The crude rate of caesarean section (CS) deliveries is considered an important global indicator when measuring access to obstetric care.1 Previous ecological analysis in primitive lower-income countries revealed that with the introduction of safe CS deliveries, small increases in CS rates, if performed in women with a medical indication, could dramatically reduce maternal and newborn mortality.1 2 3 On the other hand, CS rates in developed countries have risen steeply since the 1970s and 1980s4 without any obvious evidence of significant improvements in pregnancy outcome.1 5 High CS rates have since been an issue of international public health concern. In 1985, the World Health Organization (WHO) stated that there was no justification for any region to have a CS rate higher than 10% to 15%.6 There is a lack of scientific evidence of any substantial maternal or perinatal benefit from increasing CS rates and some studies contrarily have shown that higher rates may be associated with negative consequences to maternal and child health.7 8 Despite this, CS rates have continued to increase worldwide in middle- and high-income countries. The WHO-recommended upper limit of 15% has been grossly exceeded by most centres in developed countries over the last two decades.9

The lack of a standardised classification system to facilitate monitoring and comparison of CS rates in a consistent and action-oriented manner is one of the factors that makes changes to CS trends difficult to understand.10 Previous discussions often focused on total CS rates and did not yield information about the underlying reasons. The Robson’s Ten-group Classification System is one of the best methods that fulfils current international and institutional needs to monitor and analyse CS rates.11 The classification system divides women into 10 groups based on basic epidemiological and obstetric characteristics, including parity, previous uterine scar, preterm (<37 weeks) or term delivery, fetal presentation, singleton or multiple pregnancy, or whether labour is spontaneous or induced. The actual indication for CS is not needed for such categorisation. As the groups are totally inclusive and mutually exclusive, the classification system can be applied prospectively. All women who present to the labour ward for delivery can be promptly classified based on these readily available parameters. Specifically, these categories are11:
(1) Primiparous women with a single cephalic pregnancy, ≥37 weeks’ gestation, in spontaneous labour;
(2) Primiparous women with a single cephalic pregnancy, ≥37 weeks’ gestation, who have induction of labour or CS prior to labour onset;
(3) Multiparous women without a previous uterine scar, with a single cephalic pregnancy of ≥37 weeks’ gestation in spontaneous labour;
(4) Multiparous women without a previous uterine scar, with a single cephalic pregnancy of ≥37 weeks’ gestation, with induction of labour or CS prior to labour onset;
(5) Multiparous women with one or more previous uterine scar(s) and a single cephalic pregnancy of ≥37 weeks’ gestation;
(6) Primiparous women with a single breech pregnancy;
(7) Multiparous women with a single breech pregnancy, with/without previous uterine scar(s);
(8) Women with multiple pregnancies with/without previous uterine scar(s);
(9) Women with a single pregnancy with a transverse or oblique lie, with/without previous uterine scar(s); and
(10) Women with a single cephalic pregnancy at ≤36 weeks’ gestation.

Despite this increase in CS rates over the years in both public as well as private sectors, there has been a paucity of formal surveys on trends in CS rates in Hong Kong. This study attempted to analyse the secular trends in CS rates over 20 years at a single public tertiary training obstetric unit serving a stable population of around 0.7 to 1 million in the Kowloon East area. Applying the Robson’s classification to the data should allow identification of the subgroup(s) that are predominantly contributing to the steady increase in overall CS rate. The results of this study should determine whether the increase in CS rates is genuinely due to changes in patient epidemiology and risk factors or merely to changes in obstetric management.

The obstetric data from a single obstetric unit (United Christian Hospital, Kwun Tong) for the last 20 years (1995-2014) were retrieved from the Hospital Authority (HA) Obstetrics Clinical Information System. The annual data were supplied to the unit in an anonymous format with only secondary identifiers such as medical record number and hospital number. After compiling this 20-year database, basic patient characteristics that could constitute important epidemiological risk factors, such as the proportion with advanced maternal age of >35 years, percentage with previous CS or other uterine scars, induction of labour and multiple pregnancies, were calculated over the study period.

All cases that underwent CS in our unit during the study period were classified into one of the 10 groups according to the Robson’s classification,10 11 using prior characteristics or risk factors before delivery, including primiparous versus multiparous, preterm versus term, induction of labour versus spontaneous labour, cephalic presentation versus breech or other non-cephalic presentation, singleton versus multiple pregnancy, and previous uterine scar versus no previous scars. The CS rate of each of the 10 subgroups was then calculated for each year, and the trends and changes in the rate over the 20 years were examined. The total number of patients in each category was then stratified into four 5-year intervals to compare the four periods using 4 x 2 contingency tables and Mantel-Haenszel Chi squared tests for linear trends for each category. A P value of <0.05 was considered statistically significant. Significant trends identified in each category were then compared with observable trends in patient epidemiological factors over the same period of time. This study was approved by the Kowloon Central/Kowloon East Ethics Committee Board.

There were 86 262 births and 17 140 CSs from January 1995 to December 2014. The annual number of deliveries over the 20 years ranged from 3350 in 1995 to 5648 in 2011. The overall CS rate increased modestly from 15.4% in 1996 to 24.6% in 2014. Parallel with the gradual increase in overall CS rate, the proportion of elective CS compared with emergency CS also gradually increased from 25%-30% for 1995-2000 to 40%-45% for 2010-2014, indicating that an increasing number of CS were performed electively and the decision was made well ahead of labour, rather than as an emergency in the intrapartum period. There were significant increases in the proportion of women with previous CS (lowest 5% in 2000 to 16.2% in 2014), advanced maternal age of >35 years (lowest 13.2% in 1995 to 24.5% in 2014), induction of labour (lowest 8.5% in 2006 to 15.9% in 2014), and multiple pregnancies (1.1% in 1996 to 3.6% in 2014) during the study period (Fig 1). The crude perinatal mortality rate also fluctuated between 2.6 and 5.3 per 1000 deliveries; the adjusted perinatal mortality rate (excluding those major congenital malformations and birth weight of <750 g) also varied with an excursion of 1.9 and 3.5 per 1000 deliveries. Due to the small number variations with absolute crude perinatal mortality ranging between 10 and 27 per year, however, no obvious trends were identified during the study. The maternal mortality rate was lower than 5 per 100 000 pregnancies throughout the two decades with many years recorded as zero, so no trends could be observed due to the small variations.

Comparison of the trend in Robson’s categories 1 to 4 for primiparous and multiparous women with term spontaneous labour (category 1 and 3) or induced labour or elective CS (category 2 and 4) revealed that the group of primiparous women with term spontaneous labour (category 1) had a consistent and gradual fall in CS rate from 14.4% to 10.8%, while the group of multiparous women with term spontaneous labour (category 3) also had a slight fall from 2.1% to 1.6%. The other categories remained quite stable (Fig 2). On the contrary, obvious trends showing a dramatic increase in CS rate were observed in those with previous CS (category 5, from 29% to 61%), breech presentation at delivery (category 6 and 7, primiparous from 72% to 97% and multiparous from 69% to 96%), and multiple pregnancy (category 8, from 35% to 86%). Although the CS rate for abnormal lie or malpresentation other than breech (category 9) approached 100% throughout the period and therefore displayed no significant trend, a subtle increase in CS rate was seen in those with preterm labour (category 10, from 17% to 25%) [Fig 2]. The data were then stratified into 5-year intervals and the CS rate for each category compared using a 2 x 4 contingency table. The above observed trends were confirmed to be statistically significantly different with P<0.05 for categories 1, 5, 6, 7, 8 and 10 (Table).

The data were then reorganised to show the percentage contribution of each Robson’s category to the total CS rate for each 5-year interval (Fig 3a). As there were wide differences in the absolute number of women in these categories, it could be seen that the contribution of categories 1, 2, 5, 6 and 8 tended to overwhelm the contribution of other categories. Thus, despite the modest fall in CS rates in category 1 over the four interims from approximately 14% to 11%, the impact on reducing the overall CS rates was predominant over other categories, amounting to almost 10% of all CS recorded. This effect, however, was counterbalanced by the contribution of categories 5 and 8 that increased the CS rates, and to the accumulated effects on increasing CS rates by other categories (Fig 3b), so that the net balance was an overall rise in CS rates from 15% to 24% within the study period.

The data presented above revealed a gradual increase in overall CS rate of approximately 10% over the 20-year study period. While there was a significant reduction in the primary CS rate in low-risk primiparous women with spontaneous labour, this was counterbalanced by the ever-increasing CS rate in those with previous CS, breech presentation, multiple pregnancies, and to a lesser extent those with preterm labour. The data have demonstrated the advantages of using the Robson’s classification to analyse factors that will influence the overall CS rate.

The use of the Robson’s classification is increasing rapidly and spontaneously worldwide. Despite some limitations, the 10-group classification is easy to implement and interpret.11 12 It allows standardised comparisons of data across countries and time points, and identifies the subpopulations that drive changes in CS rates. The 10-group classification was easily applied to different levels of analysis from single-centre to multi-country datasets without problems of inconsistencies or misclassification,12 13 14 15 16 enabling specific groups of women to be clearly identified as the main contributors to the overall CS rate. Indeed, it has been demonstrated that this classification can help health care providers plan practical and effective care that targets specific groups of women to improve maternal and perinatal care.13 14 16 17

According to the WHO multi-country survey,9 CS rate was as high as 46% in China, 42% in Paraguay, and 40% in Ecuador with an overall mean of 26.4% for the 21 countries in the survey. Incremental rates as high as 18% within 3 to 4 years and a total CS rate of up to 80% have been reported in some parts of China.9 18 In Hong Kong, the annual CS rate rose steadily from 16.6% to 27.4% from 1987 to 1999, indicating a 65% increase over 12 years, with the CS rates in private institutions of approximately 27.4% higher than those in the public sector.4 The Hong Kong College of Obstetricians and Gynaecologists territory-wide audit has documented an increase in overall CS rates in Hong Kong from 27.1% in 1999 to 30.4% in 2004 and 42.1% in 2009,19 a drastic increase of 12% over a 5-year interval. The annual obstetric report of the HA in 2014 also showed varying CS rates among the eight public hospitals with obstetric services, ranging from 22.7% to 32%.20 The overall increase in CS rates of approximately 10% over the 20-year period to 24%-25% (approximately 0.5% per year on average) reported in the current study was modest in rate as well as lower in absolute value compared with the figures reported above, and those reported in other countries. The slight drop in CS rates for primiparous pregnancies with spontaneous term labour may be an important factor that mitigates the surge in CS rates in the study period.

The current data have demonstrated a modest drop in CS rate for primiparous women with spontaneous labour (category 1) and a slight fall in multiparous women with spontaneous labour (category 3) although the rate for all women with induced labour or prelabour CS (category 2 and 4) has remained constant. Review of the labour ward management protocol in the unit during the study period revealed that the adoption of evidence-based active management of labour protocols since the late 90s (including regular formal audits in CS rates and indications), the implementation of ‘best practices’ such as vigilant use of partograms,21 early amniotomy,22 and prompt oxytocin augmentation for slow progress23 could have contributed to the gradual but progressive fall in CS rates in these low-risk women. Similar measures in labour management have been shown in cluster-randomised trials to be associated with significant, albeit small declines in primary CS rates driven by the effects in low-risk pregnancies.24 Indeed the magnitude of fall of 1% to 2% in CS rates in such studies was similar to that observed over the two decades in the current study. As this category of low-risk primiparous women with spontaneous labour usually constitutes approximately ≥30% of the entire obstetric population, the effects of a modest fall in CS rates in this group will have a major impact on the overall rate. Other national studies to evaluate the effect of labour attempts and labour success on primary CS rates have shown that the fall in CS rates might not be persistent. After a slight drop in the late 90s, the rate started to rise again between 2004 and 2010.25 In addition, other meta-analyses have shown that the effects of such active management of labour, while consistently associated with shorter duration of labour and no discernible differences in neonatal and maternal outcome, might not be associated with significant reductions in CS rates.26 It remains to be seen whether the modest fall in CS rate in primiparous low-risk women in the current study will persist in future years. The effects of still other more drastic attempts to curb primary CS rates in primiparous women, including redefining labour dystocia,27 postponing the cut-off for active labour at 6-cm dilatation, allowing adequate time for second stage of labour, and encouraging operative vaginal delivery28 require further evaluation.

