Specialist gynaecologists and survival outcome in ovarian cancer: a Scottish national study of 1866 patients

Authors


Correspondence: Dr E. J. Junor, Beatson Oncology Centre, Western Infirmary, Dumbarton Road, Glasgow, UK.

Abstract

Objective To determine whether specialist gynaecological surgeons improved survival in women with ovarian cancer when compared with general gynaecologists.

Design Retrospective case note review.

Population All women diagnosed with ovarian cancer in Scotland in 1987, 1992, 1993 and 1994.

Methods Data on prognostic factors and surgical and post-operative management was extracted from case notes. Surgeons were classified as specialist gynaecologists, general gynaecologists or general surgeons by an independent committee with no knowledge of an individual's outcome. Cox's proportional hazards model was used to determine the relative risk of a patient dying, if managed by specialist and general gynaecologists, after adjustment for age, histology, tumour differentiation, presence of ascites and socio-economic status. Analysis was performed separately for each FIGO stage.

Main outcome measures Relative hazard ratios for survival up to three years.

Results Survival benefit for specialists varied according to the stage of the disease. The greatest benefit was observed among women with Stage III disease (44% of women presented at this stage) where there was a 25% (relative hazard ratio = 0.75, P= 0.005) reduction in the rate of dying for women operated on by specialist gynaecologists, compared with women operated on by general gynaecologists. Differential use of platinum chemotherapy did not explain this survival advantage. Specialist gynaecologists more often debulked tumour to c 2 cm than general gynaecologists in Stage III cases (36.3%vs 28.7%, P= 0.07). In women with Stage III carcinoma with > 2 cm remaining, survival was significantly improved for women treated by specialist gynaecologists (relative hazard ratio = 0.71, P= 0.007). No significant differences were observed for patients with Stages I, II and IV disease, although there were fewer deaths in women with early stage disease.

Conclusions Specialist gynaecologists improve survival for some women with ovarian cancer.

INTRODUCTION

Calman and Hine have proposed a network of cancer centres and cancer units for improving the outcome of cancer in the UK. Evidence exists for improved survival among patients undergoing surgery by specialist surgeons for breast1 and bowel cancer2,5, and as surgery is the major intervention in the management of ovarian cancer, with improved outcome for women in whom little or no macroscopic disease remains after operation4, it is important to ascertain what benefits might accrue from a similar approach in this disease. Several studies have shown that survival of women with ovarian cancer improves if they are operated on by gynaecologists compared with general surgeons5,6,7, but there is no clear evidence whether gynaecological oncologists (specialists) further improve outcome over general gynaecologists.

The Royal College of Obstetricians and Gynaecologists has already indicated their preference for women with ovarian cancer to be managed by specialist gynaecologists8.

A previous study5, which examined factors affecting outcome for all women diagnosed with ovarian cancer in Scotland in 1987, a nonstatistically significant but improved survival outcome for women operated on by specialists in the surgery of gynaecological cancer was observed, but not reported in the text. Thus the current study was specifically designed to test the hypothesis that improved survival would result if women with ovarian cancer were operated on by gynaecological oncologists with sample size sufficient to ensure a precise estimate of benefit.

METHODS

All women diagnosed with primary ovarian cancer in Scotland in the years 1987, 1992, 1993 and 1994 were eligible for the study. A list of 2101 registrations was obtained from the Scottish National Cancer Registry, and each case note was sought to validate the registry diagnosis, to ensure the woman was normally resident in Scotland, and was diagnosed in one of the specified years. One hundred and sixteen women (5.5%) were excluded because of an incorrect pathology or year of diagnosis. The medical records of a further 119 women (5.7%) could not be found.

Detailed information for the remaining 1866 women was abstracted by two experienced data managers specially trained by and with support from one investigator (E.J.) to resolve disagreements and to check on the quality of the information. These data included presenting features; investigations; histological type of the carcinoma of the ovary; its stage; operative procedures; volume of residual disease and subsequent referral and treatment.

Staging was performed using the standard FIGO classification on the basis of the operation notes, pathology report and the results of all available investigations. Sub-staging was only possible for Stage I and then incompletely. All histological reports were coded according to the International Classification of Disease for Oncology.

All of the surgeons who operated were classified into one of three groups: gynaecological oncologist (specialist), general gynaecologist (nonspecialist) or general surgeon. These classifications were made by the committee of gynaecologists who oversaw the 1987 study5. It was a subjective assessment determined by two criteria: 1. the group had no knowledge of the outcome for each of these groups in the 1987 or any other study; and 2. the classification was fixed for the period of the study.

