnson PE, Rush WA, Biltz G. Clinical inertia and outpatient medical errors. In: Henriksen K, Battles JB, Marks ES, Lewin DI, editors. Advances in patient safety: from research to implementation. Vol 2: Concepts and methodology. Rockville (MD): Agency for Healthcare Research and Quality (US); 2005.

23. American Diabetes Association. Standards of medical care in diabetes—2013. Diabetes Care 2013;36 Suppl 1:S11-66. Crossref

24. National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III). Third Report of the National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III) final report. Circulation 2002;106:3143-421.

25. Centers for Disease Control and Prevention (CDC). Smoking-attributable mortality, years of potential life lost, and productivity losses—United States, 2000-2004. MMWR Morb Mortal Wkly Rep 2008;57:1226-8.

26. Levey AS, Bosch JP, Lewis JB, Greene T, Rogers N, Roth D. A more accurate method to estimate glomerular filtration rate from serum creatinine: a new prediction equation. Modification of Diet in Renal Disease Study Group. Ann Intern Med 1999;130:461-70. Crossref

27. Park JE, Chiang CE, Munawar M, et al. Lipid-lowering treatment in hypercholesterolaemic patients: the CEPHEUS Pan-Asian survey. Eur J Prev Cardiol 2012;19:781-94. Crossref

28. Khoo CM, Tan ML, Wu Y, et al. Prevalence and control of hypercholesterolaemia as defined by NCEP-ATPIII guidelines and predictors of LDL-C goal attainment in a multi-ethnic Asian population. Ann Acad Med Singapore 2013;42:379-87.

29. Whitford DL, Al-Anjawi HA, Al-Baharna MM. Impact of clinical inertia on cardiovascular risk factors in patients with diabetes. Prim Care Diabetes 2014;8:133-8. Crossref

30. Goldberg KC, Melnyk SD, Simel DL. Overcoming inertia: improvement in achieving target low-density lipoprotein cholesterol. Am J Manag Care 2007;13:530-4.

31. Stone NJ, Robinson J, Lichtenstein AH, et al. 2013 ACC/AHA Guideline on the Treatment of Blood Cholesterol to Reduce Atherosclerotic Cardiovascular Risk in Adults, Journal of the American College of Cardiology 2013. Available from: https://www.joslin.org/docs/2013-ACC-AHA-Guideline-Treatment-of-Blood-Cholestero-_to-Reduce-Atherosclerotic-Cardiovascular-Risk-in-Adults.pdf. Accessed Mar 2016.

32. Simvastatin summary of product characteristics. Available from: http://www.medicines.org.uk/emc/medicine/1201/SPC. Accessed Mar 2016.

33. Aujoulat I, Jacquemin P, Rietzschel E, et al. Factors associated with clinical inertia: an integrative review. Adv Med Educ Pract 2014;5:141-7. Crossref

34. Byrnes PD. Why haven’t I changed that? Therapeutic inertia in general practice. Aust Fam Physician 2011;40:24-8.

35. Berlowitz DR, Ash AS, Glickman M, et al. Developing a quality measure for clinical inertia in diabetes care. Health Serv Res 2005;40:1836-53. Crossref

Hong Kong has experienced a tremendous change in lifestyle and the consequent clinical problems pose a challenge to the medical profession. A good example of this is infant feeding. Almost half a century ago, the prevalence of breastfeeding in Hong Kong was at its lowest rate of 5% in 1978 after a dramatic fall from 44% in 1967.1 Following the joint efforts of doctors, nurses and mothers, the prevalence of the ever breastfeeding rate in Hong Kong has rapidly climbed from 20% in 1992 to 60% in 2002 and 86% in 2014.2 The efforts of both the UNICEF Baby-Friendly Hospital Initiative and the Department of Health should be applauded.

Breast milk is the best for babies. Mothers should be encouraged to breastfeed fully for 6 months, followed by introduction of solid foods and continuation of breastfeeding for 2 years or more. Recent data have shown that only 27% of mothers can sustain breastfeeding for 4 to 6 months.2 There are areas where we, as medical professionals, can provide support. For historical reasons, however, not many local doctors and nurses have been trained to manage the clinical problems encountered by breastfeeding mothers. One such problem is breast pain.

Breast pain, which may lead to cessation of breastfeeding, is the most common complaint of lactating mothers seen in a private general paediatric clinic run by a doctor (author) trained in 2000 as an International Lactation Consultant. This study aimed to analyse the reasons for breast pain and how it can be relieved.

Clinical records of lactating mothers who presented with breast pain over a 6-month period (January to June 2015) were retrieved. Patients were self-referred after chatting online with other breastfeeding mothers. During consultation, patients were asked about the history of pain, prior treatment, breastfeeding practices, and their own diet. Breast examination was then performed, including the nipple and areola, to identify any redness or tenderness. In particular, any blockage was identified.

If redness or tenderness was generalised in either or both breasts, it was diagnosed as engorgement (Fig a). If it was confined to a segment, this implied only a lobule was involved. If gentle massage and milk expression provided relief, a blocked duct was diagnosed (Fig b). The ability to express pus (Fig c) or an area of fluctuation or skin thinning (Fig d) was indicative of breast abscess. Mastitis was diagnosed in the presence of fever and tenderness/mass that could not be relieved but had not progressed to an abscess.3 Nipples were examined for cracks (Fig e). Feeding position and latch were checked when appropriate and corrected accordingly. When there was a white spot in the nipple, it was cleared by simple expression or by using a needle to open up the blockage. If there was a shinny reddish colour of the nipple and areola together with burning, stinging, and itchiness then fungal infection was diagnosed. In such case, the baby’s mouth was also examined for the presence of oral thrush.