The rising proportion of women with previous CS who undergo repeat CS has been shown by various studies to contribute significantly to the overall rise in CS rates. For instance, at a single tertiary hospital level, it was shown that the Robson’s classification easily identified multiparous women with a previous CS scar as the leading patient group that contributed to an increase in CS rates from 38% in 1998 to 43.7% in 2011.13 Similarly, on a national scale, a French population–based study using perinatal survey data showed that a continuous rise in the CS rate was observed in three patient groups, one of which was women with previous CS.16 On an even larger scale, a WHO global survey of 97 095 women who delivered in one of 120 facilities in eight countries showed that although women with a previous CS (category 5) represented only 11.4% of the obstetric population, they were the largest contributor to the overall CS rate (26.7% of all the CSs). This highlights the great burden of repeat CS and the need to curb primary CS in order to control CS rates.17 In our study, the repeat CS rate escalated sharply from approximately 30% to 50%-60% over the two decades. This could be explained by the abandonment of the use of X-ray and computed tomographic pelvimetry as a selection tool 15 years ago to decide which patients with previous CS can undergo a trial of labour.29 As evidence accumulated that pelvimetry is imprecise and fails to predict successful trial of vaginal birth after CS,30 a liberal policy of allowing women with previous CS to choose between elective repeat CS or trial of labour was adopted since 2001. Although this policy is not based on strong evidence, the progressive increase in CS rate in this category indicates the preference of a large proportion of patients to elect repeat CS based on the relative indication of previous CS.

The Term Breech Trial published in 2000 is a good example of an important landmark study that has affected clinical protocols adopted by the unit and thus the CS rates in the study period.31 This was an authoritative randomised controlled trial which concluded that planned CS carries a reduced perinatal mortality and early neonatal morbidity for babies with breech presentation at term compared with planned vaginal birth. Although these findings have been challenged in subsequent studies,32 33 the policy of sectioning all breech babies has been widely adopted in international guidelines.34 35 Thus, while the CS rate for breech presentation was already high at approximately 70%-75% at the beginning of the study period, it increased to well over 90% in the subsequent 10 to 15 years to comply with these recommendations. Within this study period, 10%-12% of women with breech presentation at term opted for external cephalic version (ECV) and approximately 65% had achieved a successful vaginal delivery. With better counselling to achieve a higher acceptance of a trial of ECV, a decline in CS in this category can be anticipated.

The policy of allowing women with a twin pregnancy to opt for CS delivery was even more controversial. Over 90% of these CS deliveries were elective, based on maternal choice rather than emergency intrapartum obstetric indications. Over the 20 years of the study, women with a twin pregnancy in which one fetus was breech opted for CS in order to avoid a vaginal breech delivery at all costs, despite the lack of good clinical supporting evidence if the first twin is in vertex presentation.36 This further evolved into a patient expectation that all twin pregnancies should be sectioned, again despite contrary evidence from randomised controlled trials that elective CS in uncomplicated twins offers no perinatal advantage.37 The data from the current study showed that the liberal clinical policy we have adopted gradually since 2003 to accommodate such expectations has resulted in an overwhelming rise in CS rates in multiple pregnancies from >40% to >80%, far in excess of that which could be explained by a breech presentation38 or other risk factors in either twin.

Similarly, the literature has not shown any particular perinatal survival benefit for CS in preterm delivery of a cephalic-presenting fetus.39 There is also good evidence that CS delivery at very early gestations is associated with increased morbidity in the mother.40 Despite this, we observed that a large proportion of the increase in preterm CS was a result of planned iatrogenic preterm deliveries largely due to specific maternal or fetal conditions such as pre-eclampsia or early-onset fetal growth restriction with evidence of fetal compromise. The modest increase in the use of CS in these cases from approximately 19% to 23%-24% more likely reflects the obstetrician’s increasing preference for CS in the management of these cases rather than women’s choice. Nevertheless, the increase was modest and comparable with that reported in other centres.41

The overall contribution of this class to the overall CS rate was low. Stabilising induction after ECV was performed in only a small number of highly selected cases largely because of the low success rate (<30%), so that the impact of such a practice on CS rates in this category was limited. Hence the CS rate in this class remained high throughout the study period (>96%).

A strength of the current study was the large sample size collected over a long duration of two decades to allow significant trends to be observed. As a single-centre study, the impact of authoritative scientific guidelines or a change to liberal management policies that allowed patients with relative indications to undergo CS delivery could be readily identified. Although patient epidemiology, risk factors, and case-mix were believed to contribute to the rising CS rates observed within the study period, such effects were not observed in all categories. For example, advanced maternal age should have caused an increase in CS rates for low-risk primiparous women yet this was not observed. Changes in obstetric management protocols could also play an important role in these increasing trends. For instance, the rising repeat CS rate for women with previous CS from 36.7% to 57.0% during the study period grossly exaggerated the absolute increase in the number of CS performed in women with previous uterine scars. Liberal rules for multiple pregnancies as described above were not entirely evidence-based, but were often adopted to meet patient expectations. It remained a limitation that we could not test the temporal relationship of CS trend to changes in obstetric practice to establish a causal relationship.

In this study, it could be argued that the trends observed are specific to a public obstetric unit that did not entertain CS at the mother’s request in the absence of any clinical indications. However, CS rates have been observed to rise similarly in all other HA hospitals as reflected in the HA annual obstetric reports since 1999. We believe that our practice is similar to that of other public institutions in Hong Kong and that our observations can serve to encourage other obstetric units to audit their own trends, analyse the extent of rise in each Robson’s category, and identify the target groups that contribute most significantly to the rise in CS rates. Appropriate changes may then be made to clinical management protocols.

The most significant trends in an increase in CS rates were in line with the clinical practice towards CS for those with relative indications such as previous CS, breech presentation, and multiple pregnancies. The drop in CS rates for primiparous pregnancies with spontaneous term labour could be ascribed to more vigilant active labour management, and because the large absolute number in this group had the effect of mitigating the overall surge in CS rates. The overall increase in CS rates of approximately 10% over the 20-year period was modest compared with figures reported previously in Hong Kong and in other developed countries.

All authors have disclosed no conflicts of interest.

1. Betran AP, Torloni MR, Zhang J, et al. What is the optimal rate of caesarean section at population level? A systematic review of ecologic studies. Reprod Health 2015;12:57. Crossref

2. Belizán JM, Althabe F, Cafferata ML. Health consequences of the increasing caesarean section rates. Epidemiology 2007;18:485-6. Crossref

3. Betrán AP, Ye J, Moller AB, Zhang J, Gülmezoglu AM, Torloni MR. The increasing trend in caesarean section rates: global, regional and national estimates: 1990-2014. PLoS ONE 2016;11:e0148343. Crossref

4. Leung GM, Lam TH, Thach TQ, Wan S, Ho LM. Rates of cesarean births in Hong Kong: 1987-1999. Birth 2001;28:166-72. Crossref

5. Notzon FC. International differences in the use of obstetric interventions. JAMA 1990;263:3286-91. Crossref

6. Appropriate technology for birth. Lancet 1985;2:436-7.

7. Villar J, Valladares E, Wojdyla D, et al. Caesarean delivery rates and pregnancy outcomes: the 2005 WHO global survey on maternal and perinatal health in Latin America. Lancet 2006;367:1819-29. Crossref

8. Lumbiganon P, Laopaiboon M, Gulmezoglu AM, et al. Method of delivery and pregnancy outcomes in Asia: the WHO global survey on maternal and perinatal health 2007-08. Lancet 2010;375:490-9. Crossref

9. Vogel JP, Betrán AP, Vindevoghel N, et al. Use of the Robson classification to assess caesarean section trends in 21 countries: a secondary analysis of two WHO multicountry surveys. Lancet Glob Health 2015;3:e260-70. Crossref

10. Robson MS. Classification of caesarean sections. Fetal Matern Med Rev 2001;12:23-39. Crossref

11. Betrán AP, Vindevoghel N, Souza JP, Gülmezoglu AM, Torloni MR. A systematic review of the Robson classification for caesarean section: what works, doesn’t work and how to improve it. PLoS One 2014;9:e97769. Crossref

12. Torloni MR, Betran AP, Souza JP, et al. Classifications for cesarean section: a systematic review. PLoS One 2011;6:e14566. Crossref

13. Triunfo S, Ferrazzani S, Lanzone A, Scambia G. Identification of obstetric targets for reducing cesarean section rate using the Robson ten group classification in a tertiary level hospital. Eur J Obstet Gynecol Reprod Biol 2015;189:91-5. Crossref

14. Colais P, Fantini MP, Fusco D, et al. Risk adjustment models for interhospital comparison of CS rates using Robson’s ten group classification system and other socio-demographic and clinical variables. BMC Pregnancy Childbirth 2012;12:54. Crossref

15. Stavrou EP, Ford JB, Shand AW, Morris JM, Roberts CL. Epidemiology and trends for caesarean section births in New South Wales, Australia: a population-based study. BMC Pregnancy Childbirth 2011;11:8. Crossref

16. Le Ray C, Blondel B, Prunet C, Khireddine I, Deneux-Tharaux C, Goffinet F. Stabilising the caesarean rate: which target population? BJOG 2015;122:690-9. Crossref

17. Betrán AP, Gulmezoglu AM, Robson M, et al. WHO global survey on maternal and perinatal health in Latin America: classifying caesarean sections. Reprod Health 2009;6:18. Crossref

18. Huang K, Tao F, Bogg L, Tang S. Impact of alternative reimbursement strategies in the new cooperative medical scheme on caesarean delivery rates: a mixed-method study in rural China. BMC Health Serv Res 2012;12:217. Crossref

19. Yuen PM. Territory-wide audit in obstetrics and gynaecology 2014. Available from: http://www.hkcog.org.hk/hkcog/pages_3_77.html. Accessed Feb 2017.

20. Annual obstetric report 2014. Hong Kong: Hospital Authority; 2014.

21. Cox KJ, King TL. Preventing primary caesarean births: midwifery care. Clin Obstet Gynecol 2015;58:282-93. Crossref

22. Wei S, Wo BL, Qi HP, et al. Early amniotomy and early oxytocin for prevention of, or therapy for, delay in first stage spontaneous labour compared with routine care. Cochrane Database Syst Rev 2013;(8):CD006794. Crossref

23. Bugg GJ, Siddiqui F, Thornton JG. Oxytocin versus no treatment or delayed treatment for slow progress in the first stage of spontaneous labour. Cochrane Database Syst Rev 2013;(6):CD007123. Crossref

24. Chaillet N, Dumont A, Abrahamowicz M, et al. A cluster-randomized trial to reduce cesarean delivery rates in Quebec. N Engl J Med 2015;372:1710-21. Crossref

25. Simon AE, Uddin SG. National trends in primary cesarean delivery, labor attempts, and labor success, 1990-2010. Am J Obstet Gynecol 2013;209:554.e1-8. Crossref

26. Brown HC, Paranjothy S, Dowswell T, Thomas J. Package of care for active management in labour for reducing caesarean section rates in low-risk women. Cochrane Database Syst Rev 2013;(9):CD004907. Crossref

27. Boyle A, Reddy UM, Landy HJ, Huang CC, Driggers RW, Laughon SK. Primary cesarean delivery in the United States. Obstet Gynecol 2013;122:33-40. Crossref

28. American College of Obstetricians and Gynecologists, Society for Maternal–Fetal Medicine, Caughey AB, Cahill AG, Guise JM, Rouse DJ. Safe prevention of the primary cesarean delivery. Am J Obstet Gynecol 2014;210:179-93. Crossref

29. Pattinson RC, Farrell EM. Pelvimetry for fetal cephalic presentations at or near term. Cochrane Database Syst Rev 1997;(2):CD000161. Crossref

30. Dodd JM, Crowther CA, Huertas E, Guise JM, Horey D. Planned elective repeat caesarean section versus planned vaginal birth for women with a previous caesarean birth. Cochrane Database Syst Rev 2013;(12):CD004224. Crossref

31. Hannah ME, Hannah WJ, Hewson SA, Hodnett ED, Saigal S, Willan AR. Planned caesarean section versus planned vaginal birth for breech presentation at term: a randomised multicentre trial. Term Breech Trial Collaborative Group. Lancet 2000;356:1375-83. Crossref

32. Whyte H, Hannah ME, Saigal S, et al. Outcomes of children at 2 years after planned cesarean birth versus planned vaginal birth for breech presentation at term: the International Randomized Term Breech Trial. Am J Obstet Gynecol 2004;191:864-71. Crossref

33. Hannah ME, Whyte H, Hannah WJ, et al. Maternal outcomes at 2 years after planned cesarean section versus planned vaginal birth for breech presentation at term: the International Randomized Term Breech Trial. Am J Obstet Gynecol 2004;191:917-27. Crossref

34. Committee on Obstetric Practice. ACOG committee opinion: number 265, December 2001. Mode of term single breech delivery. Obstet Gynecol 2001;98:1189-90. Crossref

35. Royal College of Obstetricians and Gynaecologists. RCOG Green-top Guidelines: The management of breech presentation. Guideline No. 20b. December 2006. Available from: https://www.rcog.org.uk/globalassets/documents/guidelines/gtg-no-20b-breech-presentation.pdf. Accessed Feb 2017.

