Professor K. McPherson, Nuffield Department of Obstetrics and Gynaecology, Research Institute, Churchill Hospital, Oxford, OX3 7LJ UK.
Objectives To model the determinants of serious operative and post-operative complications of hysterectomy and their potential risk factors.
Design A prospective cohort of women undergoing hysterectomies for benign indications in 1994/1995, with a six-week postsurgery follow up.
Population and setting A total of 37,512 women from 276 NHS and 145 private hospitals in England, Wales and Northern Ireland, originally recruited to compare the outcomes of endometrial destruction with those of hysterectomy.
Methods Gynaecologists reported hysterectomies for non-malignant indications carried out during a 12-month period beginning in October 1994 and follow up data were obtained at outpatient follow up six weeks postsurgery. Odds ratios of severe complications by indication and method, adjusting for measured intrinsic risk factors, were calculated.
Main outcome measures Severe operative and post-operative complications.
Results Severe operative complications occurred in 3%. The risk decreased with age and increased with greater parity and history of serious illness. Women with symptomatic fibroids (4.4%, 95% CI 3.9–4.9) experienced more complications than women with dysfunctional uterine bleeding (3.6%, 3.2–3.8), adjusted odds ratio (OR) = 1.3 (95% CI 1.1–1.6). Laparoscopic procedures (6.1%) doubled the risk of operative complications of abdominal hysterectomy (3.6%) (adjusted OR = 1.9, 1.5–2.5). Post-operative complications occurred in around 1% of women, with a slight decrease with increasing age, and the strongest risk factor was a history of operative complications. Relative to dysfunctional uterine bleeding (1.0%), a higher risk for fibroids (1.2%) persisted after adjustments (RR = 1.5, 1.1–2.0). Both vaginal (1.2%) and laparoscopic (1.7%) techniques had significantly higher adjusted risks than abdominal operations (0.9%), RR = 1.4 (1.0–1.9) and RR = 1.6 (1.0–2.7). There were no operative deaths; 14 women died within the six-week postsurgery (a crude mortality rate of 3.8/1000, 2.5–6.4).
Conclusions Hysterectomy is a common, routine surgery with comparatively rare serious complications. However, younger women, women with more vascular pelvises, who undergo hysterectomy, especially laparoscopically assisted vaginal surgery for symptomatic fibroids, are at most risk of experiencing severe complications both operatively and post-operatively. Therefore, a less invasive alternative treatment for symptomatic fibroids could particularly benefit this group of women, while less invasive treatments for dysfunctional uterine bleeding, such as various methods of endometrial ablations or resections, would need to meet the current low levels of clinical complications in order to replace hysterectomy.
Hysterectomy is the most common major gynaecological operation with around 100,000 procedures performed annually in the UK.1 Approximately 20% of women have had the procedure by the age of 60 years; about 40% of these for dysfunctional uterine bleeding with no gynaecological pathology. The VALUE (Vaginal Abdominal or Laparoscopic Uterine Excision) study was designed to collect data on all hysterectomies that took place in the England, Northern Ireland and Wales between October 1994 and September 19952 in order to obtain an unselected group of hysterectomy cases to act as a control for women who were recruited into the MISTLETOE (Minimally Invasive Surgical Techniques: Laser Endothermal or Endoresection) study on endometrial resection/ablation.3 The study was designed also to investigate the use of hysterectomy in routine gynaecological practice and the complications associated with different procedures, routinely and inexpensively.4 This paper describes the reported incidence of severe operative and post-operative complications, and their possible determinants.
Data collection was similar to the MISTLETOE study.3 In brief, in 1994 staff at the Clinical Audit Unit of the Royal College of Obstetricians and Gynaecologists (RCOG) contacted all consultant gynaecologists at 276 National Health Service (NHS) and 145 independent hospitals in England, Wales, Northern Ireland, the Channel Islands and the Isle of Man. They were asked to supply information on all hysterectomies for benign causes performed between October 1994 and September 1995 inclusive.2 Of these, 80% of NHS and 95% independent hospitals responded. Among respondents, staff completed questionnaires at the time of the operation, at discharge and six weeks postsurgery, and sent them to the RCOG Audit Unit. The responses comprising these data were validated after the study period by selecting 30 hospitals at random, stratified equally into three reporting categories and systematically reviewing hospital case notes for prescribed study periods. The most plausible source of bias in these data is (consultant) response bias.
Pre-operative information included patient identifiers, relevant past medical history, previous gynaecological management and indications for hysterectomy. Operative information included details of hysterectomy method used, operative findings and complications. Post-operative complications and histology, if available, were reported at discharge and/or outpatient follow up at six weeks postsurgery.
