SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. References

Objective To examine the influence of operator specialty, volume of work and referral to an oncologist on the survival of women with ovarian cancer.

Design Population-based retrospective cohort study, using hospital records and Cancer Registry data.

Setting The North Western Region, UK.

Population Six hundred and ninety-one women undergoing laparotomy for histologically confirmed ovarian malignancy during 1991 to 1992.

Methods Univariate and multivariate survival analyses.

Main outcome measures Univariate survival estimates. Relative risks, derived from Cox's proportional hazards model, describing the effect on survival of surgeons vs gynaecologists as baseline, high volume vs low volume operators and referral vs nonreferral to an oncologist.

Results After adjusting for woman and disease-related prognostic factors, operation by a surgeon was shown to have an adverse impact on survival (RR = 1.58, 95% CI 1.19 to 2.10). Regardless of how a high volume operator was defined (in terms of the number of laparotomies performed), no survival advantage over low volume operators could be demonstrated. Women referred to an oncologist had significantly better survival than women not referred (RR = 0.54, 95% CI 0.43 to 0.68)

Conclusions All women undergoing surgery for ovarian cancer should have access to a gynaecological opinion and postoperatively should be referred for a nonsurgical oncological opinion.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. References

The report by the Expert Advisory Group on Cancer to the Chief Medical Officer of England and Wales, A Policy Framework for Commissioning Cancer Services1, stresses the importance of trained site-specialised surgeons, a minimum acceptable volume of work and nonsurgical oncological input. We have examined the influence on outcome of operator specialty, volume of work and referral to an oncologist, in a population based series of women with ovarian cancer.

METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. References

All women with ovarian cancer diagnosed from 1 January 1991 to 31 December 1992 were identified from records held by the North Western Regional Cancer Registry (NWRCR) and the following information abstracted from hospital records: age at diagnosis, histological diagnosis, stage, operating surgeon, operative procedure and whether there was a referral to an oncologist. Staging was performed by one of the authors (A.B.) according to FIGO 1976 criteria2 using operative records, pathology reports and the results of clinical investigations. Women were considered to have had an emergency laparotomy if this was undertaken within 48 hours of first presentation. Operating surgeons were either general surgeons or gynaecologists, the latter arbitrarily categorised as low or high volume operators according to whether they had operated on six or more patients with ovarian cancer during 1991 to 1992, or fewer than six. Variation in the completeness of operative records allowed only the following crude classification of operative procedures: inoperable (biopsy only), oophorectomy with or without hysterectomy, and omentectomy with or without oophorectomy and hysterectomy. For this reason the residual rumour load was not recorded. Where insufficient data were available to determine the value of a variable, this information was recorded as unknown. Survival was measured from date of laparotomy until death from any cause, survival times being censored after 31 December 1994. Vital status was determined from records held by the NWRCR.

The analysis, restricted to those women who had undergone laparotomy, examined the effects on survival of the following groups of variables: woman-and disease-related (age, stage, histology, tumour differentiation and emergency/elective procedure), and organisational (operator specialty, operator volume and referral to an oncologist).

Analyses were performed using the SPSS for Windows statistical package. Women were grouped for analysis according to the specialty of their operator (gynaecologist or general surgeon), referral to an oncologist (referral or nonreferral) and for those cases operated on by gynaecologists, volume of work (low or high volume). Univariate methods were used to compare the characteristics of each group and univariate survival was analysed using the Kaplan-Meier method3 and logrank tests of significance. A multi-variate survival analysis using Cox's proportional hazards model4 was then undertaken to examine, simultaneously, the effects of the study variables on survival. For this analysis the 12 histological types recorded were collapsed into three groups using the results of the univariate survival analysis: 1. borderline and germ cell tumours were classified as having ‘good’ survival; 2. mucinous adenocarcinoma, serous adenocarcinoma, endometroid, clear cell, sex chord and stromal tumours and miscellaneous tumours as ‘moderate’ survival; 3. adenocarcinoma, not otherwise specified as ‘poor’ survival. All tests of significance were performed at the 5% two-sided significance level.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. References

Eight hundred and sixty women with primary ovarian cancer were diagnosed during the study period; 169 (19.7%) were excluded from this analysis because casenotes were unavailable for review (n= 33), post-mortem diagnosis (n= 4), clinical, cytological, or radiological diagnosis which was not histologically confirmed (n= 132). The remaining 691 women (80.3%) had a laparotomy for histologically confirmed ovarian malignancy and comprise the study population.