36. Rabinovici J, Barkai G, Reichman B, Serr DM, Mashiach S. Randomized management of the second nonvertex twin: vaginal delivery or cesarean section. Am J Obstet Gynecol 1987;156:52-6. Crossref

37. Barrett JF, Hannah ME, Hutton EK, et al. A randomized trial of planned cesarean or vaginal delivery for twin pregnancy. N Engl J Med 2013;369:1295-305. Crossref

38. Lee HC, Gould JB, Boscardin WJ, El-Sayed YY, Blumenfeld YJ. Trends in cesarean delivery for twin births in the United States: 1995-2008. Obstet Gynecol 2011;118:1095-101. Crossref

39. Riskin A, Riskin-Mashiah S, Lusky A, Reichman B; Israel Neonatal Network. The relationship between delivery mode and mortality in very low birthweight singleton vertex-presenting infants. BJOG 2004;111:1365-71. Crossref

40. Malloy MH. Impact of cesarean section on neonatal mortality rates among very preterm infants in the United States, 2000-2003. Pediatrics 2008;122:285-92. Crossref

41. Biswas A, Su LL, Mattar C. Caesarean section for preterm birth and, breech presentation and twin pregnancies. Best Pract Res Clin Obstet Gynaecol 2013;27:209-19. Crossref

The recent increasing availability and popularity of assisted reproductive technology (ART) has resulted in an increase in the incidence of multiple pregnancies.1 In the United States, the incidence of triplet pregnancies has increased two- to three-fold since the early 1980s.2 In Hong Kong, there was a 2.3-fold increase in ART procedures from 2009 to 2015, reaching over 11 000 procedures per year.3 The Council on Human Reproductive Technology of Hong Kong issued the Code of Practice on Reproductive Technology & Embryo Research in 2013 and limited the number of embryos transferred per cycle to three. Despite this recommendation, there has been no drop in the number of multiple pregnancies following ART, and the rate has remained at 6% since 2010.3 Thus multiple pregnancy is still a major obstetric concern in Hong Kong.

Triplet and higher-order multiple pregnancies are well known to be associated with increased adverse outcomes including maternal medical complications, pregnancy loss, intrauterine growth restriction, and preterm delivery. Triplet pregnancies have a four-fold increased risk of birth of <29 weeks compared with twin pregnancies.4 This is of particular concern as it significantly increases the perinatal morbidity and mortality due to prematurity. The risk of infant death in triplets is 3 times higher than that in twins.4

Reduction of triplets or higher-order multiple pregnancies has been performed since the 1980s. A meta-analysis of the early prospective non-randomised studies suggested that reduction of triplet pregnancies was associated with a reduction in maternal and fetal adverse outcomes.5 More importantly, the rate of early premature delivery (<32 weeks) can significantly be reduced following fetal reduction (FR), from 26%-33% to 5.5% for FR to a singleton and 10%-17% for FR to twins.6 7 8 Nonetheless, it is associated with procedure-related pregnancy loss in 4.5%-9.6% of cases when performed by fetal intracardiac injection of potassium chloride (KCl), and 8.8%-15% for cord coagulation.8 9 10 11

The objective of this study was to compare the perinatal outcomes for triplet pregnancy with and without FR at a university hospital in Hong Kong.

This was a retrospective cohort study conducted at Prince of Wales Hospital, Hong Kong. Triplet pregnancies with an expected date of confinement from 1 January 2008 to 30 September 2014 were retrieved from the prenatal diagnostic unit database and the Specialty Clinical Information System database. Demographics, and pregnancy and perinatal outcomes were reviewed. The data retrieval and review were performed by the first author, a medical officer from the department, who was not blinded to the study hypothesis. Women who carried a triplet pregnancy were classified into three groups: expectant management, FR to twins, or FR to a singleton.

Chorionicity and amnionicity were assessed during the first trimester by ultrasound. Women with a triplet pregnancy were offered counselling about the benefits and risks of expectant management versus FR to twins or a singleton. Specifically, women were counselled that approximately one third of triplet pregnancies resulted in preterm birth before 32 weeks of gestation. The aim of FR was primarily to reduce the risk of early preterm birth. Such risk can be reduced to 10%-18% following FR to twins and to 5%-8% if reduced to a singleton, although the procedure associated with a miscarriage rate of 5%-15%, depending on the method used.6 7 8 9 10 11 Ultimately, the decision for FR to twins or a singleton was dictated first by the mother’s wishes and second by whether FR was technically feasible. All procedures were performed in accordance with the Offences Against the Person Ordinance. The FR procedures carried out at the Prince of Wales Hospital were performed under real-time ultrasound guidance by Maternal and Fetal Medicine (MFM) specialists or supervised trainees. Some procedures were performed in the private sector. Choice of fetus(es) to be reduced depended on the presence of fetal abnormalities, placental location, and technical feasibility. Fetal gender was not revealed to the parents to avoid gender selection. Fetal intracardiac KCl injection was performed in fetuses with a separate placenta, that is, in trichorionic triamniotic (TCTA) triplets or when feticide was performed in both monochorionic fetuses of a dichorionic triamniotic (DCTA) pregnancy. A 20G needle was inserted transabdominally into the fetal heart and 1-5 mL of 14.9% KCl injected until fetal asystole was observed. Bipolar cord coagulation (BPC) or radiofrequency ablation (RFA) of the umbilical cord was performed in monochorionic fetuses in DCTA or monochorionic triamniotic (MCTA) triplets. In BPC, a 2-mm or 3-mm bipolar forceps was inserted transabdominally through a 3.9-mm trocar (Karl Storz, Tuttlingen, Germany) and the umbilical cord was grasped. Electrocautery was performed at a power setting of 20-60 W for a duration of 30-60 seconds. Two to three adjacent sites on the cord were cauterised. Cessation of blood flow through the umbilical cord was confirmed by colour and/or pulsed wave Doppler. In RFA, the radiofrequency needle (LaVeen SuperSlim Needle Electrode radiofrequency probe, 17G, 15 cm long; Massachusetts, United States) was inserted percutaneously into the fetal abdomen at the site of cord insertion. The prongs of the device were deployed and radiofrequency energy was applied in a stepwise fashion starting from 30 W and progressing to a maximum of 100 W; each energy level lasted no more than 2 minutes. Energy was applied until no blood flow was observed in the umbilical cord by colour and pulsed wave Doppler and a terminal fetal bradycardia was detected.

Women were discharged within 24 hours of the procedure, and were followed up 1 week later to confirm viability of the remaining fetus(es). Women who carried triplets and elected expectant management underwent elective caesarean delivery at 34 weeks. Monochorionic twins were delivered at 37 weeks and dichorionic twins at 38 weeks. Earlier delivery was indicated if there were maternal or fetal complications. The pregnancy outcomes studied included any pregnancy loss, gestational age at delivery, birth weight, neonatal survival rate, neonatal death, and neonatal complications: respiratory distress syndrome, chronic lung disease, intraventricular haemorrhage, necrotising enterocolitis, retinopathy of prematurity, neonatal sepsis, need for neonatal intensive care unit (NICU) care, and median length of NICU stay.

Ethics approval was obtained from the Institutional Review Board (CREC Ref No: 2016.697) with informed consent waived. The SPSS (Windows version 21.0; IBM Corp, Armonk [NY], US) was used for statistical analysis. Fisher’s exact test was used for categorical data, and Student’s t test for comparing the means between the expectant management and FR groups. A P value of <0.05 was taken as statistically significant.

There were 52 triplet pregnancies during the study period. Two cases were excluded from analysis as one was lost to follow-up after 13 weeks of gestation, and the other had a termination of pregnancy (TOP) at 8 weeks for social reasons. Of the included pregnancies, 84% (42/50) were the result of an ART procedure, of which 14 resulted from ovulation induction and 28 from in-vitro fertilisation. Among those conceived by ovulation induction, seven also included intrauterine insemination. In these 42 pregnancies conceived by ART, 33 (78.6%) were TCTA, eight (19.0%) were DCTA, and one (2.4%) was MCTA.

In the 50 cases included in the analysis, 26 (52.0%) pregnancies underwent FR and the remaining 24 (48.0%) had expectant management. Of the 26 cases of FR, 22 (84.6%) were reduced from triplets to twins and four (15.4%) from triplets to a singleton. Two of the pregnant women in the expectant management group elected FR, however, this could not be done due to technical difficulties and so they were managed conservatively. All except three FR procedures were performed at the Prince of Wales Hospital between 10 and 13 weeks of gestation. The three cases of FR performed in the private sector had fetal intracardiac KCl injection between 7 and 9 weeks of gestation.

Maternal characteristics and pregnancy outcomes are shown in Table 1. Women who underwent FR were 1 to 2 years older than those who elected expectant management. Parity, method of conception, chorionicity, and method of feticide between the three groups were similar. There was no miscarriage of the whole pregnancy in any of the three groups although five pregnancies that were managed expectantly had spontaneous fetal loss of one of the fetuses, and one pregnancy with FR to twins miscarried one twin following FR. The numbers of spontaneous fetal losses before 24 weeks (ie miscarriage rate) were 6.9% (5/72) and 2.1% (1/48) in the expectant management group and FR group, respectively, and there was no statistical significance between the two groups. There were three cases of preterm rupture of membranes (ROM) following FR and in all three cases the parents elected TOP. One woman had fetal intracardiac KCl injection to reduce a TCTA triplet pregnancy to twins at 11 weeks, but ROM occurred at 15 weeks of gestation. In the other two cases where feticide was performed to reduce the triplets to a singleton, one was a case of MCTA triplets reduced to a singleton with BPC of two fetuses at 13 weeks, and the other was TCTA triplets reduced to a singleton by fetal intracardiac KCl injection in two fetuses at 11 weeks of gestation. For these two cases, ROM occurred at 1 day and 11 weeks after FR, respectively. There were only two cases of cord coagulation in this study. Apart from the case of BPC in MCTA triplets that resulted in ROM, the other case was also MCTA triplets with RFA performed at 12 weeks for FR to twin pregnancy. The procedure was uncomplicated and the twins were subsequently delivered at 35 weeks of gestation. The gestational age at delivery was significantly higher in the FR groups. The mean gestation at delivery was 32.6 weeks in the expectant management group versus 35.2 and 39.6 weeks in the FR to two fetuses and one fetus, respectively. The risk of extreme preterm delivery was also significantly lower in those with FR (P=0.001). In women with expectant management, 16.7% had extreme preterm delivery of <28 weeks, and 29.2% delivered before 32 weeks. In women with FR to two fetuses, there was no case of extreme preterm delivery of <28 weeks, and 23.8% delivered before 32 weeks. All pregnancies with FR to a singleton had term birth. None of the pregnancies was complicated by twin-to-twin transfusion syndrome and intrauterine growth restriction occurred in only two and one pregnancies in the expectant management and FR to twins groups, respectively.

The neonatal outcome of the fetuses who survived beyond 24 weeks are reported in Table 2. There were two intrauterine fetal deaths, one in the expectant management group and one in the FR to twins group. All seven neonatal and post-neonatal deaths occurred in the expectant management group. These infants were all delivered <25 weeks and died of complications of prematurity. The mean birth weights in the FR groups to twins and a singleton were 553 g and 1073 g higher than that in the expectant management group, respectively. The need for NICU care (P=0.003) and length of NICU stay (P=0.040) were significantly higher in the expectant management group with no FR. Neonatal morbidities including respiratory distress syndrome, chronic lung disease, intraventricular haemorrhage, necrotising enterocolitis, retinopathy of prematurity, and neonatal sepsis were not statistically significant between the groups.

Multiple pregnancy is an increasingly important problem in obstetric practice as a result of the success of fertility procedures. Their incidence is expected to continue to rise as fertility services both in Hong Kong and in other nearby countries such as Taiwan, Thailand, and Mainland China become more accessible and affordable. As shown in our cohort, over 80% of triplet pregnancies were conceived by ART. This is the first study of the outcomes of triplet pregnancies following FR in Hong Kong. Of note, FR may not be acceptable to all parents, and parental choice has a strong influence on decisions about intervention. To make the best informed choice, parents should be counselled adequately on the benefits and risks of expectant management versus FR. The provision of local data on perinatal outcomes following FR is an essential part of that counselling.

The primary aim of FR is to reduce neonatal morbidity consequent to prematurity. Our results show that FR in a triplet pregnancy has the benefit of increasing the gestation at delivery and reducing the risk of extreme preterm delivery earlier than 32 and 28 weeks. The mean gestation at delivery was 32.6 weeks in the expectant management group versus 35.2 and 39.6 weeks in the FR to two fetuses and one fetus, respectively. This indicates that the performance of FR in Hong Kong is comparable with that reported in the literature and our results reaffirm previously reported data in which FR in triplet pregnancies to twins can prolong the pregnancy by approximately 3 weeks.12 In women with expectant management, 16.7% and 29.2% had extreme preterm delivery before 28 weeks and 32 weeks, respectively. In women with FR to two fetuses, there was no extreme preterm delivery of <28 weeks, and 23.8% delivered before 32 weeks. The risk of preterm delivery earlier than 32-33 weeks following FR has been reported to be between 24% and 37%,7 13 14 15 which are comparable to our results. All women with FR to one fetus had term deliveries. Despite the prolongation of pregnancy, however, the overall survival following FR was not significantly different to that following expectant management (Table 1). In fact, studies of FR in triplet pregnancies have not shown an increase in perinatal survival.12 13

Prolongation of pregnancy in the FR group leads to improvement in a number of outcomes. The FR group had a significantly higher birth weight. The need for NICU care and length of NICU stay were significantly lower. The seven neonatal deaths in our cohort were all of neonates from the expectant management group who were delivered <25 weeks and died of complications of prematurity. We were, however, unable to show a difference in neonatal morbidity due to the small numbers in each group.