A total of 69 patients were seen of whom 45 had been breastfeeding for more than 1 month. All except six were in their 30s. The age of the baby was less than 1 month in 24 (35%) women, 1 to 6 months in 27 (39%), and over 6 months in 18 (26%). Only 13 (19%) used complementary infant milk formula.

Breast pain was present for less than 3 days in 35 (51%) women but for longer in the remaining 34 (49%). Pain duration exceeded 7 days in 22 (32%); 15 (22%) of whom had intermittent pain for 14 to 30 days. In 31 (45%) patients, earlier treatment had been received from various sources including Maternal and Child Health Centres (MCHCs), family doctors, general practitioners, obstetricians, doctors at an accident and emergency department, surgeons in a breast surgery clinic, or lactation consultants. Antifungal or antibacterial medication, either local or systemic, was prescribed.

Apart from breast pain, there were other additional complaints: nipple pain in eight (12%) women, sharp needle pain after feeding in eight (12%), white spot at nipple in 15 (22%), and fever in 14 (20%). All had decreased milk production by the affected breast despite frequent feeding or pumping.

The following diagnoses were made: nipple cracks (n=2), poor positioning and latch (n=7), engorgement (n=5), blocked duct (n=35), mastitis (n=13), breast abscess (n=6), and skin infection (n=1). One had all pus drained via the milk duct. Another had pus formed in the sebaceous gland at the areola and was fully drained. The remaining four were referred to a surgeon for further management. Oral antibiotics were prescribed to 21 (30%) women. Fungal infection was suspected in only one woman. Clinical details of four patients chosen for illustration are shown in the Table.

Subjects in this study represent mothers who were very dedicated to breastfeeding. Most had been breastfeeding for more than 1 month and had not given up, despite experiencing pain for quite a number of days.

Blocked duct was the most common cause of breast pain in this study group. Delay in diagnosing and treating a blocked duct can lead to a more serious condition of mastitis and breast abscess.

Engorgement, blocked duct, mastitis, and breast abscess reflect progression from a common original problem of inadequate drainage that can be due to poor positioning and latching, inadequate emptying, or overproduction.3 Obtaining a good history, performing a thorough breast examination, and milk expression can help to make the diagnosis.

Engorgement usually involves the whole breast whereas a blocked duct involves a lobule. In the latter, redness and tenderness are apparent and examination of the areola may reveal a tender swelling representing a blockage of the duct near the opening. Gentle massage and milk expression will relieve the pain and tenderness. A simple blocked duct can be relieved immediately. Nonetheless, when the swelling can only be partially relieved, it may represent tissue inflammation indicative of mastitis. Mothers were encouraged to feed more often on the affected breast. If this failed after one or two feeds, antibiotics were prescribed to prevent progression to breast abscess.

Milk is a very good medium for bacterial culture. Stasis of milk for too long may lead to infection (mastitis) and pus formation (abscess). The common guideline is to relieve a blocked duct as soon as possible, especially in the presence of fever. Once fever has persisted for longer than 24 hours, antibiotics are required. In one woman in this study, however, breast abscess was evident within the first few hours of fever and in another woman without fever, thus fever should be considered a non-specific sign. Clinical assessment was the most important. The ratio of breast abscess to mastitis was higher in this series (46.2%) compared with that reported in the literature (11.1%).3 This may have been due to a difference in sampling methods or different diagnostic criteria for mastitis. The difference between a blocked duct and mastitis can be very subtle. Presence of redness and tenderness in the breast with little effort to clear the blockage may be classified as mastitis. What is of more concern is the possible delay in management that allows untreated mastitis to progress to breast abscess. An abscess can be drained through the duct manually, but needle aspiration under ultrasound guidance or incision may be required in some cases. If there is an incision, the wound must be left open for continuous drainage and the mother may be forced to stop breastfeeding.

Since most of these infections are due to Staphylococcus aureus, Streptococcus, or Escherichia coli, antibiotics chosen should be amoxicillin with clavulanate, cloxacillin, or cefuroxime; all of which are compatible with continuation of breastfeeding.3

Nipple pain may indicate a blocked duct because the duct beneath the areola is swollen. There should be some tenderness although not as much as that of the affected breast lobule. After relief of the blockage, nipple pain will resolve.

A white spot at the nipple may also indicate a blocked duct. Blockage of a lobule and then stasis of milk at the opening of the duct can lead to further blockage by milk that has a high fat or high calcium content. This spot will be white in colour, sometimes referred to as a bleb. It can be removed by milk expression, needle or local application of vegetable oil. This should not be confused with thrush.

Concern has already been raised about the general overdiagnosis of fungal infection as a cause of breast pain, nipple pain, or white spot.4 Patients treated for a presumed ‘yeast infection’ might have shown improvement in symptoms as a result of the anti-inflammatory effect of the antifungal drugs or because the blocked duct resolved on its own. Fungal infection of the breast and nipple may be considered if a blocked duct has been excluded and is often associated with other risk factors. Examples are consuming a diet with high sugar content that promotes growth of fungus, mother having received antibiotics, a maternal history of vaginal candidiasis, or baby’s oral mucosa with thrush.4 All these risk factors were not found in any of the mothers in this study. Excruciating pain after a feed is a non-specific sign. It is more likely to be due to a blocked duct or inadequate emptying of the breast, as shown in this study. These patients had failed to improve after being given local or systemic antifungal treatment in their previous consultations prior to presentation to this clinic. Pain was relieved only after the blocked duct was cleared.