We investigated the associations between severe operative and post-operative complications and age, comorbidity, the indication for surgery, pre-operative use of antibiotics, grade of surgeon, grade of supervisor and method of hysterectomy. Severe complications were death, deep venous thrombosis, pulmonary embolism, myocardial infarction, renal failure, cerebrovascular accident, septicaemia, necrotising fasciitis, secondary haemorrhage, fistula, ureteric obstruction and visceral damage. Post-operative complications were categorised as early (reported before hospital discharge), or late (reported after discharge and before the six-week hospital check up). We assumed that there were no late post-operative complications for women with no medical event form.
Ethical approval for the study was obtained from the ethical committees of six (then) District Health Authorities and the London School of Hygiene and Tropical Medicine. The study was also approved by the Royal College of Obstetricians and Gynaecologists and the British Medical Association.
Simple univariate tests of proportions of the complete data between main groups preceded adjustment for confounders. SPSS and STATA were used to model the proportion of women who experienced operative complications. Logistic regression analyses calculated odds ratios (OR) for severe operative complications (presence or absence), according to method, indication and age at surgery, history of serious illness, grade and supervision of surgeon, and their interactions. As there were several possible post-operative complications, Poisson regression was used for analysing each of them, controlling for the same set of confounding variables, as well as operative complications and prophylactic antibiotic use. All reported risks were adjusted for other risk factors, and hence, were likely to represent the main effect attributable to that subgroup. Otherwise, the risks of several independent factors were multiplicative. The baseline category was the most prevalent group in the cohort: women aged 40–49, with dysfunctional uterine bleeding, with no history of relevant co-morbidity having abdominal hysterectomy performed by a non-supervised consultant surgeon.
The number of hysterectomies reported was 37,512, approximately 40% of the total in the study time period. After excluding 217 cases (209 women with cancer, 6 postpartum hysterectomies and 2 who did not have hysterectomies), the number of cases was 37,295. There were 26,973 medical event forms at the six-week follow up (72%). The study sample was broadly representative of women in the reference population in terms of age, indications and surgical method. Validation process uncovered some disparities concerning interpretations of the ‘main’ indication. However, as case notes suffered from a degree of ambiguity in post hoc interpretation, we felt it best to rely on the data forms uniformly submitted at surgery (both available for 429 women), except where case notes appeared contemporaneously unambiguous.
The sample has been described in detail elsewhere.2 Briefly, the median age was 45 years (range 12–95); 88% were married or cohabiting, and 27% were current smokers. Dysfunctional uterine bleeding was the most common indication (46%). Other indications were: fibroids (19%), prolapse (19%), endometriosis and adenomyosis (5%), pelvic mass (3%) and other miscellaneous (8%). Consultants carried out 58% of hysterectomies, and of those attributed to non-consultants, 34% were supervised. Of unsupervised hysterectomies, senior house officers carried out less than 2%. One hundred and fifty-two of 194 consultants used laparoscopic methods. Eleven percent of laparoscopic operations were performed by non-consultants, of which 65% were recorded as unsupervised, while 3% overall were supervised by consultants.
For 37,295 cases with data, the proportions of abdominal (total and subtotal), vaginal and laparoscopically assisted vaginal hysterectomy were respectively 67%, 30% and 3%. Antibiotic prophylaxis was used in 72% of operations, with single shots most commonly used (59%). A full course was rare (<1%).
Fourteen deaths were reported by six weeks postsurgery, a mortality rate of 0.38 per thousand (0.25–0.64). None died in theatre and eight died before hospital discharge. The median age at death was 58 years. No deaths were associated with the laparoscopic technique.
The number of women, aged over 20 at surgery, suffering operative complication was 1295/37,173 (3.5%). There was no difference according to whether consultants (3.5%, 716/20,426) or non-consultants (3.4%, 508/14,772) operated, P(χ21) = 0.74. The proportions of operative complications for supervised and non-supervised procedures were 4.0% (238/5902) and 3.4% (1046/31,007), P(χ21) = 0.01. Visceral damage was mostly associated with the bladder (220 bladder, 19 ureter cases), with no difference between the three hysterectomy methods (0.5–0.6%). The ureteric damage was positively associated with endometriosis, but the trend did not reach significance. The highest proportion of complications occurred in women with fibroids. Women with a history of pertinent serious illness experienced more complications (4.8%) than women without it (3.4%), P(χ21) < 0.001. Laparoscopically assisted vaginal hysterectomy resulted in significantly higher risks of operative complications (6.1%) than the abdominal (3.6%) and/or vaginal hysterectomy (3.1%), P(χ22) < 0.001.