The rates of spontaneous loss of the whole pregnancy in reduced versus non-reduced triplets have previously been reported to be 8.1% and 4.4%, respectively,7 although such loss rate can be reduced with increasing experience so that it is comparable with that in non-reduced triplets.6 10 11 In our study, there was no spontaneous total pregnancy loss. This may be because all procedures were carried out by MFM specialists or trainees with expertise in invasive fetal procedures. There were fetal losses in both the expectant management and FR groups, but it was not statistically significant. Three cases had prelabour ROM after FR, and these parents elected TOP due to poor prognosis. Most studies of FR report a procedure-related pregnancy loss or miscarriage rate,6 7 8 10 11 but there are no data for the rate of prelabour ROM after FR in triplet pregnancies. In complicated monochorionic pregnancies, the rate of prelabour ROM following cord coagulation has been reported to be 20% to 30%.9 16 17 On the contrary, the prelabour ROM rate was 12% after fetal intracardiac KCl injection in multichorionic pregnancies.18

This study found that only maternal age influenced a decision to undergo FR but this is contrary to the findings of other studies.7 13 19 A possible explanation is that in women with advanced age, multiple pregnancy will add additional maternal risks during the pregnancy and may influence a decision to undergo FR. When choosing FR to twins or a singleton, our results showed that the chorionicity of the triplet pregnancies most likely affected their choice: 64% (21/33) of those with TCTA triplets chose to have FR although 95% (20/21) opted for reduction to twins. In DCTA and MCTA triplets, only 27% and 33% proceeded to FR, respectively. As our study only had two cases of RFA or BPC performed in MCTA triplets, we cannot conclude any reason for a low rate of FR by cord coagulation. It is, however, possible that women believed there was a higher risk of miscarriage associated with RFA or BPC.

It must be emphasised that in women who choose to reduce one fetus in a DCTA pregnancy, the best option is to reduce one fetus in the monochorionic pair, not the fetus with a separate placenta, by fetal intracardiac KCl injection. Although the latter is technically easier, there would be continued risks of twin-to-twin transfusion syndrome or twin anaemia polycythaemia sequence when the singleton fetus is reduced to keep the monochorionic twins, and this poses significant risks to the pregnancy. The miscarriage risk before 24 weeks following FR of the fetus with a separate placenta has been reported to be as high as 23.5%.8 Therefore, women who elect to have FR of DCTA triplets to twins should be referred to a tertiary unit with expertise in FR and where more advanced techniques are readily available. Nevertheless, the best perinatal outcome in any type of triplets will be achieved by reducing two fetuses resulting in a singleton pregnancy.8 Women along with their partner should be aware of this and be adequately counselled so that they may make an informed decision.

This is the first study of FR in triplet pregnancy in Hong Kong. It provides valuable data on the local experience in FR that is useful in parental counselling. The strength of this study is that comprehensive perinatal outcomes were obtained in all pregnancies except two.

This study has limitations. First, the number of cases included was small, although we believe this is the largest cohort possible to be reported in Hong Kong as our hospital has the highest number of deliveries and receives referrals for FR from private obstetricians and other obstetric units in Hong Kong. Second, the number of cord coagulation procedures was limited, and no further analysis was possible to determine which cord coagulation procedure is superior.

Approximately 50% of women with a triplet pregnancy in Hong Kong elected to undergo FR, which can significantly prolong the gestation at delivery and significantly reduce preterm delivery of <32 weeks, although it is associated with risk of miscarriage and complications such as ROM. Women carrying a triplet pregnancy should be adequately counselled about the benefits and risks of FR so that they can make an informed decision.

All authors have disclosed no conflicts of interest.

1. Pison G, Monden C, Smits J. Twinning rates in developed countries: trends and explanations. Popul Dev Rev 2015;41:629-49. Crossref

2. Martin JA, Osterman MJ, Thoma ME. Declines in triplet and higher-order multiple births in the United States, 1998-2014. NCHS Data Brief 2016;(243):1-8.

3. Council on Human Reproductive Technology, Hong Kong. Code of practice on reproductive technology & embryo research. Available from: http://www.chrt.org.hk/english/service/service_cod.html. Accessed 26 Jan 2017.

4. Luke B, Brown MB. Maternal morbidity and infant death in twin vs triplet and quadruplet pregnancies. Am J Obstet Gynecol 2008;198:401.e1-10. Crossref

5. Dodd J, Crowther C. Multifetal pregnancy reduction of triplet and higher-order multiple pregnancies to twins. Fertil Steril 2004;81:1420-2. Crossref

6. Evans MI, Berkowitz RL, Wapner RJ, et al. Improvement in outcomes of multifetal pregnancy reduction with increased experience. Am J Obstet Gynecol 2001;184:97-103. Crossref

7. Papageorghiou AT, Avgidou K, Bakoulas V, Sebire NJ, Nicolaides KH. Risks of miscarriage and early preterm birth in trichorionic triplet pregnancies with embryo reduction versus expectant management: new data and systematic review. Hum Reprod 2006;21:1912-7. Crossref

8. Morlando M, Ferrara L, D’Antonio F, et al. Dichorionic triplet pregnancies: risk of miscarriage and severe preterm delivery with fetal reduction versus expectant management. Outcomes of a cohort study and systematic review. BJOG 2015;122:1053-60. Crossref

9. Rossi AC, D’Addario V. Umbilical cord occlusion for selective feticide in complicated monochorionic twins: a systematic review of literature. Am J Obstet Gynecol 2009;200:123-9. Crossref

10. Stone J, Eddleman K, Lynch L, Berkowitz RL. A single center experience with 1000 consecutive cases of multifetal pregnancy reduction. Am J Obstet Gynecol 2002;187:1163-7. Crossref

11. Stone J, Ferrara L, Kamrath J, et al. Contemporary outcomes with the latest 1000 cases of multifetal pregnancy reduction (MPR). Am J Obstet Gynecol 2008;199:406.e1-4. Crossref

12. van de Mheen L, Everwijn SM, Knapen MF, et al. The effectiveness of multifetal pregnancy reduction in trichorionic triplet gestation. Am J Obstet Gynecol 2014;211:536.e1-6. Crossref

13. Chaveeva P, Kosinski P, Puglia D, Poon LC, Nicolaides KH. Trichorionic and dichorionic triplet pregnancies at 10-14 weeks: outcome after embryo reduction compared to expectant management. Fetal Diagn Ther 2013;34:199-205. Crossref

14. Wimalasundera RC. Selective reduction and termination of multiple pregnancies. Semin Fetal Neonatal Med 2010;15:327-35. Crossref

15. Antsaklis A, Souka AP, Daskalakis G, et al. Embryo reduction versus expectant management in triplet pregnancies. J Matern Fetal Neonatal Med 2004;16:219-22. Crossref

16. Bebbington MW, Danzer E, Moldenhauer J, Khalek N, Johnson MP. Radiofrequency ablation vs bipolar umbilical cord coagulation in the management of complicated monochorionic pregnancies. Ultrasound Obstet Gynecol 2012;40:319-24. Crossref

17. van den Bos EM, van Klink JM, Middeldorp JM, Klumper FJ, Oepkes D, Lopriore E. Perinatal outcome after selective feticide in monochorionic twin pregnancies. Ultrasound Obstet Gynecol 2013;41:653-8. Crossref

18. Mohammed AB, Farid I, Ahmed B, Ghany EA. Obstetric and neonatal outcome of multifetal pregnancy reduction. Middle East Fertil Soc J 2015;20:176-81. Crossref

19. Boulot P, Vignal J, Vergnes C, Dechaud H, Faure JM, Hedon B. Multifetal reduction of triplets to twins: a prospective comparison of pregnancy outcome. Hum Reprod 2000;15:1619-23. Crossref

Single-level lumbar discectomy is a very common surgical procedure and has been proven to be beneficial for patients with lumbar disc herniation (LDH). Recurrent lumbar disc herniation (rLDH) is defined as disc herniation at the same level, regardless of ipsilateral or contralateral herniation, in a patient who has experienced a pain-free interval of at least 6 months after surgery.1 2 The true incidence of same-level rLDH after lumbar discectomy is unclear. The recurrence rate of LDH has been reported to be 5% to 15%.1 2 3 4 5 There have been many studies designed to determine the recurrence of LDH, and various risk factors suggested including disc degeneration, trauma, age, smoking, gender, and obesity.1 3 6 Radiologically identifiable factors, such as disc degeneration, disc height, and sagittal range of motion have been shown to be related to spinal instability and consequently to rLDH.7 8 9 In this retrospective study, we analysed the influence of disc degeneration, endplate changes, surgical technique, and patient’s clinical characteristics on rLDH.

We examined factors that could influence the recurrence of LDH, especially in those with the highest recurrence rate, and to minimise the biomechanical changes at every level. We retrospectively reviewed the medical records of patients with L4-L5 LDH who underwent lumbar discectomy between August 2004 and September 2009 at the Neurosurgery Department of Ataturk Education and Research Hospital in Ankara, Turkey. This hospital is a tertiary referral hospital and a centre for education and scientific research. This study was approved by the Ethics Committee of Ankara Ataturk Education and Research Hospital, Turkey. The principles outlined in the Declaration of Helsinki have also been followed.

Patients were excluded if they had any of the following: prior lumbar surgery at another institution, segmental instability, vertebral fractures and spinal infections, other types of degenerative disc disease, tumours, pregnancy, and age over 75 years. Patients were included if they had radicular pain for at least 3 months that was refractory to 6 weeks of conservative treatment with or without neurological deficit, numbness in the lumbar spine, buttock, and/or lower extremity, age between 21 and 75 years, and magnetic resonance imaging (MRI) and/or computed tomography demonstrating anatomical unilateral LDH correlating with symptoms. In the rLDH group, patients were additionally required to have had a pain-free interval of at least 6 months following the first surgery. We compared the patients’ demographic and clinical characteristics (age, sex, body mass index [BMI], diabetes mellitus, smoking, herniation type), preoperative radiological parameters (Pfirrmann disc degeneration grade, Modic endplate changes, disc height), surgical technique (microdiscectomy, open discectomy), and duration of symptoms. All surgeries were performed by the same group of surgeons via microdiscectomy or open discectomy technique as previously reported and standardised by Williams10 and Mixter.11 The type of herniation was classified as protrusion, extrusion, or sequestration after retrospective review of surgical records and MRI studies. The staff in the radiological department were blinded to the outcome of the study. Lumbar MRI and simple radiographic examinations were performed in all patients before surgery. Intervertebral disc height measurements were calculated using the lateral radiographs. The degree of disc degeneration was assessed on T2-weighted sagittal MRI sequences. Disc degeneration was classified by our radiological department in a retrospective and blinded manner according to modified Pfirrmann criteria as shown in Table 1.12 Modic endplate changes were also classified with the help of our radiological department on T1/T2-weighted sagittal MRI sequences, again blinded to the outcome.13 14

Data analysis was performed using the Statistical Package for the Social Sciences (Windows version 11.5; SPSS Inc, Chicago [IL], United States). Test of normality was applied to parametric data. Parametric numerical data were compared with Student’s t test and non-parametric data with Mann-Whitney U test. Chi squared test was used with Fisher’s exact test to compare categorical and nominal variables as appropriate. Wilcoxon signed-rank test was used to compare differences between paired data. To identify the associations between recurrence and diabetes mellitus and smoking, logistic regression analysis was used. In detail, recurrence was included in the regression model as a dependent variable; smoking status and existence of diabetes mellitus were included as covariants and the enter mode was used. P value was used to determine whether the differences were statistically significant. A P value of <0.05 was considered statistically significant.

A retrospective analysis of 600 patients who underwent surgery between August 2004 and September 2009 for only a single-level LDH was performed. The level of disc herniation was L1-L2 in four (0.7%), L2-L3 in 17 (2.8%), L3-L4 in 39 (6.5%), L4-L5 in 289 (48.2%), and L5-S1 in 251 (41.8%) patients. Of the 600 patients, 44 had rLDH; their respective distributions of disc levels were 0, 1, 3, 25, and 15. The total recurrence rate was 7.3%. Recurrence rate for L4-L5 level was the highest at 8.6%. The mean follow-up time of all patients was 323 days. The mean symptom duration was 78 days for the non-recurrent group and 77 days for the recurrent group. Patients in the recurrent group were pain-free for at least 6 months after the first operation with a median of 240 days (range, 188-1260 days).

After applying the exclusion/inclusion criteria, we retrospectively analysed 126 patients who underwent primary single-level L4-L5 lumbar discectomy and who were reoperated on for rLDH. The patients were divided into recurrent (n=25) and non-recurrent (n=101) group (Fig).