There was a general misunderstanding among the lactating mothers that eating more animal foods could improve milk supply. Previous studies have shown that the protein intake of local lactating mothers is much higher than that of those in other countries. At 3 months postpartum, Hong Kong mothers had a protein intake of 98 g/day5 compared to 81 g/day in the UK6 and 80 g/day in Japan.7 In the first month after delivery (known locally as the confinement period), the protein intake was even higher (133 g/day)5 than at 3 months. During this month, local mothers usually consume a special diet consisting of much more pork, fish, chicken, egg, and milk.

The practice of eating a special diet with additional animal foods during confinement may be unique to Hong Kong Chinese population and is likely a long Chinese tradition. The original rationale was to replenish the blood loss of childbirth and may have been necessary at a time when the general population had barely enough food. Prior to the 1960s, our ancestors usually ate a plant-based diet, with pork available only in the Chinese New Year or during some festivals. There was very little over-nutrition. Time has changed. The diet of adults today is generally high in animal protein8 and fat. Further increase will lead to new clinical problems, not just weight gain in mothers but also increased risk of blockage and inflammation in breastfeeding mothers. A diet that contains much more meat has been shown to be associated with higher inflammatory index scores9 and one of these is C-reactive protein.10

The quantity and quality of fat in breast milk can be affected by the fat in the maternal diet. Lactating mothers in Chongqing (a major city in Southwest China) consumed a diet wherein fat came from lard. Total fat in the breast milk was higher in Chongqing: 38 g/L compared with 32 g/L in Hong Kong.11 Chongqing mothers did not appear to have problems of blocked ducts or mastitis. Thus, a high-fat diet per se may not cause mastitis, it is the quality of fat that matters. Mothers who consume a diet high in saturated fat may be more prone to duct blockage.3 Mothers with a recurrent blocked duct were often advised to change their diet to one with more polyunsaturated fat or use a supplement, lecithin.3 A dietary source of lecithin is mainly soy or eggs. It would appear to be a good practice for lactating mothers in Chongqing to eat lots of eggs. However, in view of the possibility of egg allergy, Hong Kong mothers may be better advised to eat more soy products. Mothers in this study group appeared to eat very few soy products.

High milk production together with inadequate emptying definitely poses a problem. Many Hong Kong mothers took both Chinese remedies (herbs, fish soups) and drank western teas (eg fenugreek) to increase milk production. Nearly all breastfeeding mothers had a breast pump. Some mothers pumped milk more often to produce an excess for later use. Indeed quite a number of the studied mothers had plenty of stored milk in their refrigerator. Working mothers may have stopped pumping during weekends. Such irregular breast emptying may cause the problem of milk stasis. The presence of fatigue, stress, and an imbalanced diet can encourage inflammation that can easily progress to mastitis. Recurrence of blocked duct/mastitis may occur if the mother’s diet and practice of feeding or pumping are not corrected.

A diet rich in white sugar or corn syrup, pastries, and cakes can enhance the growth of fungus but was generally not observed in our subjects. This may explain why fungal infection was rare. A natural well-balanced diet with whole grains, plenty of vegetables and fruits, and no excessive animal products should be recommended. Refined sugary foods, foods with chemicals, colouring agents, and preservatives should be avoided.

A substantial number of nurses had passed the examination that qualified them as an International Lactation Consultant. They worked mainly in the maternity wards of hospitals and MCHCs in Hong Kong. They were very successful in initiating breastfeeding. Some hospitals ran a breastfeeding clinic to support mothers after discharge from the maternity ward but they were not available round the clock. Most mothers with breastfeeding problems attended a MCHC to seek for help. Other mothers chose to see their family doctors. In general, doctors had little training in dealing with problems related to breastfeeding. In 2011, the Department of Health produced a self-learning kit on breastfeeding for any doctor who was interested, but it is difficult for the public to identify such doctors.

Breast pain can sometimes be unbearable. Some patients described it as worse than labour pain. It is unknown from this study how many mothers had stopped breastfeeding because of the pain or how many ended up in hospital with a high fever and abscess that required surgery. Many patients in this study stated that after earlier treatment failed, they had no idea where else to seek further help. Others hesitated to seek medical help because they were afraid they would be told to stop breastfeeding. The mothers in this study were perhaps exceptional. They had tried very hard to find a solution for their pain even though it might have taken a number of days. These mothers deserve a better medical service. The Secretary for Food and Health has stated that the government is very supportive of breastfeeding and is ready to collaborate with health care professional bodies or non-governmental organisations in training personnel and promoting breastfeeding.12 Setting up breastfeeding clinics is the correct approach. These clinics can be run by MCHCs or a Baby-Friendly Hospital and should be full time and open to all. Doctors and lactation consultants can accumulate clinical experience faster and can then act as professional trainers. There is also a need for more local research on the diet and health of lactating mothers, especially those in confinement, so that appropriate education can be delivered to doctors, lactation consultants, midwives, peer counsellors, confinement nannies, and the public.