Table 1 shows study population characteristics and the results of the univariate survival analysis. Age, stage, histological type and rumour differentiation were significant prognostic factors, but survival was not significantly worse in women undergoing an emergency, compared with an elective laparotomy. Women operated on by surgeons, compared with gynaecologists, had a significantly worse survival, as did those operated on by high volume, compared with low volume gynaecologists. Gynaecologists were more likely than surgeons to undertake an operative procedure which involved removing the omentum, but the frequency of this procedure did not vary between low and high volume gynaecologists (Table 2). Women referred to an oncologist were not shown to have a significant survival advantage in the univariate analysis.

Table 1.  Determinants of outcome: univariate survival comparison (crude survival). Total number of cases = 691.
  Survival(%) 
 Cases n (%)1 year2 years3 yearsMedian survival (months)
  1. Censored median survival times.

  2. Logrank test for heterogeneity.

  3. *Refers only to the 596 cases whose laparotomies were performed by gynaecologists.

Age[χ2(4)= 98.7, P < 0.0001]    
0–44105 (15)89.582.982.947.8
45–54115 (17)83.566.156.947.3
55–64189 (27)69.853.452.147.8
65–74168 (24)53.041.132.513.9
geqslant R: gt-or-equal, slanted75114 (17)43.931.627.18.3
Stage[χ2(4)= 295.8, P<0.0001]     
I222 (32)94.691.989.147.8
II88 (13)83.068.264.147.7
III269 (39)49.830.123.511.7
IV53 (8)35.915.110.65.9
Not known59 (9)42.427.120.69.2
Degree of differentiation [χ2(3)= 112.7, P <0.0001]     
Well177 (26)77.467.262.447.8
Moderate125 (18)64.046.440.620.7
Poor212 (31)51.929.322.012.5
Not known177 (26)75.773.572.947.8
Histological type: survival [χ2(2)= 168.4, P <0.0001]     
Good128 (19)95.395.395.347.8
Moderate455 (66)65.350.643.525.2
Poor108 (16)38.915.714.47.6
Nature of admission[χ2(1)= 1.2, P = 0.27]     
Elective640 (93)67.053.849.133.3
Emergency51 (7)62.849.039.419.6
Operator specialty[χ2(1)= 58.5, P < 0.0001]     
Gynaecol.596 (86)71.657.653.047.8
Surgeon95 (14)35.827.419.76.8
Nonsurgical oncologist[χ2(1)= 0.5, P = 0.48]     
Nonreferral322 (47)63.055.953.347.8
Referral369 (53)69.951.244.325.9
Volume of work*2(1)= 6.0, P < 0.05]     
Low (1-5)92 (15)78.369.666.147.8
High(geqslant R: gt-or-equal, slanted6)504 (85)70.455.450.737.8
Table 2.  Distribution of operative procedures by operator specialty and volume of work. Values are given as n (%). TAH = total abdominal hysterectomy; BSD = bilateral salpingo-oophorectomy.
  Operative procedure  
  1. *χ2 test for heterogeneity.

 No.InoperableOophorectomy ±TAHOmentectomy ±TAH & BSOX2(2)P
Gynaecologist596110(19)226 (38)260 (44)  
Surgeon9557 (60)23 (24)15 (16)78.7< 0.0001*
Low volume gynaecologist9214 (15)35 (38)43 (47)  
High volume gynaecologist50496 (19)191 (38)217 (43)0.90.65*

The interpretation of the univariate survival comparisons is confounded by imbalances in the distribution of prognostic factors (Table 3). Adverse prognostic factors (older age, late stage, the ‘poor’ histological types and poorly differentiated tumours) were more likely to be found in women operated on by surgeons, in women operated on by high volume gynaecologists, and in women referred to an oncologist.

Table 3.  Distribution of study variables by operator specialty, referral to an oncologist and operator volume. Values are given as n (%) unless otherwise indicated.
 GynaecologistSurgeonNonreferral to oncologistReferral to oncologistLow volume gynaecologistHigh volume gynaecologist
  1. χ2 test for heterogeneity: *P< 0.05; P<0.01; P<0.001.