Univariate analysis for demographics, and clinical and radiological characteristics of the recurrent and non-recurrent groups is shown in Table 2. There were no significant differences in the age, sex, type of disc herniation, type of surgery, and duration of symptoms between the two groups. Of the patients, 13 (52.0%) in the recurrent group and 42 (41.6%) in the non-recurrent group were smokers. In comparison with non-smokers, smokers had a 50% higher recurrence rate (odds ratio [OR]=1.52; 95% confidence interval [CI], 0.63-3.66). Five (20.0%) patients in the recurrent group and 10 (9.9%) in the non-recurrent group had diabetes mellitus. Patients with diabetes mellitus had twice the recurrence rate of those without (OR=2.27; 95% CI, 0.70-7.38). Preoperative mean intervertebral disc heights were significantly different between the recurrent (19.1 mm) and non-recurrent (15.0 mm) groups (P<0.001). A higher preoperative intervertebral disc space might be a risk factor for recurrence. A higher BMI was a statistically significant factor in the recurrent group (P=0.042). Modic endplate changes were statistically higher in the recurrent group than in the non-recurrent group (P=0.032). In the non-recurrent group, 13 (12.9%) patients showed grade II, 51 (50.5%) showed grade III, 24 (23.8%) showed grade IVA, and 13 (12.9%) showed grade IVB Pfirrmann disc degeneration. Although Pfirrmann disc degeneration was not statistically significant between the two groups (P=0.079), it might still be a moderate marker for a potential risk of recurrence.

The comparative analysis of preoperative clinical and radiological characteristics of the recurrent group is shown in Table 3. Before the first surgery, one (4%) patient showed grade II, 10 (40%) showed grade III, eight (32%) showed grade IVA, and six (24%) showed grade IVB Pfirrmann disc degeneration. In one patient, grade II Pfirrmann degeneration progressed to IVA before the second surgery. Six of 10 patients with grade III changed to IVB, three changed to IVA, and one remained unchanged with a Pfirrmann degeneration grade III after recurrence.

These results led us to consider the relationship between grade of degenerated disc and herniation type. Herniation type was compared with Pfirrmann disc degeneration degree and Modic endplate changes. Patients with extrusion-sequestration herniation had a statistically significant higher Pfirrmann disc degeneration in contrast to patients with protrusions (P=0.016). Nonetheless, there was no correlation between Modic changes and herniation type (P=0.279).

Degenerative disc disease remains a poorly understood phenomenon because of the lack of precise definitions for healthy and degenerated discs. Decreased nutrition is the final common pathway for degenerative disc disease and the status of the endplate plays a crucial role in controlling the extent of diffusion and is the only source of nutrition.15 A recurrence rate of 5% to 15% for LDH has been reported.1 2 3 4 5 Differentiation of recurrent disc herniation from scar formation will allow for improved treatment choices and selection of patients who may benefit from a second surgery. Gadolinium-enhanced MRI is thought to be the best modality to differentiate between these two diagnoses. There is much debate about the risk factors for rLDH and it is very difficult to define them because many clinical and complicated biomechanical parameters are involved.

In this study, we analysed the influence of disc degeneration, endplate changes, surgical technique, and patient’s clinical characteristics (age, gender, BMI, symptom duration, herniation type, smoking status, and diabetes). Kim et al6 reported old age, high BMI, protrusion type of disc herniation, and positive Modic changes as risk factors after percutaneous endoscopic discectomy. Swartz and Trost,2 however, found that age, gender, smoking status, level of herniation, and duration of symptoms were not associated with rLDH. We showed that disc height, BMI, and Modic endplate changes were significantly correlated with a higher incidence of rLDH. Although diabetes and smoking were not statistically significant in our study, patients with diabetes had twice the recurrence rate of those without. Furthermore, patients who were a smoker had a 50% higher recurrence rate in contrast to non-smokers. The exact mechanism by which smoking contributes to disc degeneration is incompletely understood, but may be related to disc annulus nutrition and oxygenation, as well as increases in intradiscal pressure due to excessive coughing. Vascular insufficiency as a result of atheromas should also be considered.16 17 18 19 20 Analogous with our results, these presumptions may account for smoking as a cause of rLDH. In contrast with these findings, however, some studies found no relationship between smoking and rLDH.21 22 23

Clinical studies of disc height and recurrence have shown that degenerative segments with preserved disc height have a latent instability compared with segments with collapsed discs.8 Other studies have shown that the restabilisation stage begins when disc height is reduced by 50%.7 Similar to these studies, our study showed that preoperative higher intervertebral disc space measurements were significantly more important in recurrence (P<0.001).

Pfirrmann disc degeneration grade was not statistically significant in the recurrent group in contrast to non-recurrent group (P=0.079), but patients with extrusion and sequestration had a statistically significant higher Pfirrmann disc degeneration than patients with protrusions (P=0.016). These findings provide evidence that the healing processes that occur in the outer lamellas after annular injury may not be sufficient for effective reconstitution of the external annulus in degenerated discs.24 25 Increases in disc degeneration cause larger volumes of herniation type. Studies of Modic endplate changes after lumbar discectomy have shown incremental changes in disc degeneration grade.18 26 It is accepted that Modic type 1 changes are dynamically unstable and inflammatory lesions, whereas type 2 lesions are much more stable and unchangeable.26 Therefore, posterior lumbar interbody fusion combined with pedicle screw fixation is suggested for degenerative lumbar disc disease with Modic changes.27 Another study suggested treatment of Modic type 1 and 2 lesions with degenerative disc disease with posterior dynamic stabilisation.28 Our study showed that Modic changes were statistically higher in the recurrent group than the non-recurrent group (P=0.032). These findings suggest that patients with LDH and higher preoperative disc heights and Modic changes have a higher risk and tendency to recurrence of LDH. Although our study does not include different modalities to include lumbar disc diseases with Modic changes, the results might suggest a supplemental approach such as posterior stabilisation and fusion, or newly proposed treatment options with dynamic posterior stabilisation.

This study has several limitations. It would have a greater impact if we had included a larger subgroup population, especially for rLDH. To investigate factors that influence recurrence of L4-L5 disc herniation, however, several clinical and radiological parameters such as canal diameter, facet angle, annular defect size, location of the herniation type etc would need to be considered. The aim of this study was to focus on the effect of disc height, endplate changes, and disc degeneration in rLDH at L4-L5 level. As only univariate analyses were performed, we have no adjustment for potential confounding, hence the independent effects of the risk factors could not be documented. A prospective study would obtain more precise results, especially due to standardised sampling and classification of data.

This study suggests that patients who had LDH with higher preoperative disc height, higher BMI, and Modic endplate changes have a higher tendency for rLDH. Well-planned and well-conducted large-scale prospective cohort studies are essential to firmly evaluate and determine factors involved in rLDH.

We thank Ms Fatma Kubra Erbay from Department of Micro and Nanotechnology, TOBB University of Economics and Technology, Ankara, Turkey for her great effort in revising the statistical results of this study.

All authors have disclosed no conflicts of interest.

1. Suk KS, Lee HM, Moon SH, Kimm NH. Recurrent lumbar disc herniation: results of operative management. Spine (Phila Pa 1976) 2001;26:672-6. Crossref

2. Swartz KR, Trost GR. Recurrent lumbar disc herniation. Neurosurg Focus 2003;15:E10. Crossref

3. Connolly ES. Surgery for recurrent lumbar disc herniation. Clin Neurosurg 1992;39:211-6.

4. Fandiño J, Botana C, Viladrich A, Gomez-Bueno J. Reoperation after lumbar disc surgery: results in 130 cases. Acta Neurochir (Wien) 1993;122:102-4. Crossref

5. Mobbs RJ, Newcombe RL, Chandran KN. Lumbar discectomy and the diabetic patient: incidence and outcome. J Clin Neurosci 2001;8:10-3. Crossref

6. Kim JM, Lee SH, Ahn Y, Yoon DH, Lee CD, Lim ST. Recurrence after successful percutaneous endoscopic lumbar discectomy. Minim Invasive Neurosurg 2007;50:82-5. Crossref

7. Axelsson P, Karlsson BS. Intervertebral mobility in the progressive degenerative process. A radiostereometric analysis. Eur Spine J 2004;13:567-72. Crossref

8. Hasegawa K, Kitahara K, Hara T, Takano K, Shimoda H, Homma T. Evaluation of lumbar segmental instability in degenerative diseases by using a new intraoperative measurement system. J Neurosurg Spine 2008;8:255-62. Crossref

9. Zhao F, Pollintine P, Hole BD, Dolan P, Adams MA. Discogenic origins of spinal instability. Spine (Phila Pa 1976) 2005;30:2621-30. Crossref

10. Williams RW. Microlumbar discectomy: a conservative surgical approach to the virgin herniated lumbar disc. Spine (Phila Pa 1976) 1978;3:175-82. Crossref

11. Mixter WJ. Pitfalls in the surgery of the ruptured intervertebral disk. J Fla Med Assoc 1952;39:159-67.

12. Pfirrmann CW, Metzdorf A, Zanetti M, Hodler J, Boos N. Magnetic resonance classification of lumbar intervertebral disc degeneration. Spine (Phila Pa 1976) 2001;26:1873-8. Crossref

13. Modic MT, Masaryk TJ, Ross JS, Carter JR. Imaging of degenerative disk disease. Radiology 1998;168:177-86. Crossref

14. Modic MT, Steinberg PM, Ross JS, Masaryk TJ, Carter JR. Degenerative disk disease: assessment of changes in vertebral body marrow with MR imaging. Radiology 1988;166(1 Pt 1):193-9. Crossref

15. Rajasekaran S, Venkatadass K, Naresh Babu J, Ganesh K, Shetty AP. Pharmacological enhancement of disc diffusion and differentiation of healthy, ageing and degenerated discs: results from in-vivo serial post-contrast MRI studies in 365 human lumbar discs. Eur Spine J 2008;17:626-43. Crossref

16. Akmal M, Kesani A, Anand B, Singh A, Wiseman M, Goodship A. Effect of nicotine on spinal disc cells: a cellular mechanism for disc degeneration. Spine (Phila Pa 1976) 2004;29:568-75. Crossref

17. Frymoyer JW, Pope MH, Costanza MC, Rosen JC, Goggin JE, Wilder DG. Epidemiologic studies of low-back pain. Spine (Phila Pa 1976) 1980;5:419-23. Crossref

18. Iwahashi M, Matsuzaki H, Tokuhashi Y, Wakabayashi K, Uematsu Y. Mechanism of intervertebral disc degeneration caused by nicotine in rabbits to explicate intervertebral disc disorders caused by smoking. Spine (Phila Pa 1976) 2002;27:1396-401. Crossref

19. Nemoto Y, Matsuzaki H, Tokuhasi Y, et al. Histological changes in intervertebral discs after smoking cessation: experimental study using a rat passive smoking model. J Orthop Sci 2006;11:191-7. Crossref

20. Stairmand JW, Holm S, Urban JP. Factors influencing oxygen concentration gradients in the intervertebral disc. A theoretical analysis. Spine (Phila Pa 1976) 1991;16:444-9. Crossref

21. Kara B, Tulum Z, Acar U. Functional results and the risk factors of reoperations after lumbar disc surgery. Eur Spine J 2005;14:43-8. Crossref

22. Meredith DS, Huang RC, Nguyen J, Lyman S. Obesity increases the risk of recurrent herniated nucleus pulposus after lumbar microdiscectomy. Spine J 2010;10:575-80. Crossref

23. Palma L, Carangelo B, Muzii VF, Mariottini A, Zalaffi A, Capitani S. Microsurgery for recurrent lumbar disk herniation at the same level and side: do patients fare worse? Experience with 95 consecutive cases. Surg Neurol 2008;70:619-21. Crossref

24. Hampton D, Laros G, McCarron R, Franks D. Healing potential of the anulus fibrosus. Spine (Phila Pa 1976) 1989;14:398-401. Crossref

25. Osti OL, Vernon-Roberts B, Fraser RD. 1990 Volvo Award in experimental studies. Anulus tears and intervertebral disc degeneration. An experimental study using an animal model. Spine (Phila Pa 1976) 1990;15:762-7. Crossref

26. Rahme R, Moussa R, Bou-Nassif R, et al. What happens to Modic changes following lumbar discectomy? Analysis of a cohort of 41 patients with a 3- to 5-year follow-up period. J Neurosurg Spine 2010;13:562-7. Crossref

27. Kwon YM, Chin DK, Jin BH, Kim KS, Cho YE, Kuh SU. Long term efficacy of posterior lumbar interbody fusion with standard cages alone in lumbar disc diseases combined with modic changes. J Korean Neurosurg Soc 2009;46:322-7. Crossref

28. Öktenoglu T, Ozer AF, Sasani M, et al. Posterior dynamic stabilization in the treatment of lumbar degenerative disc disease: 2-year follow-up. Minim Invasive Neurosurg 2010;53:112-6. Crossref

The advent of biologic agents has revolutionised the treatment of patients with rheumatoid arthritis (RA), psoriatic arthritis (PSA), and spondyloarthritis (SPA). The outcome is now greatly improved. This, however, comes at the price of a clear heightened risk of active tuberculosis (TB) as a progression of latent TB infection (LTBI).1 Therefore, it is universally agreed that screening for LTBI prior to biologic therapy is necessary, especially in endemic areas such as Hong Kong.2 Unfortunately, there remains controversy regarding how best to accomplish this task.