This study was limited by its retrospective nature. There was a lack of standard protocols for data recording and retrieval. Not all women were followed up to determine if they had completely recovered since it is difficult to do so in a private clinic. What is certain is that those with engorgement and a blocked duct felt immediate relief the moment they left the clinic. It is also quite possible that many cases of breast pain were treated by other doctors and lactation consultants. The data of this study may thus not be representative of Hong Kong in general.

Blocked duct was the most common cause of breast pain in lactating mothers. Without prompt relief it may progress to mastitis/breast abscess or the mother may choose to stop breastfeeding. It may be a suitable time for Hong Kong to set up one or more public full-time breastfeeding clinics in order to provide a better service for lactating mothers and to facilitate professional training and research.

The author has disclosed no conflicts of interest.

1. Baber FM. The current situation in Hong Kong. Hong Kong Pract 1981;5:132-7.

2. Baby-Friendly Hospital Initiative Hong Kong Association. Available from: http://www.babyfriendly.org.hk/en/breastfeeding-in-hk/breastfeeding-trend/. Accessed Feb 2016.

3. Lawrence RA, Lawrence RM. Breast feeding: a guide for the medical profession. 5th ed. St Louis: Mosby; 1999: 273-83.

4. Wilson-Clay B, Hoover K. The breastfeeding atlas. 5th ed. US: LactNews Press; 2013: 57-8.

5. Chan SM, Nelson EA, Leung SS, Cheng JC. Bone mineral density and calcium metabolism of Hong Kong Chinese postpartum women—a 1-y longitudinal study. Eur J Clin Nutr 2005;59:868-76. Crossref

6. Black AE, Wiles SJ, Paul AA. The nutrient intakes of pregnant and lactating mothers of good socio-economic status in Cambridge, UK: some implications for recommended daily allowances of minor nutrients. Br J Nutr 1986;56:59-72. Crossref

7. Takimoto H, Yoshiike N, Katagiri A, Ishida H, Abe S. Nutritional status of pregnant and lactating women in Japan: a comparison with non-pregnant/non-lactating controls in the National Nutrition Survey. J Obstet Gynaecol Res 2003;29:96-103. Crossref

8. Leung SS, Woo J, Ho S, Lam TH, Janus ED. Hong Kong dietary survey. Aust J Nutr Diet 1988;55(Suppl):S11-4.

9. Morimoto Y, Beckford F, Cooney RV, Franke AA, Maskarinec G. Adherence to cancer prevention recommendations and antioxidant and inflammatory status in premenopausal women. Br J Nutr 2015;114:134-43. Crossref

10. Turner-McGrievy GM, Wirth MD, Shivappa N, et al. Randomization to plant-based dietary approaches leads to larger short-term improvements in Dietary Inflammatory Index scores and macronutrient intake compared with diets that contain meat. Nutr Res 2015;35:97-106. Crossref

11. Chen ZY, Kwan KY, Tong KK, Ratnayake WM, Li HQ, Leung SS. Breast milk fatty acid composition: a comparative study between Hong Kong and Chongqing Chinese. Lipids 1997;32:1061-7. Crossref

12. Hong Kong Paediatric Society, Hong Kong Paediatric Foundation. Summit on Breastfeeding and Early Childhood Nutrition in the First 1000 Days 2015; Abstract: 18.

With ageing of the world’s population, the prevalence of dementia increases: 46.8 million people worldwide were living with dementia in 2015. This is projected to reach 74.7 million in 2030 and 131.5 million in 2050, with 60% suffering from Alzheimer’s disease (AD).1 In Hong Kong, the prevalence of mild dementia has been reported to be 8.9% for adults aged 70 years or over, with 64.6% suffering from AD.2 Appropriate management of demented patients begins with correct diagnosis of dementia subtype that allows earlier implementation of disease-specific treatment. In particular, cholinesterase inhibitors (ChEIs) or N-methyl-D-aspartate receptor antagonists are mostly suitable for the treatment of AD. The current clinical diagnostic guidelines for various types of dementia have limited sensitivities and specificities, however. The sensitivity and specificity of clinical diagnostic criteria for AD, dementia with Lewy bodies (DLB), and frontotemporal dementia (FTD) have been reported as 81% and 70%, 50% and 80%, 85% and 95%, respectively.3 4 5 6 In the most recent diagnostic criteria for AD, additional use of biomarkers of AD has been recommended by the National Institute on Aging and Alzheimer’s Association to improve the accuracy of AD diagnosis.3 Biomarkers for the diagnosis of AD include cerebrospinal fluid (CSF), amyloid pathological imaging (eg carbon 11–labelled Pittsburgh compound B [¹¹C-PIB] positron emission tomography [PET]), and functional imaging (eg [¹⁸F]-2-fluoro-2-deoxy-D-glucose [¹⁸F-FDG] PET) that yield sensitivities and specificities of at least 90% and 85%, respectively in the diagnosis of AD, DLB, and FTD.3 7 8 9 10 11 Because of the invasive nature of lumbar puncture in the collection of CSF, neuroimaging modalities such as ¹⁸F-FDG PET and ¹¹C-PIB PET are more accepted in routine clinical practice to improve the diagnosis of dementia subtype.