TOTAL596 (86)95 (14)322 (47)369 (53)92 (15)504 (85)
Age      
Median [range]60 [18-92]70 [29-90]64 [18-92]59 [21-85]59 [27-85]60 [18-92]
0–44102 (17)3 (3)57 (18)48 (13)18 (20)84 (17)
45–54109 (18)6 (6)31 (10)84 (23)11 (12)98 (19)
55–64166 (28)23 (24)75 (23)114 (31)27 (29)139 (28)
65–74134 (23)34 (36)80 (25)88 (24)24 (26)110 (22)
geqslant R: gt-or-equal, slanted7585 (14)29 (31)79 (25)35 (10)12 (13)73 (15)
Stage      
I207 (35)15 (16)149 (46)73 (20)43 (47)164 (33)
II80 (13)8 (8)33 (10)55 (15)7 (8)73 (15)
III222 (37)47 (50)88 (27)181 (49)27 (29)195 (39)
IV45 (8)8 (8)20 (6)33 (9)5 (5)40 (8)
Not known42 (7)17 (18)32 (10)27 (7)10 (11)32 (6)*
Degree of differentiation      
Well165 (28)12 (13)74 (23)103 (28)23 (25)142 (28)
Moderate103 (17)22 (23)43 (13)82 (22)15 (16)88 (18)
Poor180 (30)32 (34)84 (26)128 (35)18 (20)162 (32)
Not known148 (25)29 (31)*121 (38)56 (15)36 (39)112 (22)
Histological type: survival      
Good120 (20)8 (8)109 (34)19 (5)33 (36)87 (17)
Moderate401 (67)54 (57)166 (52)289 (78)49 (53)352 (70)
Poor75 (13)33 (35)47 (15)61 (17)10 (11)65 (13)
Nature of admission      
Elective555 (93)81 (85)294 (91)342 (93)89 (97)466 (93)
Emergency37 (6)14 (15)25 (8)26 (7)3 (3)34 (7)
Not known4 (1)0 (0)3 (1)1 (0)0 (0)4 (1)
Operator specialty      
Gynaecologist596 (100)267 (83)329 (89)92 (100)504 (100)
Surgeon95 (100)55 (17)40 (11)*
Nonsurgical oncologist      
Nonreferral267 (45)55 (58)322 (100)39 (42)228 (45)
Referral329 (55)40(42)*369 (100)53 (58)276 (55)

The multivariate analysis confirmed age, stage, histological type and operator specialty as significant independent prognostic factors. Degree of tumour differentiation was of borderline significance. Referral to an oncologist was now found to be associated with an improved survival when included in the proportional hazards model containing these five variables (Table 4). In an analysis restricted to those women operated on by gynaecologists, no significant association with volume of work was found after adjusting for imbalances in the distribution of these prognostic factors (relative risk of death in high, compared with low volume = 1.19, P= 0.31). In a further analysis of the 302 women with Stage II or Stage III disease gynaecologists were stratified into low (< 6), intermediate (6–10) and high volume (geqslant R: gt-or-equal, slanted 11) operators according to the number of Stage II/III cases they had operated on (Table 5). Again, no significant association between volume and outcome was found in the univariate analysis, nor after correcting for imbalances in the distribution of other prognostic factors (Wald test χ2(2)= 0.85, P= 0.65).

Table 4.  Determinants of outcome: multivariate survival comparison (proportional hazards model).
 No. of deathsRR95% CI
  1. Before 1 January 1995. Ratio of hazards.

  2. *Wald test for heterogeneity.