The tuberculin skin test (TST) has been the routine screening tool in Hong Kong for the past decade.3 Its accuracy, however, is far from perfect, especially in patients who have been vaccinated with Bacillus Calmette–Guérin (BCG), are immunocompromised, or have been infected with atypical mycobacterium.4 Recently, interferon-gamma release assays (IGRAs) that measure interferon-gamma secretion in response to Mycobacterium tuberculosis–specific antigens have become available to detect LTBI. They have been shown to offer improved specificity and probably sensitivity.5 6 Other shortcomings of the TST, such as the need for return visits and reader variability, are also overcome. One of the IGRAs, the ASACIR.TB (A.TB; Haikou VTI Biological Institute, Hainan, China), has shown encouraging results in a large-scale clinical trial conducted in China and might be more appropriate in Chinese populations.7

On the other hand, IGRAs are not flawless. The rate of indeterminate results has been reported to be as high as 40%.8 The immunocompromised state of arthritic patients will also induce a depressed response to a T-cell reaction leading to an inaccurate IGRA result. There are recent data to argue that an IGRA alone is insufficient to identify all patients at risk.9 10 Furthermore, various studies have suggested very different concordance figures between the IGRA and the TST, likely as a result of heterogeneity (eg differing background TB prevalence, variable immunosuppressive therapies, or underlying BCG status).11

This study aimed to evaluate the agreement between the IGRA and the TST in the diagnosis of LTBI in patients with arthritic diseases scheduled to receive biologic agents in Hong Kong.

We reviewed 38 patients with RA, PSA, or SPA at a local hospital in Hong Kong from August 2013 to April 2014. They were diagnosed according to the 2010 classification criteria for RA of the American College of Rheumatology/European League Against Rheumatism, the Classification Criteria for Psoriatic Arthritis, and the Assessment of SpondyloArthritis international Society classification criteria, respectively. Patients were included if they were considered candidates for biologic agents. Concurrent medications were documented. Candidates were excluded if they had active TB infection, a history of incomplete TB treatment, or no measured induration. Patients underwent both the IGRA and the TST simultaneously. Those who tested positive for either test and who were due to be prescribed biologic agents were given latent TB treatment with isoniazid or rifampicin for 9 months. All patients were followed up regularly for 1 year and the development of active TB infection was evaluated. This study conforms to the provisions of the Declaration of Helsinki and the guidelines of the local ethical committee. Informed consent was considered not necessary due to the retrospective nature of the study.

The TST was performed by rheumatologists. A 0.1 mL of 2-TU PPD (tuberculin units of purified protein derivative) was injected intradermally into the volar aspect of the forearm. The indurations were measured in millimetres after 48 hours of inoculation by rheumatologists who were blinded to the IGRA results. According to the local guideline, induration of ≥10 mm was considered a positive result of LTBI.3

We performed the A.TB IGRA test (Haikou VTI Biological Institute) in all study patients. This assay employs Haikou VTI’s patented technology (US patent number 7754219) that enables intracellular delivery of the full-length protein CFP-10 and the antigen ESAT-6 to stimulate antigen-specific T-cells through the major histocompatibility complex class 1 pathway.12 13 The assay was performed according to the user manual. In brief, negative control phosphate buffered saline (N), positive control concanavalin A (P), and the TB stimulators CFP-10 and ESAT-6 (T) were mixed with fresh heparinised whole blood and incubated for approximately 24 hours at 37.8°C. The plasma was collected and stored at 48°C for up to 2 weeks. The interferon-gamma level in the plasma was then determined by enzyme-linked immunosorbent assay. If N was <0.5 IU/mL and (T-N)/(P-N) ≥0.6, or if N was ≥0.5 IU/mL and (T-N)/(P-N) ≥0.85, the test was considered to be positive (A.TB+), otherwise the result was negative (A.TB–).

Descriptive statistics were presented as frequencies and means ± standard deviations as appropriate. The concordance between TST and IGRA was evaluated by the Cohen’s weighted k statistic. A kappa value of >0.6 represents substantial agreement, 0.41 to 0.60 moderate agreement, 0.21 to 0.40 fair agreement, and <0.21 slight agreement. The concordance was subanalysed in patients with and without prednisolone.

The demographic and clinical characteristics of 38 patients are summarised in Table 1. All patients were residents of Hong Kong. Half of the patients were prescribed systemic steroid therapy that comprised prednisolone at a dose of 2.5 mg daily to 15 mg daily. All except three patients with SPA were on various conventional disease-modifying antirheumatic drugs.

The results of the concomitant TST and IGRA are shown in Table 2. Of the 38 patients, based on an induration of 10-mm diameter as the cut-off value, 13 (34.2%) had a positive TST, 11 (28.9%) had a positive IGRA. The agreement between A.TB IGRA test and TST was 73.7%. Six patients were TST+/A.TB– and four were TST–/A.TB+. Subgroup analysis showed that four of the six divergent TST+/A.TB– results were in patients on systemic steroid, and only three patients with systemic steroid were A.TB+ versus eight patients without. When positive TST was defined as an induration of 5-mm diameter, half of the patients had a positive TST, among them nine were TST+/A.TB–. Only one was TST–/A.TB+. In patients prescribed a systemic steroid, with 5-mm induration as positivity, TST missed no patients who had positive IGRAs.

Analysis of the agreement between the two tests, assessed by kappa statistic, showed only fair strength in our study, with a kappa value of 0.39 (Table 3). When a 5-mm induration was taken as a positive TST, however, the agreement between the two tests improved to moderate with a kappa value of 0.47. Subgroup analysis revealed that the agreement between both tests improved further (kappa=0.57) in patients not taking a concurrent systemic steroid. For patients taking a systemic steroid, the agreement was only slight (kappa=0.066). Again, the agreement of the TST and IGRA was substantial (kappa=0.69) in patients not on systemic steroid therapy when a 5-mm induration was regarded as positive.

At the end of the study, 32 of the initial 38 patients had received biologic agents. None of them developed active TB during the 1-year follow-up period.

In clinical practice there is no gold standard test for diagnosing LTBI. Both IGRA and TST have strengths and weaknesses. In a meta-analysis performed on an unselected population, the specificity of IGRA was 99% in a non-BCG–vaccinated population and 96% in a BCG-vaccinated population.14 The specificity of the TST was 97% in a non-BCG–vaccinated population but dropped to 59% in a BCG-vaccinated population.14 In addition to BCG history, comparison between the two tests must take into account the underlying disease, the immunosuppression status, and the background TB burden of the population being screened.

In this study, we found only a fair agreement (kappa=0.39) between the results of TST and A.TB IGRA in arthritis patients from a TB-endemic area. Some studies have also reported a discrepancy between the two tests in countries with intermediate TB burden.15 16 Nonetheless, the reported agreement between TST and the IGRA was good (kappa of 0.72 in United Kingdom17 and 0.87 in Denmark18). Consequently, the incidence of TB is a crucial determinant of agreement between the two tests.

In this study the concordance between the TST and the IGRA was also affected by immune status. There was only slight agreement in patients taking concurrent prednisolone, but the agreement improved when these patients were excluded from the analysis. Some previous studies of immunosuppressed RA patients have shown similarly poor concordance between the two tests regardless of TB burden.19 20 There was also discordance between the two tests in LTBI diagnosis among individuals infected with the human immunodeficiency virus.21 The patients on prednisolone in our study had a lower rate of positivity for both tests. It seems intuitive to assume that an immunosuppressed state will induce a depressed response to a T-cell reaction. In the literature, a systematic review showed that both positive IGRA and positive TST results were significantly influenced by immunosuppressive therapy.22

The Hong Kong guideline for the TST cut-off value for LTBI diagnosis before anti-tumour necrosis factor treatment is an induration of >10 mm.3 In the current study, we showed that the TST cut-off value that achieved better agreement between IGRA and TST results was 5 mm. If we cannot rely on IGRA to diagnose LTBI, it may be more appropriate to lower that TST cut-off to 5 mm. We also showed that our approach to LTBI screening with both TST and IGRA was successful in preventing the development of active TB in patients who would receive biologic therapy. This dual testing strategy might be especially applicable to patients on systemic steroid, as they are at higher risk of developing active TB and the two test results are more discordant. The consequent improved sensitivity will invariably lower the specificity and cause a potential overtreatment. In these high-risk settings, however, it is reasonable to favour sensitivity in screening for LTBI.

Only five patients could give a definite history of BCG vaccination. For other patients, such information was uncertain. While this might reflect the local clinical situation, it is one of the limitations of the present study. Despite the mechanistic similarity, because of the different interpretation methods employed for the test results and the lack of comparative trials of the performance of A.TB IGRA and other IGRAs, the conclusions drawn from the current study may not be applicable to patients who are given a different IGRA. Further studies using individual IGRAs may be warranted to address the same question.

In arthritis patients in a TB-endemic population, the level of agreement between TST and A.TB IGRA for detecting LTBI was only fair. Although 10 mm is the usual cut-off for TST, the level of agreement between the two tests improved from fair to moderate when a 5-mm cut-off was used. A dual testing strategy with TST and IGRA appeared to be effective and should be pursued, especially in patients who are prescribed a systemic steroid. The issue of potential overtreatment is yet to be evaluated.

All authors have disclosed no conflicts of interst.

1. Furst DE. The risk of infections with biologic therapies for rheumatoid arthritis. Semin Arthritis Rheum 2010;39:327-46. Crossref

2. World Health Organization. Guidelines on the management of latent tuberculosis infection. Geneva, Switzerland: World Health Organization; 2015.

3. Mok CC. Consensus statements on the indications and monitoring of anti-tumor necrosis factor (TNF) therapy for rheumatic diseases in Hong Kong. Hong Kong Bull Rheum Dis 2005;5:19-25.

4. Huebner RE, Schein MF, Bass JB Jr. The tuberculin skin test. Clin Infect Dis 1993;17:968-75. Crossref

5. Matulis G, Jüni P, Villiger PM, Gadola SD. Detection of latent tuberculosis in immunosuppressed patients with autoimmune diseases: performance of a Mycobacterium tuberculosis antigen-specific interferon gamma assay. Ann Rheum Dis 2008;67:84-90. Crossref

6. Ponce de Leon D, Acevedo-Vasquez E, Alvizuri S, et al. Comparison of an interferon-gamma assay with tuberculin skin testing for detection of tuberculosis (TB) infection in patients with rheumatoid arthritis in a TB-endemic population. J Rheumatol 2008;35:776-81.

7. Song Q, Guo H, Zhong H, et al. Evaluation of a new interferon-gamma release assay and comparison to tuberculin skin test during a tuberculosis outbreak. Int J Infect Dis 2012;16:e522-6. Crossref

8. Ferrara G, Losi M, Meacci M, et al. Routine hospital use of a new commercial whole blood interferon-gamma assay for the diagnosis of tuberculosis infection. Am J Respir Crit Care Med 2005;172:631-5. Crossref

9. Kleinert S, Tony HP, Krueger K, et al. Screening for latent tuberculosis infection: performance of tuberculin skin test and interferon-gamma release assays under real-life conditions. Ann Rheum Dis 2012;71:1791-5. Crossref

10. Mariette X, Baron G, Tubach F, et al. Influence of replacing tuberculin skin test with ex vivo interferon gamma release assays on decision to administer prophylactic antituberculosis antibiotics before anti-TNF therapy. Ann Rheum Dis 2012;71:1783-90. Crossref

11. Winthrop KL, Weinblatt ME, Daley CL. You can’t always get what you want, but if you try sometimes (with two tests—TST and IGRA—for tuberculosis) you get what you need. Ann Rheum Dis 2012;71:1757-60. Crossref

12. Cao H, Agrawal D, Kushner N, Touzjian N, Essex M, Lu Y. Delivery of exogenous protein antigens to major histocompatibility complex class I pathway in cytosol. J Infect Dis 2002;185:244-51. Crossref

13. McEvers K, Elrefaei M, Norris P, et al. Modified anthrax fusion proteins deliver HIV antigens through MHC class I and II pathways. Vaccine 2005;23:4128-35. Crossref

14. Pai M, Zwerling A, Menzies D. Systematic review: T-cell–based assays for the diagnosis of latent tuberculosis infection: an update. Ann Intern Med 2008;149:177-84. Crossref

15. Yilmaz N, Zehra Aydin S, Inanc N, Karakurt S, Direskeneli H, Yavuz S. Comparison of QuantiFERON-TB Gold test and tuberculin skin test for the identification of latent Mycobacterium tuberculosis infection in lupus patients. Lupus 2012;21:491-5. Crossref

16. Lee JH, Sohn HS, Chun JH, et al. Poor agreement between QuantiFERON-TB Gold test and tuberculin skin test results for the diagnosis of latent tuberculosis infection in rheumatoid arthritis patients and healthy controls. Korean J Intern Med 2014;29:76-84. Crossref

17. Ewer K, Deeks J, Alvarez L, et al. Comparison of T-cell-based assay with tuberculin skin test for diagnosis of Mycobacterium tuberculosis infection in a school tuberculosis outbreak. Lancet 2003;361:1168-73. Crossref

18. Brock I, Weldingh K, Lillebaek T, Follmann F, Andersen P. Comparison of tuberculin skin test and new specific blood test in tuberculosis contacts. Am J Respir Crit Care Med 2004;170:65-9. Crossref

19. Huang YW, Shen GH, Lee JJ, Yang WT. Latent tuberculosis infection among close contacts of multidrug-resistant tuberculosis patients in central Taiwan. Int J Tuberc Lung Dis 2010;14:1430-5.