The most common functional neuroimaging is with ¹⁸F-FDG12 and the most common pathological neuroimaging is with ¹¹C-PIB.13 These molecular imaging markers are imaged using PET. The ¹⁸F-FDG measures metabolic activity of the brain; ¹⁸F-FDG PET distinguishes well between AD and non-AD dementia.11 In a systematic review, the sensitivity and specificity for ¹⁸F-FDG PET in distinguishing between AD and DLB was 83%-99% and 71%-93%, respectively; and the sensitivity and specificity for ¹⁸F-FDG PET in distinguishing between AD and FTD was 97.6%-99% and 65%-86%, respectively.11 In the same systematic review, ¹⁸F-FDG PET predicted patients with mild cognitive impairment (MCI) deteriorating into dementia with sensitivity and specificity of 81%-82% and 86%-90%, respectively.11 Besides, ¹¹C-PIB can detect the presence of fibrillar amyloid plaques that are a neuropathological marker of AD.13 Correlation studies with neuropathology have shown a sensitivity of 90% and specificity of 100%; ¹¹C-PIB can reasonably distinguish AD from other types of dementia, eg FTD.13 Using neuropathology as the gold standard, the sensitivity and specificity was 89% and 83%, respectively.13 The presence of ¹¹C-PIB retention also predicts the progression of patients with MCI: 50% progress to AD in 1 year and 80% progress to AD within 3 years.14

Previous studies with ¹⁸F-FDG and ¹¹C-PIB PET have focused on highly selected diagnostic groups, and only a few studies have studied their impact in the routine clinical setting of a memory clinic at a tertiary university hospital. The latter are referral centres, and often encounter patients with complicated diagnostic issues. Ossenkoppele et al15 reported a cohort of 145 patients who underwent ¹⁸F-FDG and ¹¹C-PIB PET after clinical assessment. Change in clinical diagnosis was required in 23% with the diagnostic confidence increased from a mean of 71% to 87%. Diagnosis remained unchanged in 96% after PET over the next 2 years.15 In seven patients with MCI and positive amyloid deposition on ¹¹C-PIB PET, six progressed to AD during follow-up (5 had AD pattern of hypometabolism on ¹⁸F-FDG PET).15 In a retrospective study of 94 patients with MCI or dementia, Laforce et al16 showed that ¹⁸F-FDG PET brain scan led to a change in diagnosis in 29% of patients, and reduced the frequency of atypical or unclear diagnoses from 39.4% to 16%.

To the best of our knowledge, there are no published data on the impact of molecular neuroimaging on accuracy of diagnosis of AD or other dementias in the Chinese population. We hypothesised that brain ¹⁸F-FDG with or without ¹¹C-PIB PET imaging can improve the accuracy of diagnosis of common dementia subtypes in a memory clinic. The objective of this study was to investigate the impact of brain ¹⁸F-FDG with or without ¹¹C-PIB imaging in improving the accuracy of diagnosis of dementia subtype in a local memory clinic in Hong Kong.

This was a retrospective study conducted at the Memory Clinic of Queen Mary Hospital, the University of Hong Kong. Patients were referred by general practitioners, neurologists, geriatricians, surgeons, or psychiatrists. All patient records between January 2007 and December 2014 were reviewed. Inclusion criteria were a clinical diagnosis of MCI, dementia of any type, or unclassifiable dementia; and ¹⁸F-FDG with or without ¹¹C-PIB PET performed within 3 months after the initial clinical diagnosis. The initial clinical assessment was performed by a geriatrician experienced in dementia care and included detailed history taking from primary carers of the patient, physical examination, cognitive assessment, and laboratory studies (including thyroid function test, vitamin B₁₂ level, folate level, and syphilis serology [Venereal Disease Research Laboratory]). Clinical criteria for AD, FTD, DLB, and vascular dementia (VaD) were employed to establish the clinical diagnosis initially, without using any biomarker. The diagnosis of different dementia subtype before neuroimaging was based on the respective diagnostic guidelines. Patients with AD were diagnosed according to the NINCDS-ADRDA (National Institute of Neurological and Communicative Disorders and Stroke and Alzheimer’s Disease and Related Disorders Association) diagnostic criteria.17 Patients with DLB were diagnosed by the McKeith criteria.4 Behavioural variant (bv) of FTD was diagnosed by revised diagnostic criteria reported by the International bvFTD Criteria Consortium5 and language variant of FTD was diagnosed by latest published criteria.6 Patients with VaD were diagnosed according to the criteria of the NINDS-AIREN (National Institute of Neurological Disorders and Stroke/Association Internationale pour la Recherche et l’Enseignement en Neurosciences).18 In this study, we reviewed the medical records of eligible subjects and collected data including age, sex, education level, Mini-Mental State Examination score, Clinical Dementia Rating scale score, molecular imaging report including the standardised uptake value ratio (SUVR) of ¹¹C-PIB PET, and the pre- and post-imaging diagnoses. For patients who were diagnosed with MCI, their progression during subsequent follow-up visits was also reviewed.

The need for ¹⁸F-FDG with or without ¹¹C-PIB PET was determined by the geriatrician who performed the initial clinical assessment. The images were evaluated by a radiologist with more than 10 years of experience in reading PET scans. Dementias were classified using the generally accepted criteria. Patients were fasted for at least 4 hours before the PET. The serum glucose level was measured in all patients. For ¹⁸F-FDG PET, the patient was rested in a dimly lit room with eyes closed for 30 minutes prior to injection of ¹⁸F-FDG via a venous catheter. Another 30 minutes of rest was observed before starting the acquisition. The acquired data were semi-quantitatively compared with age-stratified normal controls using three-dimensional stereotactic surface projections. For PIB imaging, acquisition was performed at 5 minutes and 35 minutes after ¹¹C-PIB injection via a venous catheter, and SUVR images of ¹¹C-PIB between 5 and 35 minutes were generated. Cerebellar grey matter was chosen as reference tissue. In this study, ¹¹C-PIB PET scans were rated as positive (PIB+; if binding occurred in more than one cortical brain region; ie frontal, parietal, temporal, or occipital) or negative (PIB–; if predominantly white matter binding).