Age [χ2(4)= 23.44, P < 0.001*]   
0-44181.00Baseline
45-54491.700.98–2.96
55-64911.560.93–2.62
65-741102.101.25–3.52
geqslant R: gt-or-equal, slanted75832.891.70–4.91
Stage [χ2(4)= 118.97, P < 0.0001*]   
I221.00Baseline
II312.601.47–4.58
III2038.155.02–13.22
IV4814.598.38–25.41
Not known478.614.97–14.91
Degree of differentiation [χ2(3)= 7.50, P = 0.06*]   
Well641.00Baseline
Moderate740.910.64–1.28
Poor1651.320.96–1.80
Not known481.080.72–1.60
Histological type: survival [χ2(2)= 18.96, P < 0.001*]   
Good60.200.08–0.48
Moderate2521.00Baseline
Poor931.381.06–1.81
Operator specialty [χ2(1)= 10.15, P <0.01*]   
Gynaecologist2771.00Baseline
Surgeon741.581.19–2.10
Nonsurgical oncologist [χ2(1)= 26.92, P > 0.0001*]   
Nonreferral1491.00Baseline
Referral2020.540.43–0.68
Table 5.  Determinants of outcome: volume of work for Stage II and Stage III tumours: univariate survival comparison. Total number of cases = 302. χ2(2)= 0.04, P= 0.98 (logrank test for heterogeneity).
   Survival(%) 
  Cases n(%)1 year2 years3 yearsMedian Survival (months)
  1. *Volume defined as number of operations performed on women with Stage II or Stage III tumours.

Volume of work*      
Low(1-5)127 (42)64.5742.5238.6919.64
Intermediate(6-10)114 (38)61.4042.9636.9417.93
High(geqslant R: gt-or-equal, slanted11)61 (20)62.3044.2636.1319.57

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. References

The distribution of woman- and disease-related characteristics in this series is similar to that reported in two population based surveys of the management of ovarian cancer previously undertaken in the United Kingdom5,6. Both suggested that women operated on by general surgeons have worse survival than those operated on by gynaecologists. The survival differential observed in our series persisted after controlling for patient and disease characteristics, and is likely to reflect differences in the surgical approach of the two specialties as women operated on by surgeons were less likely to have had an extended procedure. It is not always possible to confirm a diagnosis of ovarian cancer before laparotomy, and therefore any future reconfiguration of cancer services must ensure access to a gynaecological opinion when surgery for intra-abdominal malignancy is planned. A significant advantage for women managed in a multidisciplinary team has previously been described, so it is a matter for concern that 47% of women in this series were not referred to an oncologist.

Unpublished data, submitted in evidence to the Expert Advisory Group, suggested a close relation between the number of patients treated by individual operators and outcome. We were unable to confirm this relation, irrespective of where the volume of work threshold was set, nor can it be explained by differences between low and high volume gynaecologists in the proportion of different operative procedures undertaken (Table 2). A further analysis, confined to cases of Stage II/III disease only, also revealed no relation. Our findings support those of a similar survey undertaken in the West Midlands and may reflect a more uniform approach by gynaecologists in these regions to the management of ovarian cancer. But perhaps we should not be surprised by the lack of association of volume with outcome. Current volume levels have been determined historically by the requirement for general gynaecologists in a district general hospital to meet the needs of their local catchment population. We might expect these arrangements to change when the report by the Expert Advisory Group is implemented. In the interim, it may be naive to imagine that mere repetition guarantees surgical competence, and there is a danger that preoccupation with quotas will detract from the importance of training, protocols, audit and professional updates.

CONCLUSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. References

The Expert Advisory Group's report has provided a valuable stimulus towards a reappraisal of the optimum configuration of cancer services and its implementation is being vigorously pursued in some regions. More reflection may be necessary to ensure that the critical elements of effective management for each site of cancer are not overlooked in a precipitous rush towards structural changes.

Acknowledgements

We wish to acknowledge the help and advice given throughout the period of this study by Dr P. Prior and Mrs S. Wilson of the Centre for Cancer Epidemiology, and Mrs J. Kennedy of the North West Regional Cancer Registry.

References

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. References
  • 1
    An Expert Advisory Group on Cancer. A Policy Framework for Commissioning Cancer Services [report to the Chief Medical Officer of England and Wales]. London : HMSO, 1995.
  • 2
    International Union Against Cancer. TNM Atlas: Illustrated guide to the Classification of Malignant Tumours. New York : Springer-Verlag, 1992.
  • 3
    Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958; 53: 457481.
  • 4
    Cox DR. Regression models and life tables. J R Stat Soc 1972; 34: 187220.
  • 5
    Kehoe S, Powell J, Wilson S, Woodman C. The influence of the operating surgeon's specialisation on patient survival in ovarian carcinoma. Br J Cancer 1994; 70: 10141017.
  • 6
    Junor EJ, Hole DJ, Gillis CR. Management of ovarian cancer: referral to a multidisciplinary team matters. BrJ Cancer 1994; 70: 363370.