20. Shalabi NM, Houssen ME. Discrepancy between the tuberculin skin test and the levels of serum interferon-gamma in the diagnosis of tubercular infection in contacts. Clin Biochem 2009;42:1596-601. Crossref

21. Mandalakas AM, Hesseling AC, Chegou NN, et al. High level of discordant IGRA results in HIV-infected adults and children. Int J Tuberc Lung Dis 2008;12:417-23.

22. Shahidi N, Fu YT, Qian H, Bressler B. Performance of interferon-gamma release assays in patients with inflammatory bowel disease: a systematic review and meta-analysis. Inflamm Bowel Dis 2012;18:2034-42. Crossref

Oesophageal perforation is uncommon, yet its management remains a substantial challenge to surgeons. Diagnosis and treatment are often delayed due to lack of clinical suspicion and accurate diagnostic tools. Hence, reported mortality rates range from 10% to 25%.1 2 3

Oesophageal perforation can occur spontaneously from forceful vomiting (Boerhaave’s syndrome), or in pre-existing pathology (such as oesophageal cancer) or can be associated with ingestion of a foreign body. Iatrogenic perforation usually occurs after therapeutic endoscopic procedures such as dilatation, and is the predominant cause of perforation reported in many studies.1 2 4 5

Diagnosis and treatment within 24 hours of perforation are critical if favourable outcomes are to be achieved.1 6 After diagnosis and the initial phase of resuscitation, there is a wide range of treatment options, which are informed by the presentation, aetiology, location of perforation, and the extent of mediastinal or intrathoracic contamination. Surgery remains the mainstay of treatment; the conventional operative approach is considered to be primary repair of the perforation site and drainage.7 8 9 Some surgeons advocate primary repair only for those patients presented within 24 hours of perforation,10 while others would try primary repair as the initial treatment regardless of the timing of presentation.9 11 Endoscopic treatment, including stenting, is becoming an increasingly popular means of treating oesophageal perforation in selected patients, and reportedly has a high technical success rate.12 13 14 15 16

Oesophageal perforation should be managed in specialised centres. In this study, we report the characteristics, treatment, and outcomes of a cohort of patients with oesophageal perforation treated at a single tertiary centre in Hong Kong over a period of 16 years.

We retrospectively identified patients treated for perforation of the oesophagus at a university teaching hospital in Hong Kong between January 1997 and December 2013. Patients’ demographic characteristics, presentation, investigations, management, and outcomes were recorded.

Diagnosis of perforation was confirmed by one or more of the following methods: oesophagogastroduodenoscopy (OGD), water-soluble contrast swallow study, and contrast-enhanced computed tomography imaging of the neck, thorax, and abdomen. After confirmation of the diagnosis, patients were resuscitated to address homoeostatic and haemodynamic disturbances, followed by definitive treatment. All patients were kept ‘nil by mouth’, administered parenteral broad-spectrum antibiotics and proton pump inhibitors, and chest drain(s) was inserted if clinically indicated. Patients with significant haemodynamic instability or respiratory distress requiring intubation and mechanical ventilation were admitted to the intensive care unit (ICU) for optimisation before definitive treatment.

Definitive treatment depended on the location of the perforation, its aetiology, the extent of mediastinal and intrathoracic contamination, and the patient’s physical status. In general, patients with malignant perforation or perforation contained within the mediastinal pleura were treated non-operatively. For the former, self-expanding metallic stents were inserted under fluoroscopic guidance. In selected patients with a benign cause of perforation and limited contamination, a polyester oesophageal stent (Polyflex; Boston Scientific, Natick [MA], United States) was placed under fluoroscopic guidance. For patients in whom the site of perforation could not be identified, and in the absence of clinical signs of sepsis, a conservative management strategy was adopted. This entailed placement of a nasogastric feeding tube under endoscopic guidance, followed by enteral feeding for 7 days. Thereafter, a water-soluble contrast swallow study was undertaken to confirm the absence of a leak before oral feeding was resumed.

When a surgical management strategy was decided, patients with perforation of the intra-abdominal oesophagus were treated with laparotomy, primary repair of the perforation, and feeding jejunostomy. For an intrathoracic perforation with significant contamination of the pleural cavity, thoracotomy and primary repair was the preferred approach. A left-sided thoracotomy was the usual approach for Boerhaave’s perforation of the distal thoracic oesophagus. Necrotic tissue was debrided, the edges of the perforation were trimmed, and the defect was closed with fine sutures in two layers. The mucosal edges of the perforation were approximated using interrupted absorbable sutures, and the muscular defect was approximated using interrupted monofilament absorbable sutures. Lung decortication was performed. One drain was placed in close proximity to the repair, generally accompanied by one basal and one apical large-bore chest drain. Feeding jejunostomy was performed in selected patients. Postoperatively patients remained nil by mouth, and were given nutritional support and intravenous antibiotics. A contrast swallow study was generally performed 7 to 10 days postoperatively; oral intake was commenced if there was no evidence of leak. The choice of antibiotics and duration of treatment were guided by microbiology culture findings.

In selected patients who presented late, and in those who developed a persistent leak after primary repair, oesophageal exclusion (cervical oesophagostomy and jejunostomy) followed by second-stage oesophagectomy might be considered. In the first stage, the oesophagus was excluded proximally in the neck with an oesophagostomy, and the abdominal oesophagus was stapled. A drain was placed from the neck into the oesophageal stump for decompression. Oesophagectomy was performed once sepsis had subsided. A gastric tube was used for reconstruction via the retrosternal route, and cervical oesophagogastrostomy was performed.

The principles outlined in the Declaration of Helsinki have been followed.

Continuous data were represented as the median (range), unless otherwise stated. Fisher’s exact test was used to compare categorical variables and the Mann-Whitney U test for continuous variables. We undertook univariate analysis to identify factors associated with hospital mortality. P<0.05 was considered statistically significant. Data were analysed using SPSS 20.0 (IBM Corp, Armonk [NY], United States).

During the study period, 43 patients with oesophageal perforation were identified. Patients’ demographic and clinical characteristics are summarised in Table 1. The median age of the cohort was 66 years (range, 30-98 years); 28 (65.1%) were men. Medical co-morbidities were present in 27 (62.8%) patients, and pre-existing oesophageal pathologies were present in 14 (32.6%; of whom half had oesophageal cancer). Spontaneous perforation occurred in 22 (51.2%) patients: 15 occurred as a result of Boerhaave’s syndrome and seven as a result of malignant perforation. Fifteen (34.9%) patients had an iatrogenic perforation: 13 occurred after an endoscopic procedure (three after endoscopic retrograde cholangiopancreatography and 10 after OGD), one occurred after attempted endotracheal intubation, and one occurred during thyroidectomy. Of the 10 OGDs, eight had been therapeutic. Six (14.0%) perforations were associated with ingestion of a foreign body.

Chest pain and vomiting were the most common presenting symptoms in patients with spontaneous perforation, occurring in 13 and 10 patients, respectively. Surgical emphysema and dysphagia were the least common presenting signs and symptoms; both were only present in two patients. Over half the patients presented and were diagnosed within 24 hours of symptom onset. Of the cohort of 43 patients, 29 (67.4%) underwent two out of the three diagnostic imaging modalities. The presenting symptoms and investigations of patients with spontaneous perforation are shown in Table 2.

The management and outcomes of patients are shown in the Figure. Of the 15 patients with Boerhaave’s perforation, 10 underwent primary repair: four repairs were complicated by a leak and two patients subsequently required oesophagectomy. The remaining five patients were initially treated non-operatively: four underwent endoscopic stent placement and one endoscopic clipping of the perforation. Three patients required subsequent operations: one underwent oesophagectomy, one bypass operation, and one surgical drainage. There were no deaths in the group of patients with Boerhaave’s syndrome.

Seven patients had malignant perforation: five were treated with endoscopic placement of a metallic stent. All but one of these procedures were successful; the patient in whom stenting failed underwent oesophagectomy. Five (71.4%) of the seven patients with malignant perforation died during their hospital stay.

There were 15 iatrogenic perforations (Fig). Nine of these patients underwent early operative treatment: five underwent primary repair, one exclusion, two drainage, and one oesophagectomy. There were no leaks in those who underwent primary repair. Six patients were initially treated non-operatively, four with stents, one with a feeding tube, and one was judged to be unfit for treatment. Two of the six patients initially treated non-operatively ultimately required surgery, one underwent exclusion, and the other surgical drainage. Five of the 15 patients with iatrogenic perforations died during their hospital stay, with a mortality rate of 33.3%.

Six oesophageal perforations were associated with foreign body ingestion (Fig). Three patients were treated non-operatively; of the remainder, one underwent primary repair, one exclusion, and one surgical drainage. None of these patients died during hospitalisation.

Overall, 16 of the 43 patients underwent primary repairs in the initial treatment, and four (25%) developed clinical signs of leak subsequently. All were from Boerhaave’s perforation. Two required oesophagectomy while two were managed conservatively.

Overall, six of the 43 patients underwent oesophagectomy, generally as a salvage treatment due to failure of other treatment modalities. Three patients with Boerhaave’s syndrome required oesophagectomy, two with persistent leak after primary repair and one with a persistent leak after stenting. All had presented >24 hours from symptom onset. One patient with a perforated oesophageal cancer developed a leak after stenting and required oesophagectomy. Two patients with iatrogenic perforation in the presence of caustic strictures underwent oesophagectomy. Only one patient who underwent oesophagectomy died in hospital.

Overall, 10 patients died in hospital, with a mortality rate of 23.3%. The 30-day mortality rate was 16.3%. The median length of hospital stay was 36.5 days (range, 6-241 days), and median ICU stay was 6 days (range, 0-71 days).

All 10 patients who died had pre-existing oesophageal disease; five had cancer of the oesophagus, one caustic stricture, and four had oesophageal varices secondary to hepatic cirrhosis. Malignant perforation had a substantially higher mortality rate of 71.4%. The median survival for patients with perforated oesophageal cancer was 28.5 days (range, 13-848 days).

The results of univariate analysis of factors potentially associated with hospital mortality are shown in Table 3. The presence of pulmonary disease, hepatic disease (liver cirrhosis), and malignant perforation were significantly associated with hospital mortality (P=0.03, <0.01, and <0.01, respectively), but the site of perforation and timing of presentation were not.

Oesophageal perforation may be difficult to diagnose. Patients can present with a wide variety of symptoms, which can be non-specific. It is not uncommon for the diagnosis to be missed in the acute phase. Computed tomography imaging (preferably with oral contrast) should be undertaken when the index of clinical suspicion is high, because it allows the site of mediastinal or intra-abdominal collections to be identified and rules out other pathologies. Of note, OGD performed by an experienced endoscopist using minimal insufflation is an effective means of detecting the site and size of perforation, and is reported to have a sensitivity and specificity of 100% and 83% for intrathoracic perforation, respectively.17 A positive OGD therefore has a substantial influence on clinical decision making.

Spontaneous perforation was the most common aetiology in our cohort; around one third was associated with underlying cancer of the oesophagus. Squamous cell carcinoma remains the most common malignant cell type globally, despite the rising incidence of adenocarcinoma in the western population. Patients often present at an advanced stage. Of those patients with malignant perforations, all but one had a squamous cell carcinoma of the intrathoracic oesophagus. Perforation either occurs spontaneously or results from concurrent chemoradiotherapy. Ohtsu et al18 reported a perforation rate of 13.9% (five out of 36 patients) in cases of T4-stage cancer of the oesophagus with concurrent chemoradiotherapy. In our cohort, perforation occurred shortly after completion of radiotherapy in one patient.

The prognosis for patients with perforated oesophageal cancer is poor. The disease is often inoperable and in these circumstances treatment is palliative.19 20 Non-operative treatment, such as insertion of a metallic covered stent, is the usual practice at our centre. Stenting of the intrathoracic portion of the oesophagus is technically straightforward and is successful in most cases. Sealing of the perforation site can be confirmed by a subsequent contrast study, and oral intake can be resumed in the absence of a leak. Nevertheless, the prognosis of this group of patients is poor despite the successful placement of a stent, and the hospital mortality rate remains high. Patients most often succumb as a consequence of sepsis caused by the perforation.