The pattern of ¹⁸F-FDG PET hypometabolism that is suggestive of each subtype of dementia is as follows6 12 19:

(1) AD—uni- or bi-lateral parietotemporal hypometabolism with posterior cingulate gyrus involvement or bilateral parietal and precuneal hypometabolism.

(2) DLB—same as AD with added hypometabolism in occipital lobes.

(3) bvFTD—uni- or bi-lateral frontotemporal hypometabolism with or without less-severe parietal hypometabolism.

(4) Semantic dementia—anterior temporal lobe hypometabolism.

(5) Progressive non-fluent aphasia—left posterior frontoinsular hypometabolism.

(6) VaD—well-defined focal defects not fitting the above described patterns.

Descriptive statistics were used for data analyses. Continuous variables were expressed as mean ± standard deviation or median (interquartile range) as appropriate. Categorical data were expressed as number and percentages. The agreement between pre- or post-imaging diagnoses of dementia subtype was analysed by the Cohen’s kappa (κ) statistic. The Cohen’s κ reflected the degree of agreement: <0 = no agreement, 0-0.20 = slight agreement, 0.21-0.40 = fair agreement, 0.41-0.60 = moderate agreement, 0.61-0.80 = substantial agreement, and 0.81-1.00 = almost perfect agreement. All analyses were performed with the Statistical Package for the Social Sciences (Windows version 18.0; SPSS Inc, Chicago [IL], US).

A total of 109 patients (56.9% were female) were recruited of whom 102 had dementia and seven had MCI. Both ¹⁸F-FDG and ¹¹C-PIB PET data were available for 45 (41.3%) patients, and 64 patients underwent ¹⁸F-FDG only. The final diagnosis of the 102 demented patients after neuroimaging is shown in Table 1.

The accuracy of clinical diagnoses is summarised in Table 2. Overall, PET scans confirmed the clinical impression in 63.7% of patients, and corrected the diagnosis in 36.3%. Using the result of PET scan as the gold standard, the frequency of accurate initial clinical diagnosis was low for FTD, VaD, and mixed dementia (14.3%, 28.6%, and 0%, respectively). The accuracy of clinical diagnosis for AD and DLB was 81.5% and 44.4%, respectively. After excluding subjects with an initial MCI diagnosis, the agreement between the initial and final post-imaging dementia diagnosis was only fair, with a Cohen’s κ of 0.25 (95% confidence interval, 0.05-0.45).

Table 3 lists the diagnosis of subjects before and after the availability of ¹⁸F-FDG with or without ¹¹C-PIB PET neuroimaging. For subjects with a final diagnosis of AD (n=65), 18.5% (n=12) were initially diagnosed with non-AD dementia (including 3 with DLB, 2 with FTD, 4 with VaD, and 3 with other dementia) and subsequently received symptomatic AD therapy (ie ChEIs and/or memantine). For the 21 subjects who underwent PIB PET imaging, 19% (n=4) of those with AD (PIB+) were initially diagnosed with non-AD dementia. For subjects with an initial diagnosis of AD (n=74), 28.4% (n=21) had a change in diagnosis (including 4 DLB, 6 FTD, 4 VaD, 3 mixed AD plus VaD, and 4 with other dementia). Excluding subjects with DLB and mixed AD plus VaD, 13.7% of all subjects (14 out of 102) had discontinued their previous symptomatic AD therapy. For subjects with a final diagnosis of FTD (n=7), 85.7% (n=6) were initially misdiagnosed as AD. For subjects with a final diagnosis of DLB (n=9), 44.4% (n=4) were misdiagnosed as AD.

Five patients were diagnosed with unclassifiable dementia following neuroimaging, which comprised four females and one male with a mean age of 78 ± 9.4 years. All presented with amnesia. In addition, one patient presented with apraxia and dysexecutive syndrome and another presented with hyperorality. All of them were PIB^-. An AD pattern of hypometabolism was present in four patients (2 with hypometabolism in posterior cingulate gyrus and 2 with hypometabolism in temporoparietal lobes). Isolated hypometabolism in the temporal lobes was present in one patient.

The clinical information of the seven amnesic MCI subjects are summarised in Table 4. None of the three subjects without imaging risk factors for AD deteriorated over a follow-up period of 1 to 5 years. Of the four amnesic MCI subjects with imaging risk factors, two deteriorated into AD over a follow-up period of 5 years.

In this study, we showed that ¹⁸F-FDG with or without ¹¹C-PIB PET clarified and improved the accuracy of dementia diagnosis in 36.3% of patients, and confirmed the initial diagnosis in 63.7%. Using the results of PET scan as the gold standard, the accuracy of clinical diagnosis was low for FTD, VaD, and mixed dementia collectively. On the one hand, 11.7% of patients (ie 12 out of 102) were started on symptomatic AD therapy after the ¹⁸F-FDG with or without ¹¹C-PIB PET neuroimaging investigations. On the other hand, 13.7% of patients (ie 14 out of 102) discontinued symptomatic AD therapy after ¹⁸F-FDG with or without ¹¹C-PIB PET because they did not have AD.