The records of the Registrar General (Scotland) were examined for the cause and date of death. All deaths until 31 December 1996 were included, providing a minimum follow up time of two years and a maximum follow up of 10 years.

Statistical details

The size of the current study was calculated from the results of the original 1987 Scottish study5. The intention was to be able to detect a relative hazard ratio of 0.82 with 90% power at the 5% significance level. The assumption was made that general gynaecologists would treat three times the number of women as specialists. This required the accrual of 725 deaths. With approximately 185 deaths expected in three years of follow up in women treated by gynaecologists in each study year, data throughout Scotland for four years was required. Cases diagnosed in the years 1992, 1993 and 1994 were added to the 1987 data to complete the four years. These years were chosen to provide a balance between current practice and sufficient follow up to produce a reliable three-year estimate of survival in ovarian cancer. All analyses are based on survival to three years, as the majority of relapses and deaths occur within this period, and three of the four diagnostic years in the study have been followed for this length of time.

Kaplan-Meier survival curves have been used to provide a graphical presentation of the survival pattern of patients treated by gynaecological oncologists, general gynaecologists and general surgeons. This technique does not allow for any statistical adjustment for prognostic factors. Survival comparison taking into account prognostic and other explanatory variables has been undertaken using Cox's proportional hazards model.

Five prognostic factors have been used, in addition to FIGO staging, to allow for confounding variables seen by each category of surgeon. These were degree of differentiation (well, moderate, poor and not recorded), age (< 55, 55–64, 65–74, 75+), presence of ascites (yes, no), histology (reduced from 11 to 3 categories on the basis of prognosis) and Carstairs' deprivation category (affluent, intermediate and deprived).

In order to explore differences between category of surgeon, two explanatory variables have been used in the analysis: volume of residual disease after operation and type of chemotherapy given. Volume of residual disease was recorded as: bulky; > 2 cm; c 2 cm or minimal; or no disease on the basis of entries in the casenotes. After preliminary analysis, this variable was simplified to two categories (< 2 cm or > 2 cm) because of the lack of any additional discrimination in survival analysis with four levels. Chemotherapy was classified into four groups: platinum alone; platinum in combination with a second drug; alkylating agent alone; or no chemotherapy. At this time taxanes were not used as first line chemotherapy.

Models have first been fitted using category of surgeon with the standard prognostic variables. Where differences exist, the additional variables of type of chemotherapy and extent of residual disease have been added to the model to see whether this removes (i.e. explains) the effect of category of surgeon. Because Cox's analysis produced a significant interaction (P= 0.04) between stage and category of surgeon (implying that differences between categories of surgeon varied according to the stage of the disease), results have been reported separately for each stage.

RESULTS

Table 1 presents the characteristics and the distribution of prognostic factors of women operated on by different categories of surgeon. Specialist gynaecologists tended to see more late stage cases (FIGO Stages 111, IV) than general gynaecologists, with general surgeons seeing the highest proportion. This pattern was also apparent for age of the women, differentiation of the cancer and presence of ascites. In general, general gynaecologists cared for women with a better prognosis, specialist gynaecologists an intermediate group, and general surgeons women with the poorest prognosis. The importance of Stage III cases on overall outcome in ovarian cancer can be appreciated from the fact that 44% (830/1866) of women present at this stage.

Table 1.  Characteristics of women operated on by specialist, nonspecialist gynaecologists and general surgeons.
 No. of patients (n=1866*)3 year survival rate (%)Specialist gynaecologist (%) (n= 351)Nonspecialist gynaecologist (%) (n= 1032)General surgeon (%) (n= 216)
  1. *Includes 267 patients in whom no operation was performed.

  2. **Insufficient information.