Many treatment options are available for non-malignant perforation, and the treatment strategy should be tailored to the individual. Factors to be considered include the site of perforation, extent of contamination, pre-existing oesophageal disease, and patient co-morbidities. Operative treatment is favoured for perforation of the intra-abdominal oesophagus or perforation that involves the oesophagogastric junction (OGJ). These patients often present with abdominal pain and peritonitis. Laparotomy, primary repair of the perforation, and fashioning of a feeding jejunostomy allow alimentation in the event of persistent leak. The placement of an oesophageal stent that crosses the OGJ has a higher chance of migration and is not recommended.

The intrathoracic oesophagus is the most common site of perforation. Of the three most common causes (Boerhaave’s syndrome, iatrogenic perforation, and foreign body ingestion), Boerhaave’s syndrome is the most challenging. Traditionally, Boerhaave’s syndrome is associated with a mortality rate of up to 30%.11 Patients may present late, the site of perforation is usually at the distal thoracic oesophagus, and there may be extensive contamination due to the high pressure generated by vomiting. Contamination with food particles is common. Operative treatment with primary closure of the perforation and drainage is favoured by many7 8 9; this is also our preferred approach. Many surgeons advocate primary repair irrespective of the timing of presentation.9 11 21 22 Leak rates after primary repair range from 17% to 32%.9 11 21 22 23 24 Minor leaks can be managed conservatively with drainage, while further surgery (usually exclusion) is required for larger leaks and in the presence of sepsis. Lin et al23 reported that the incidence of postoperative leak was 37.5% in patients in whom treatment was delayed for more than 48 hours, compared with 0% in those who were treated more promptly. Wright et al22 reported that three out of the four leaks in their patient cohort were repaired more than 24 hours after perforation. The incidence of leak after primary repair was 25.0% in our study, which is comparable to other reports in the literature. Of the four leaks, two patients required reoperation and ultimately oesophagectomy; both had presented more than 24 hours after symptom onset.

Endoscopic stenting for benign perforation has been reported in several small case series. Freeman et al13 14 have reported the outcomes of stent placement in patients with iatrogenic and spontaneous perforation. They proposed a hybrid approach, namely a combination of endoscopic and minimally invasive surgical techniques to drain intrathoracic and/or intra-abdominal collections. The main advantage of this strategy is the avoidance of thoracotomy and/or laparotomy. The incidence of stent migration was reported to be approximately 20% in their cohort of patients with spontaneous oesophageal perforation.14 Relative contra-indications to stent insertion include a perforation that crosses the OGJ and circumferential necrosis of the oesophagus. In our experience, operative treatment is recommended for the treatment of Boerhaave’s syndrome unless the patient is unfit for surgery or declines surgical treatment. Five patients in our series initially treated with stenting subsequently required surgery, of whom four had benign perforations (two with Boerhaave’s syndrome and two with iatrogenic perforations). One patient in our cohort with oesophageal dissection complicated by perforation underwent stenting in another hospital before transferring to our centre; this patient developed a persistent leak after stenting. In that case, the placement of the stent appeared to have aggravated the leak, and oesophagectomy was eventually required.25 In our opinion, stent placement in benign perforation is only suitable for selected patients who present early and have minimal contamination. However, stenting may allow more time for optimisation of a patient’s condition if they are initially judged not to be fit for surgery.

Oesophagectomy as a treatment for perforations was first reported in the 1950s.26 Single-stage oesophageal resection and reconstruction was first reported by Hendren and Henderson in 1968.27 Altorjay et al28 reported a hospital mortality rate of 3.7% in a series of patients undergoing oesophagectomy for intrathoracic perforation; in this series iatrogenic perforation represented 55.6% of all perforations. Some surgeons have opined that oesophagectomy may be superior to primary repair in the presence of pre-existing oesophageal disease and of extensive perforation with substantial sepsis, while the general condition of the patient should always be taken into account.28 29 There is no consensus about the optimum surgical approach and timing of reconstruction after oesophagectomy. We advocate primary repair as the initial treatment irrespective of the timing of presentation, and oesophagectomy is considered a salvage treatment. In our experience, patients with persistent leak after primary repair and sepsis should undergo oesophageal exclusion to control sepsis before oesophagectomy is contemplated. Oesophagectomy with primary reconstruction can be performed safely after patient optimisation. Oesophagectomy was undertaken in six patients in our cohort; three of these patients had pre-existing oesophageal disease. All patients had a cervical oesophagogastric anastomosis fashioned via the retrosternal route. A cervical anastomosis distant from the infected mediastinum appears to be a safe option.29 Thoracotomy is the most common surgical approach, but Yeo et al30 reported using transhiatal oesophagectomy to treat perforated oesophageal cancer in four patients. Thoracotomy is avoided in the transhiatal approach, but this technique can only be considered in perforations of the distal oesophagus and in the presence of minimal mediastinal contamination.

Oesophageal perforation after foreign body ingestion in adults is more common in China as a result of its dietary culture. The foreign body is usually a fish, chicken, or pork bone. An impacted foreign body can usually be retrieved endoscopically; however, oesophageal perforation can occur if there is deep penetration of the foreign body or extensive manipulation during retrieval. The site of perforation is usually the cervical oesophagus, followed by the intrathoracic oesophagus. In severe cases, operative management is indicated; the approach is dependent on the site of perforation, and the site and size of any collection. The aim of management is to drain any collection, remove any residual foreign body, repair the perforated site, and protect the airway. In the absence of sepsis and imaging appearances of a peri-oesophageal collection, conservative treatment may be warranted. Operative drainage may be necessary if there is a sizeable collection and if there is sepsis. Mediastinitis and sepsis are more likely after intrathoracic perforation, and would dictate treatment strategy.

It is essential to identify factors associated with mortality after oesophageal perforation so as to improve treatment and outcomes. Early diagnosis and management (in the ‘golden 24 hours’) are reportedly associated with superior outcomes.1 6 Malignant perforation, sepsis, the need for mechanical ventilation on presentation, and pulmonary co-morbidity are reported to have a significant impact on overall survival.5 In our cohort, pulmonary co-morbidity, hepatic disease, and malignant perforation were associated with risk of death. A recent meta-analysis of 75 studies that included 2971 patients reported a pooled mortality rate of 11.9% (95% confidence interval, 9.7%-14.3%).3 Of the different aetiologies, spontaneous perforation had the highest mortality rate of 14.8%.3

Oesophageal perforation remains a difficult condition to treat despite advances in surgery, endoscopic treatment, and ICU care. The mortality rate is still substantial with modern therapies. The presence of pre-existing pulmonary disease, hepatic disease, and perforation associated with malignancy was significantly associated with hospital mortality in our cohort. Oesophagectomy for salvage had a reasonable success rate in selected patients.

The authors have disclosed no conflicts of interest.

1. Vallböhmer D, Hölscher AH, Hölscher M, et al. Options in the management of esophageal perforation: analysis over a 12-year period. Dis Esophagus 2010;23:185-90. Crossref

2. Søreide JA, Konradsson A, Sandvik OM, Øvrebø K, Viste A. Esophageal perforation: clinical patterns and outcomes from a patient cohort of Western Norway. Dig Surg 2012;29:494-502. Crossref

3. Biancari F, D’Andrea V, Paone R, et al. Current treatment and outcome of esophageal perforations in adults: systematic review and meta-analysis of 75 studies. World J Surg 2013;37:1051-9. Crossref

4. Abbas G, Schuchert MJ, Pettiford BL, et al. Contemporaneous management of esophageal perforation. Surgery 2009;146:749-55. Crossref

5. Bhatia P, Fortin D, Inculet RI, Malthaner RA. Current concepts in the management of esophageal perforations: a twenty-seven year Canadian experience. Ann Thorac Surg 2011;92:209-15. CrossRef

6. Shaker H, Elsayed H, Whittle I, Hussein S, Shackcloth M. The influence of the ‘golden 24-h rule’ on the prognosis of oesophageal perforation in the modern era. Eur J Cardiothorac Surg 2010;38:216-22. Crossref

7. Brinster CJ, Singhal S, Lee L, Marshall MB, Kaiser LR, Kucharczuk JC. Evolving options in the management of esophageal perforation. Ann Thorac Surg 2004;77:1475-83. Crossref

8. Eroglu A, Can Kürkcüogu I, Karaoganogu N, Tekinbaş C, Yimaz O, Başog M. Esophageal perforation: the importance of early diagnosis and primary repair. Dis Esophagus 2004;17:91-4. Crossref

9. Jougon J, Mc Bride T, Delcambre F, Minniti A, Velly JF. Primary esophageal repair for Boerhaave’s syndrome whatever the free interval between perforation and treatment. Eur J Cardiothorac Surg 2004;25:475-9. Crossref

10. Flynn AE, Verrier ED, Way LW, Thomas AN, Pellegrini CA. Esophageal perforation. Arch Surg 1989;124:1211-4. Crossref

11. Lawrence DR, Ohri SK, Moxon RE, Townsend ER, Fountain SW. Primary esophageal repair for Boerhaave’s syndrome. Ann Thorac Surg 1999;67:818-20. Crossref

12. Fischer A, Thomusch O, Benz S, von Dobschuetz E, Baier P, Hopt UT. Nonoperative treatment of 15 benign esophageal perforations with self-expandable covered metal stents. Ann Thorac Surg 2006;81:467-72. Crossref

13. Freeman RK, Van Woerkom JM, Ascioti AJ. Esophageal stent placement for the treatment of iatrogenic intrathoracic esophageal perforation. Ann Thorac Surg 2007;83:2003-7. Crossref

14. Freeman RK, Van Woerkom JM, Vyverberg A, Ascioti AJ. Esophageal stent placement for the treatment of spontaneous esophageal perforations. Ann Thorac Surg 2009;88:194-8. Crossref

15. Kiernan PD, Khandhar SJ, Fortes DL, Sheridan MJ, Hetrick V. Thoracic esophageal perforations. Am Surg 2010;76:1355-62.

16. Dasari BV, Neely D, Kennedy A, et al. The role of esophageal stents in the management of esophageal anastomotic leaks and benign esophageal perforations. Ann Surg 2014;259:852-60. Crossref

17. Horwitz B, Krevsky B, Buckman RF Jr, Fisher RS, Dabezies MA. Endoscopic evaluation of penetrating esophageal injuries. Am J Gastroenterol 1993;88:1249-53.

18. Ohtsu A, Boku N, Muro K, et al. Definitive chemoradiotherapy for T4 and/or M1 lymph node squamous cell carcinoma of the esophagus. J Clin Oncol 1999;17:2915-21.

19. Di Franco F, Lam PJ, Karat D, Hayes N, Griffin SM. Iatrogenic perforation of localized oesophageal cancer. Br J Surg 2008;95:837-9. Crossref

20. Jethwa P, Lala A, Powell J, McConkey CC, Gillison EW, Spychal RT. A regional audit of iatrogenic perforation of tumours of the oesophagus and cardia. Aliment Pharmacol Ther 2005;21:479-84. Crossref

21. Whyte RI, Iannettoni MD, Orringer MB. Intrathoracic esophageal perforation. The merit of primary repair. J Thorac Cardiovasc Surg 1995;109:140-4. Crossref

22. Wright CD, Mathisen DJ, Wain JC, Moncure AC, Hilgenberg AD, Grillo HC. Reinforced primary repair of thoracic esophageal perforation. Ann Thorac Surg 1995;60:245-8. Crossref

23. Lin Y, Jiang G, Liu L, et al. Management of thoracic esophageal perforation. World J Surg 2014;38:1093-9. Crossref

24. Richardson JD. Management of esophageal perforations: the value of aggressive surgical treatment. Am J Surg 2005;190:161-5. Crossref

25. Zhu RY, Law TT, Tong D, Tam G, Law S. Spontaneous circumferential intramural esophageal dissection complicated with esophageal perforation and esophageal-pleural fistula: a case report and literature review. Dis Esophagus 2016;29:872-9. Crossref

26. Johnson J, Schwegman CW, MacVaugh H III. Early esophagogastrectomy in the treatment of iatrogenic perforation of the distal esophagostomy. J Thorac Cardiovasc Surg 1956;32:827-31.

27. Hendren WH, Henderson BM. Immediate esophagectomy for instrumental perforation of the thoracic esophagus. Ann Surg 1968;168:997-1003. Crossref

28. Altorjay A, Kiss J, Vörös A, Szirányi E. The role of esophagectomy in the management of esophageal perforation. Ann Thorac Surg 1998;65:1433-6. Crossref

29. Orringer MB, Stirling MC. Esophagectomy for esophageal disruption. Ann Thorac Surg 1990;49:35-42. Crossref

30. Yeo CJ, Killemoe KD, Klein AS, Zinner MJ. Treatment of instrumental perforation of esophageal malignancy by transhiatal esophagectomy. Arch Surg 1988;123:1016-8. Crossref