Because both indication and method were related to exogenous risk of complications, we examined these rates adjusted for age, operator, supervision, history of illness and parity. The adjusted odds ratios showed evidence for a negative linear trend with increasing age. Table 1 shows the crude rates by category and adjusted odds for putative effects on risk. A test for trend showed a decrease in odds of 11% for each (10-year) age category, P(χ21) = 0.002.
We have inspected interactions between indication for surgery (A) and (1) age, (2) method, (3) grade of surgeon and (4) grade of supervisor, and between method of surgery (B) and age (1). The two significant interactions were between indication for surgery and age (A1, P= 0.036) (Fig. 1) and between method and age (B1, P= 0.002) (Fig. 2). The continuous reduction in risk associated with increasing age was only evident with vaginal surgery. Similarly, the continuous reduction in risk with age was apparent among women with fibroids, but not women with dysfunctional uterine bleeding, for whom the odds were decreasing slightly if under 49 and increasing afterwards. Women with prolapse had a lower risk of complications if aged over 50 (OR = 0.59, 0.46–0.75).
Examining just operative haemorrhage and visceral damage complications in a logistic regression model (not shown), the overall OR decreased with age from 1.2 for the 20–29 year age group to 0.52 in the over 60 category. A separate test for trend was significant (P < 0.001), with a decrease in odds of 19% for each increase in age category. Similarly to all complications, compared with dysfunctional uterine bleeding, there was an increase of operative haemorrhage and visceral damage only for women with fibroids (OR = 1.4, 1.2–1.7). Serious illness increased the odds of these complications by 41% (1.1–1.8), and there was hardly any difference in the effect of grade of surgeon and supervisor. Laparoscopically assisted vaginal hysterectomy had a higher odds compared with the abdominal technique, and there was no evidence of different odds of complications with the vaginal method.
Among women aged over 20, 383/37,173 (1%) severe post-operative complications were reported, and 2848 (8%) other post-operative complications, not reported here.
Crude analyses did not show a difference in the proportion of severe post-operative complications across age categories or a large difference among surgical indications. There was no convincing support for a difference between women who were given prophylactic antibiotics and those who were not (P= 0.54). As expected, women with a history of pertinent serious illness had more complications (P= 0.024). There was no difference between women who were operated by consultant surgeons and those who were operated by non-consultants (P= 0.61), or between supervision categories (P= 0.67). Women who had laparoscopically assisted vaginal hysterectomy had more complications than those who had the vaginal and abdominal procedure (P= 0.015). Over 7% of women who had experienced operative complications also encountered post-operative complications, compared with 0.8% among women who had no complications during surgery (P < 0.001).
A Poisson regression model was used to analyse the adjusted effect of surgical technique on severe post-operative complications (Table 2). Rate ratios (RR) were adjusted simultaneously for age, main indication for surgery, use of prophylactic antibiotics, history of serious illness, grade of surgeon and grade of supervisor. The rates decreased with increasing age, but a separate test for trend with age revealed no compelling evidence of a linear trend (P= 0.16). There was no evidence of any significant interaction between age and method, or between indication and age, method, grade of surgeon or grade of supervisor.
This present study investigates the severe complication rates recorded in an unselected series of over 37,000 hysterectomies performed in the mid-1990s, and thus enables a comparison of clinical experience across the whole range of practice. We have no information on individual surgeons' experience, as the study was blinded to any identification. We have modelled the severe complication rate using measures of potential confounding, and compared the rates for different indications and different method of hysterectomy as the best estimate of independent effects on risk. Since severe operative complication rates were around 3% and post-operative rates around 1%, understanding the causes of complication rates in clinical trials demands very large numbers to reliably detect systematic effects. Clearly, the justifications (relative to other clinical priorities) for massive clinical trials of common, accepted surgical methods are weak. This study attempts to bridge the knowledge gap by analysing as near to a representative sample of surgical practice and experience as could be readily obtained inexpensively, to identify the major determinants of serious complications in the absence of randomised comparison. Most randomised trials do attempt rigorous comparison of the benefits associated with different methods of treating benign disease of the uterus but include around 300 women at most.5 Therefore, they can show effects on duration of surgery or on length of stay6 but not the complex effects of age and method shown here.
This was a heterogeneous group of women by age and by indication, with largely unmeasured intrinsic risk. Older women were more likely to have vaginal hysterectomies as part of a pelvic floor repair, and the overall risk was lower than for abdominal surgery after age 50. The overall risk of complications of hysterectomy for symptomatic fibroids certainly decreased with age, while for other indications it was relatively constant.