Stage     
  I4138019.430.811.1
  II167499.111.55.6
  III8302054.744.069.9
  IV298615.413.010.6
  Unstaged**15851.40.72.8
Age     
  < 554435421.931.111.1
  55–644813730.227 030.1
  65–745202431.626.532.9
  ≥754221316.215.425.9
Differentiation     
  Well124643.79.46.0
  Moderate3353817.121.920.8
  Poor7412753.341.245.4
  Not known6664525.927.527.8
Histology     
  Good prognosis105846.67.71.4
  Intermediate10304166.764.946.8
  Poor7311826.827.451.9
Ascites     
  Absent7184425.640.044.0
  Present11482874.460.056.0
Depivation     
  Affluent4494129.323.821.8
  Intermediate11473851.064.063.9
  Deprived2702619.712.214.4
All patients186632   
All surgical patients159937   

Figure 1 shows the unadjusted Kaplan-Meier survival curves for women operated on by specialist gynaecologists, general gynaecologists and general surgeons for each FIGO stage. The hazard ratios for each of these comparisons are shown in Table 2, with adjustment made for five prognostic factors (e.g. age, differentiation, histology, presence of ascites and deprivation category). A highly significant difference is apparent for women with Stage III disease, based on 621 deaths, with a reduction in the death rate of 25% (P= 0.005) for women operated on by a specialist gynaecologist and an increase of 32% (P= 0.009) in the rate of dying in women operated on by a general surgeon, taking the pattern of survival of those treated by general gynaecologists as the baseline. The median survival among women with Stage III carcinoma was 18 months for those treated by specialist gynaecologists and 13 months for those treated by general gynaecologists. An interesting feature of the survival curves for these women is that the difference seems to be greatest at about 18 months following surgery and thereafter appears to diminish. In contrast, the survival curve in women treated by general surgeons shows increasing divergence for the whole three years of follow up. Inferences relating to Stage I, Stage II and Stage IV cases cannot be drawn as the number of deaths within each of these stages is small. No significant differences are seen for women in relation to surgical category, and the confidence intervals are wide (e.g. relative hazard ratio for Stage II cases treated by a general surgeon is 0.67, but this is based on only six deaths and the confidence interval is 0.25 to 1.75).

Figure 1.

Kaplan-Meier survival curves by stage and specialty of person performing initial surgery.

Table 2.  Analysis of pattern of survival for women treated by specialist and nonspecialist gynaecologists and general surgeons. Values are given as hazard ratio (95% CI), unless otherwise indicated.
 Specialist gynaecologistNonspecialist gynaecologist (baseline)General surgeon
  1. *5 prognostic factors were age, degree of differentiation, histological type, presence of ascites and deprivation category.

  2. Only Stage m women showed differences which were statistically significant.

Stage 1 (79 deaths) (n)13615
  Unadjusted1.0111.14
  Adjusted for 5 factors*0.83 (0.45–1.55)11.01 (0.40–2.57)
Stage II (80 deaths) (n)17576
  Unadjusted1.0311.29
  Adjusted for 5 factors1.00 (0.56–1.80)10.67 (0.25–1.75)
Stage I and I1 (159 deaths) (n)3011811
  Adjusted for 5 factors0.93 (0.62–1.41)10.80 (0.42–1.53)
Stage III (621 deaths) (n)141345135
  Unadjusted0.8411.46
  Adjusted for 5 factors0.75 (0.62492)11.32 (1.07–1.63)
  P0.005 0.009
Stage IV (194 deaths) (n)4912322
  Unadjusted0.9511.31
  Adjusted for 5 factors1.01 (0.71–1.45)11.26 (0.78–2.04)

Is the observed difference between specialist and genera1 gynaecologists for Stage III cases due to the extent of surgery or the type of chemotherapy? This is addressed in Table 3. Substantial survival benefits are associated with the use of platinum chemotherapy and more so when it is used in combination with alkylating agents, but differential use of these agents in women treated by specialist and general gynaecologists and general surgeons (Table 4) makes only a small impact on the survival differences seen. Thus the difference would seem to be due to the surgical intervention.

Table 3.  Hazard ratios (95% CI) for Stage III cases including chemotherapy as an explanatory factor for inter-surgical category differences.
 Specialist gynaecologist (n=192)Nonspecialist gynaecologist (baseline) (n=454)General surgeon (n=151)
  1. *Adjustment for chemotherapy involves including an additional factor in the Cox's proportional hazards model which has 4 categories: platinum in combination; platinum alone; alkylating agents; no chemotherapy.

Adjusting for 5 prognostic factors0.75 (0.62–0.92)11.32 (1.07–1.63)
  P0.004 0.007
Additional adjustment for chemotherapy*0.77 (0.63495)11.25 (1.01–1.54)
  P0.01 0.04
Additional adjustment for chemotherapy and extent of residual disease0.78 (0.64495)11.07 (0.86–1.33)
  P0.02 0.54
Table 4.  Chemotherapy received by Stage III women in relation to surgeon. Values are given as% (n).
SurgeonPlatinum + alkylating agentPlatinum aloneAlkylating agentNo chemotherapy
  1. Comparison between specialist and general gynaecologist: χ2= 9.1, P= 0.03.