We also showed that the accuracy of clinical diagnosis of DLB and FTD was low (44.4% and 14.3%, respectively). This finding was in agreement with a previous study.20 Both DLB and FTD are commonly misdiagnosed clinically as AD (50% for DLB and 85.7% for FTD).20 We have previously reported that 100% of our patients with biomarkers that confirmed DLB and FTD presented with memory impairment in our memory clinic.20 A previous study also reported that 26% of DLB patients were initially misdiagnosed with AD, and 57% of these DLB patients presented with memory impairment.21 We understand that an accurate diagnosis of DLB is very important for subsequent management. Patients with DLB are particularly sensitive to neuroleptics.21 Neuroleptic sensitivity can present as drowsiness, confusion, abrupt worsening of parkinsonism, postural hypotension, or neuroleptic malignant syndrome.21 Other clinical features of DLB that need to be observed and tackled include well-formed visual hallucinations, rapid eye movement sleep behavioural disorder, and autonomic symptoms (including postural hypotension, sialorrhoea, and urinary and bowel symptoms).21 By accurately establishing the diagnosis of DLB, careful observation of classic DLB symptoms may reduce unnecessary investigations. Regarding therapeutic implications, DLB is characterised by far greater cholinergic deficits than AD. Hence, most DLB patients will benefit from ChEIs, and the extent of symptomatic improvement should be monitored after such therapy.22

Similarly, FTD may be misdiagnosed as AD. The former can also present initially with memory impairment, as illustrated by our FTD patients. There is increasing evidence that elderly patients with FTD often present with memory impairment.5 23 24 In one autopsy study, 64% (n=7) of 11 elderly patients with FTD had anterograde memory loss.23 Current treatment guidelines do not advise giving ChEIs or memantine treatments to FTD patients. Thus, such medications should be stopped to prevent unnecessary adverse effects.25

In the past few years, disease-modifying treatments (eg bapineuzumab) have failed to demonstrate their efficacy in clinical trials with AD patients.26 Detailed post-hoc analyses with AD biomarkers have shown the problem of diagnosing AD in subjects recruited in these studies. Only approximately 80% of these subjects had AD amyloid pathology, according to the presence of amyloid PET scan.26 Thus, including ¹¹C-PIB PET to confirm brain amyloid in study inclusion criteria can help ensure recruitment of genuine AD patients to future clinical trials of disease-modifying treatments for AD.27 Given the minimally invasive nature of ¹¹C-PIB PET compared with CSF amyloid-beta (Aβ) 42 measurements,7 it is likely to be a more acceptable choice for patients in clinical trials. At present, there are ongoing clinical trials of AD treatments including secretase inhibitors, Aβ aggregation inhibitors, Aβ and tau immunotherapy.27 We believe that ¹¹C-PIB PET will play an important role in these clinical trials.

It is considered that ¹⁸F-FDG and ¹¹C-PIB PET may detect underlying AD in patients with MCI.28 In the present study, 50% of MCI patients (ie 2 out of 4) with ¹⁸F-FDG and ¹¹C-PIB PET imaging findings positive for AD showed deterioration over a follow-up period of 5 years. Although recommending PET brain imaging in MCI patients is still debatable, we believe that this investigation can help clinicians to better plan future and long-term treatments. In particular, disease-modifying drugs for AD or MCI due to AD may prove to be effective in the coming decade. Finally, in the present study, five patients were diagnosed with unclassifiable dementia. In the four patients with an AD pattern of hypometabolism, AD may still be present as they may have diffuse plaques or amorphous plaques that do not bind well to PIB. Alternatively they may have another type of dementia that requires pathological confirmation, eg argyrophilic grain disease or neurofibrillary tangle–only dementia.29 We will follow up the remaining patient with isolated hypometabolism in the temporal lobes to see whether additional FTD features develop.

There were several limitations to the present study. This was a retrospective case series and as such we were unable to collect further information such as the pre-imaging or post-imaging confidence of diagnosis. The diagnosis of dementia relied on the clinical diagnostic criteria without pathological confirmation. Therefore, we were also unable to compare the relative accuracy of clinical diagnosis and PET diagnosis with pathological diagnosis. For patients with MCI, some were not followed up for sufficiently long to ascertain whether or not they had deteriorated and developed dementia. Structural imaging (including computed tomography or magnetic resonance imaging) of the brain was not analysed as a separate variable but integrated into the pre-functional imaging clinical diagnoses of dementia subtypes. Our case series is likely to have selection bias as PET imaging is mostly a self-paid service in Hong Kong. The exception is for patients who are retired civil servants or recipients of Comprehensive Social Security Assistance. Demented patients who could not afford PET may differ to the patients selected. Although the PET images were analysed and read by radiologists experienced in PET, the interpretations depended heavily on individual experience and training; also, radiologists were not blinded to clinical information written on the request form. Despite these limitations, our study should be more reflective of day-to-day practice in a memory clinic and how ¹⁸F-FDG with or without¹¹C-PIB PET imaging may assist clinical diagnosis.

In this study, ¹⁸F-FDG with or without ¹¹C-PIB brain imaging improved the accuracy of diagnosis of dementia subtype in 36% of patients with underlying AD, DLB, VaD, and FTD who presented to our memory clinic.

All authors have disclosed no conflicts of interest.

1. Alzheimer’s Disease International World Alzheimer Report 2015: executive summary. Available from: http://www.alz.co.uk/research/WorldAlzheimerReport2015-sheet.pdf. Accessed Sep 2015.