All operative complications showed some reduction in risk with age and increase with fibroids. Laparoscopic procedures were associated with a near doubling of risk, as previously reported7 but recently disputed,8 which could be a manifestation of patient selection or varying operative experience, now possibly systematically improved. Of these, major operative haemorrhage showed a very similar pattern as it constituted the majority of complications. Post-operatively, the strongest predictor was a history of operative complications; otherwise fibroids as an indication resulted in more complications. Notwithstanding the results of recent small clinical trials,9 our results indicated a doubling in adjusted rates of haemorrhage and visceral damage of laparoscopically assisted vaginal hysterectomy compared with abdominal surgery. The seniority of the operator and the supervisor did not affect this increase in risk, although there was some evidence to suggest that consultants experienced a higher risk—presumably because of unmeasured case severity differences.
Severe operative complications were more likely in women with a history of serious illness and among those with a higher number of pregnancies. The highest risk was evident among women with fibroids, but the observed continuous reduction in risk associated with age among these women was not apparent among women with dysfunctional uterine bleeding.
As far as post-operative complications are concerned, the major determinant is a history of operative complications. The rate was higher among laparoscopic (1.7%) and vaginal methods (1.2%) than with abdominal methods (0.9%), and again among women with fibroids (1.2%) than among women with dysfunctional uterine bleeding (1%). The adjusted odds were 64% (laparoscopically assisted vaginal hysterectomy) and 39% (vaginal) higher than after abdominal surgery (baseline), and again 46% higher among women with fibroids than women with dysfunctional uterine bleeding (baseline). The finding of a 0.5–0.6% rate of visceral damage is similar to findings from other studies which described rates of visceral damage of 0.5–2.0%.10 Overall mortality rates in the study were low and were compatible with findings reported from other large cohort studies that ranged from 0.6 to 1.6 per thousand.11–13
In the validation process, we (AM) reviewed hospital case notes for the 1st, 4th, 7th and 10th month of the study. When compared with the study sample, newly identified cases (1431 women) differed only (P < 0.05) with a higher likelihood of conserved ovaries (OR = 1.2), respiratory and cardiovascular complications (OR = 2.6), major haemorrhage (OR = 2.1) and post-operative complications before discharge (OR = 3.1). Comparing case note records of the same women with study responses suggested good agreement on all data recording (Cohen's kappa >0.6), except moderate disagreement on indication for surgery dysfunctional uterine bleeding or pelvic mass (kappa 0.4–0.6) and poor agreement on fibroids and post-operative complications (kappa <0.4), for reasons mentioned above. Some residual under-estimation of risk herein remains possible due to response bias.
This cohort study provides useful insights in aggregated determinants of risk factors in severe complications of hysterectomies performed for benign causes. We did not examine variation in complication rate by operator or by centre. These results have established that severe operative complications may occur in around 4.4% of hysterectomies (e.g. among the women treated for symptomatic fibroids). The risks remained elevated even after appropriate adjustments, both operatively and six weeks post-operatively. The highest observed risk group was younger women who underwent laparoscopically assisted vaginal hysterectomy for fibroids as the main indication, and those with history of related serious illness.
The most common operative complications have a reduced risk with increasing age, but risk increased with parity. Laparoscopically assisted vaginal hysterectomy showed a doubling of risk compared with vaginal or abdominal surgery. Post-operative complications were higher among women with operative complications and, when adjusted, the highest risks were among women with fibroids. Clearly, these observations represent aggregated average effects over the complete spectrum of clinical practice in the mid-1990s. Individual clinical decisions can only be guided by these observations. This study can provide little insight into the effect of changing surgical method or operator to reduce serious complications—as that would require an experimental approach. However, a 50% increase attributable to a history of serious illness across the board and a 4% increase for each live birth are important indicators of risk. Likewise, a nearly 30% increase among consultant-supervised compared with unsupervised procedures perhaps only serves to emphasise the complexity of interpreting complication rates as indices of quality of care.
Hysterectomy is a common, routine surgery with comparatively rare serious complications. However, younger women, because of their more vascular pelvises, who undergo laparoscopically assisted vaginal hysterectomy for symptomatic fibroids, are at most risk of experiencing severe complications both operatively and post-operatively. Therefore, a less invasive alternative treatment for symptomatic fibroids could particularly benefit this group of patients, while less invasive treatments for dysfunctional uterine bleeding, such as various methods of endometrial ablation or resection, would need to meet the current low levels of clinical complications in order to replace hysterectomy. On the other hand, hysterectomy disturbs the blood flow to, and the mechanics of, the remaining pelvic organs. Moreover, it often results in early or surgical menopause, and relatively rare serious operative and post-operative complications may mask other significant medium and long term health tradeoffs. In a long run, and in the light of the recent NICE recommendations, the wider application of less invasive surgeries for benign indications would almost certainly result in a healthier female population.
The authors would like to thank collaborating consultants, their teams and theatre staff for completing the forms; the women for allowing this study, and members of the Steering Group.