Specialist gynaecologist (n= 192)27.6 (53) 23.3 (106) 13.9 (21)43.2 (83) 35.2 (160) 34.4 (52)14.6 (28) 18.3 (83) 17.9 (27)14.6 (28) 23.1 (105) 33.8 (51)
General gynaecologist (n= 454)23.3 (106)35.2 (160)18.3 (83)23.1 (105)
General surgeon (n= 151)13.9 (21)34.4 (52)17.9 (27)33.8 (51)

The ability of the surgeons to debulk Stage I11 tumours to < 2 cm in diameter is shown in Table 5. The specialist gynaecologists were the most successful, with 36.3% successfully debulked, although the difference between them and general gynaecologists (28.7%) was not large. In contrast, only 11.6% of Stage I11 cases operated on by general surgeons were debulked to < 2 cm diameter. The impact of successful debulking on survival differences for Stage I11 cases is shown in Table 3 (final line). The poorer survival seen in the women treated by general surgeons disappears substantially when extent of residual disease is included in the survival model, indicating that this is the explanation for their poorer survival. However, the hazard ratio for specialist compared with general gynaecologist (relative hazard ratio = 0.78, P= 0.02) is unchanged from that observed after fitting type of surgeon and chemotherapy (relative hazard ratio = 0.77, P= 0.02).

Table 5.  Extent of residual disease after operation in Stage III women in relation to surgeon. Values are given as % (n).
 Residual tumour
Surgeon< 2cm>2cm
  1. Comparison between specialist and general gynaecologist: χ2= 3.2, P= 0.07.

  2. *Information on volume of tumour remaining is unknown for 2, 22 and 5 women operated on by specialist, general gynaecologist and general surgeons, respectively.

Specialist gynaecologist (n= 190)*36.3 (69)63.7 (121)
General gynaecologist (n= 432)28.7 (124)71.3 (308)
General surgeon (n= 146)11.6 (17)88.4 (129)

Comparison of survival curves for Stage I11 cases who did and did not have tumour debulked to < 2 cm is presented in Fig. 2. A significant interaction term exists again (P= 0.02) for surgical category and extent of residual disease. No difference is apparent between specialist and general gynaecologists for those with tumour < 2 cm diameter remaining (relative hazard ratio = 1.09) but a clear difference is apparent for those with larger deposits of tumour remaining (relative hazard ratio = 0.71, P= 0.007). This would seem to indicate that it is the more difficult Stage 111 cases who benefit most in survival from being operated on by a specialist gynaecologist. This is a very large subgroup of all the women with ovarian cancer, with 480 deaths of the 975 deaths in our study.

Figure 2.

Kaplan-Meier survival curves of Stage III patients by volume of residual disease and specialty of person performing the initial surgery.

DISCUSSION

This population-based study of all 1866 cases of ovarian cancer diagnosed in Scotland over four years demonstrates the benefit in survival of being treated by specialist gynaecologists. This benefit relates strongly to women with Stage III disease, is less evident for those with Stage I or 11 disease and was not found for those with Stage IV disease. This benefit to Stage III women was most pronounced between one and two years of follow up, with a maximum survival benefit of 12% at about 18 months. After two years there is a tendency for the two survival curves to come together. While benefit is significant in Stage 111 cases only, 44% of women present at this stage and 66% of ovarian cancer deaths occur in women with Stage III carcinoma of the ovary. It is also the group in which one might logically expect to see benefit with specialist care. In women with Stage I or 11 ovarian cancer it would be anticipated that resection of all or most of the tumour would be achieved by a total hysterectomy and bilateral salpingo-oophorectomy, unlike Stage III disease where tumour nodules are implanted outwith the pelvis.

The survival benefit for Stage 111 cases is clearly associated with being managed by a specialist, although this can come about in two ways: it could be due to the extent of the surgery, or more effective chemotherapy, especially with platinum. More women treated by specialist gynaecologists received platinum compared with those managed by general gynaecologists (71%vs 59%). The impact on survival was assessed by adding a term in the Cox's model to represent the effect on survival of additional chemotherapy after fitting a model containing the category of surgeon. The hazard ratio changed very little from 0.75 to 0.77 (Table 3). Thus the part played by the specialist gynaecologist appears to account for the difference in survival and not any vanation in the prescription of platinum between the two groups.