2. Lam LC, Tam CW, Lui VW, et al. Prevalence of very mild and mild dementia in community-dwelling older Chinese people in Hong Kong. Int Psychogeriatr 2008;20:135-48. Crossref

3. McKhann GM, Knopman DS, Chertkow H, et al. The diagnosis of dementia due to Alzheimer’s disease: recommendations from the National Institute on Aging and Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 2011;7:263-9. Crossref

4. McKeith IG, Dickson DW, Lowe J, et al. Diagnosis and management of dementia with Lewy bodies: third report of the DLB Consortium. Neurology 2005;65:1863-72. Crossref

5. Rascovsky K, Hodges JR, Knopman D, et al. Sensitivity of revised diagnostic criteria for the behavioural variant of frontotemporal dementia. Brain 2011;134:2456-77. Crossref

6. Harris JM, Gall C, Thompson JC, et al. Classification and pathology of primary progressive aphasia. Neurology 2013;81:1832-9. Crossref

7. Shea YF, Chu LW, Zhou L, et al. Cerebrospinal fluid biomarkers of Alzheimer’s disease in Chinese patients: a pilot study. Am J Alzheimers Dis Other Demen 2013;28:769-75. Crossref

8. Duits FH, Teunissen CE, Bouwman FH, et al. The cerebrospinal fluid “Alzheimer profile”: easily said, but what does it mean? Alzheimers Dement 2014;10:713-723.e2. Crossref

9. Sinha N, Firbank M, O’Brien JT. Biomarkers in dementia with Lewy bodies: a review. Int J Geriatr Psychiatry 2012;27:443-53. Crossref

10. Harris JM, Gall C, Thompson JC, et al. Sensitivity and specificity of FTDC criteria for behavioral variant frontotemporal dementia. Neurology 2013;80:1881-7. Crossref

11. Davison CM, O’Brien JT. A comparison of FDG-PET and blood flow SPECT in the diagnosis of neurodegenerative dementias: a systematic review. Int J Geriatr Psychiatry 2014;29:551-61. Crossref

12. Schöll M, Damián A, Engler H. Fluorodeoxyglucose PET in neurology and psychiatry. PET Clin 2014;9:371-90. Crossref

13. Vandenberghe R, Adamczuk K, Dupont P, Laere KV, Chételat G. Amyloid PET in clinical practice: Its place in the multidimensional space of Alzheimer’s disease. Neuroimage Clin 2013;2:497-511. Crossref

14. Cummings JL. Biomarkers in Alzheimer’s disease drug development. Alzheimers Dement 2011;7:e13-44. Crossref

15. Ossenkoppele R, Prins ND, Pijnenburg YA, et al. Impact of molecular imaging on the diagnostic process in a memory clinic. Alzheimers Dement 2013;9:414-21. Crossref

16. Laforce R Jr, Buteau JP, Paquet N, Verret L, Houde M, Bouchard RW. The value of PET in mild cognitive impairment, typical and atypical/unclear dementias: A retrospective memory clinic study. Am J Alzheimers Dis Other Demen 2010;25:324-32. Crossref

17. McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan EM. Clinical diagnosis of Alzheimer’s disease: report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer’s Disease. Neurology 1984;34:939-44. Crossref

18. Román GC, Tatemichi TK, Erkinjuntti T, et al. Vascular dementia: diagnostic criteria for research studies. Report of the NINDS-AIREN International Workshop. Neurology 1993;43:250-60. Crossref

19. Waldö ML. The frontotemporal dementias. Psychiatr Clin North Am 2015;38:193-209. Crossref

20. Shea YF, Ha J, Chu LW. Comparisons of clinical symptoms in biomarker-confirmed Alzheimer’s disease, dementia with Lewy bodies, and frontotemporal dementia patients in a local memory clinic. Psychogeriatrics 2014;15:235-41. Crossref

21. Zweig YR, Galvin JE. Lewy body dementia: the impact on patients and caregivers. Alzheimers Res Ther 2014;6:21. Crossref

22. Gauthier S. Pharmacotherapy of Parkinson disease dementia and Lewy body dementia. Front Neurol Neurosci 2009;24:135-9. Crossref

23. Baborie A, Griffiths TD, Jaros E, et al. Frontotemporal dementia in elderly individuals. Arch Neurol 2012;69:1052-60. Crossref

24. Hornberger M, Piguet O. Episodic memory in frontotemporal dementia: a critical review. Brain 2012;135:678-92. Crossref

25. Portugal Mda G, Marinho V, Laks J. Pharmacological treatment of frontotemporal lobar degeneration: systematic review. Rev Bras Psiquiatr 2011;33:81-90. Crossref

26. Blennow K, Mattsson N, Schöll M, Hansson O, Zetterberg H. Amyloid biomarkers in Alzheimer’s disease. Trends Pharmacol Sci 2015;36:297-309. Crossref

27. Wisniewski T, Goñi F. Immunotherapeutic approaches for Alzheimer’s disease. Neuron 2015;85:1162-76. Crossref

28. Langa KM, Levine DA. The diagnosis and management of mild cognitive impairment: a clinical review. JAMA 2014;312:2551-61. Crossref

29. Kovacs GG. Tauopathies. In: Kovacs GG, editor. Neuropathology of neurodegenerative diseases: a practical guide. Cambridge: Cambridge University Press; 2015: 125-8.