Our only direct measure of the extent of surgical intervention is the volume of residual tumour, although we can also use survival as an indicator of successful surgery. The specialist gynaecologists appear to be more successful in debulking tumour to c 2 cm diameter than the general gynaecologists (36%vs 29%P= 0.07). Having debulked to this level, survival was similar. Women with Stage 111 cancer with tumour > 2 cm diameter remaining after surgery showed the greatest survival benefit from specialist surgical care. It may be that specialists were more successful at debulking than general gynaecologists, but because of the imprecise way of measuring residual tumour (< 2 cm or > 2 cm) this could not be quantified. As residual tumour > 2 cm constituted 73% of the Stage III cases, any improvement in treatment in this group makes a substantial contribution to overall improvement in survival in ovarian cancer and would justify the need for specialist gynaecologists.

The independent benefit of receiving platinum chemotherapy is again evident in this study, as in our earlier report5. Women receiving platinum in combination with alkylating agents had a reduced risk of dying compared with those receiving platinum alone (relative hazard ratio = 0.75, 95% CI 0.62–0.90) and the hazard ratio for alkylating agents relative to platinum alone was 1.62 (95% CI 1.30–2.01). This was based on an analysis of women who were aged younger than 75 years, with a survival time greater than 60 days and excluded Stage 1A cases. Women with Stage 1A carcinoma of the ovary may not need chemotherapy, and elderly women or women who are too ill may be considered unsuitable for chemotherapy.

The role of multidisciplinary, combined clinics, evaluated in our 1987 study, was difficult to assess in relation to specialist surgeons because of the small number of cases operated on by specialists but not then treated in a combined clinic. However, if we restrict our analysis to the women who were operated on by general gynaecologists, it would appear that the benefits of the combined clinic were less apparent in this study. We interpret this as increased awareness among general gynaecologists of the optimal treatment for ovarian cancer.

Two cautionary points should be made. The first is that the follow up period for this analysis is only three years and much of the additional survival benefit due to specialist care is seen in women with tumour still remaining. Thus the benefit is unlikely to result in additional cures but is important as an opportunity for providing improved quality as well as length of life. The second is that the results are based on observational data. We have done our best to minimise selective bias concerning the classification of the surgeons. Surgeons were classified before the study started and the classification remained constant throughout the years of the study. The authors had no influence on the allocation of women to the surgeons. As in all observational studies, we can never be absolutely sure that bias has been completely avoided. However, no randomised trial has been performed to evaluate the benefit of being treated by a specialist gynaecologist, nor is it likely that such a trial will be performed.

Few other studies have reported on specialisation in ovarian cancer. The largest, though not population-based, has been the National Study of Ovarian Cancer which examined the care of 12,316 women from 904 hospitals between 1983 and 1988 with established cancer programmes in the United States. No overall significant survival advantage was found between specialist and general gynaecologists9 although sub-stage analysis showed a significantly better survival in Stage IIB disease (P c 0.03). No adjustment was made for prognostic factors, but the large number of cases may well have led to fairly balanced groups. A smaller study from California (263 women) of Stage IIIC and N carinoma of the ovary did find benefit associated with care by a specialist gynaecologist but the comparison group included general sugeons10. A further study of 87 women receiving identical chemotherapy for Stage I or 11 ovarian cancer found improved survival in women operated on by specialists, but concluded that this was due to more comprehensive staging at operation, which ensured that the women included in the specialist group were truly Stage I or 11. Again the comparative group included general surgeons11.

In the UK the Royal College of Obstetricians and Gynaecologists recognises as specialists those fully accredited, subspecialty trained gynaecologists and the Directors of Training Programmes. All except two of our specialists complied with this definition. The additional two gynaecologists were deemed to merit inclusion by a committee of specialist and nonspecialists at the initiation of the 1987 study. We anticipate that the inclusion of the additional two gynaecologists may be disputed, but it is important to remember that the ‘specialists’ were defined and designated prior to the collection of any data. Our results were unaffected by exclusion of these two specialists.

At present there are insufficient specialist gynaecological oncologists to operate on all cases of ovarian cancer in Scotland and in the light of this additional evidence the future service provision may need to be reconsidered8.

Acknowledgements

This work was funded by grants from the Clinical Resource and Audit Group and the Scottish Cancer Therapy Network. We are grateful to Professor C. R. Gillis whose initial observations inspired the study.

Ancillary