Survival outcome of care by specialist surgeons in breast cancer: a study of 3786 patients in the west of ScotlandBMJ 1996; 312 doi: https://doi.org/10.1136/bmj.312.7024.145 (Published 20 January 1996) Cite this as: BMJ 1996;312:145
- Correspondence to: Dr Gillis.
- Accepted 19 October 1995
Objective: To compare survival outcome for patients with breast cancer cared for by specialist and non-specialist surgeons in a geographically defined area.
Design: Retrospective study of all female patients aged under 75 years in the area treated between 1980 and June 1988 (before breast screening began). Patients were identified from the cancer registry and from pathology records of all hospitals in the area. Specialist surgeons were identified by one author. All other surgeons caring for patients from the area were considered non-specialists.
Setting: A geographically defined population in urban west of Scotland.
Subjects: 3786 patients with histologically verified breast cancer operated on between 1 January 1980 and 30 June 1988 and followed to 31 December 1993.
Main outcome measures: Five and 10 year survival rates for specialists and non-specialists; relative hazard ratios derived from Cox's proportional hazards model adjusted for prognostic factors—age, socioeconomic status, tumour size, and nodal involvement.
Results: The five year survival rate was 9% higher and the 10 year survival 8% higher for patients cared for by specialist surgeons. A reduction in risk of dying of 16% (95% confidence interval 6% to 25%) was found after adjustment for age, tumour size, socioeconomic status, and nodal involvement. The benefit of specialist care was apparent for all age groups, for small and large tumours, and for tumours that did and did not affect the nodes and was consistent across all socioeconomic categories.
Conclusions: Survival differences of the magnitude demonstrated have implications for the provision of services for the treatment of women with breast cancer. There is a need to improve equity in the treatment of breast cancer.
Accounting for variations in survival outcome in cancer treatment could lead to improvements in survival. The first stage is an accurate definition of the problem. Breast cancer affects one in 12 women in the United Kingdom. It is more a disease of the affluent than the deprived.1 We have shown that the affluent survive better even though the distribution of stage at presentation is the same in all socioeconomic categories.2 This present study arose from an attempt to investigate one aspect of this finding: whether the better survival observed in the affluent relates more to their treatment than to factors relating to socioeconomic status. We had information on patient identification, social class, stage of disease, and survival. We knew surgeons considered to have a specialist interest in breast cancer, thus identification of survival outcomes in relation to specialist and non-specialist care seemed the way forward.
Patients and methods
We identified all patients with breast cancer aged under 75 years whose disease was recorded in the west of Scotland cancer registry as histologically verified and who lived in a geographically defined area between 1 January 1980 and 30 June 1988 (when the United Kingdom national breast screening programme began locally).
The study population was defined by identifying the postcode sectors that formed the natural catchment areas for each of 10 hospitals concerned, which remain anonymous for reasons of confidentiality.
This procedure produced a population base of 1.5 million and 3786 histologically verified cases of breast cancer in women. We were unable to find the pathology records for only 2% of cases. Cases which were not histologically verified were excluded from the study. Of patients with breast cancer within the study area, 89% or more were treated at their local hospital. The pathology departments of all hospitals in the study area gave access to their data on tumour size and nodal involvement. Histological grade3 was available from only two of the 10 hospitals studied and applied to 412 patients from these.
The specialist surgeons were chosen by local perception. Over the time period of the study they each demonstrated the following indicators of specialist interest. These were setting up of a dedicated breast clinic; a defined association with pathologists and oncologists; organising and facilitating clinical trials; and maintaining a separate record of all patients with breast cancer in their care.
They kindly supplied the names of all patients in the care of their teams within the time period 1980 to mid-1988. A total of 918 patients were categorised in this way. The 2868 remaining patients were considered to have received non-specialist care. An independent scrutiny of surgeons' names on the pathology reports confirmed that the names of the patients supplied by the specialists were accurate and complete. We also confirmed that the remaining patients were indeed cared for by those categorised as non-specialists. The names of the non-specialists in the hospitals studied were not abstracted. Survival of all patients up to the end of 1993 was ascertained from death certificates provided by the registrar general (Scotland) to the cancer registration system. Information on socioeconomic status was derived for the postcode sectors included in the study area by using the Carstairs deprivation index.4
We carried out an initial analysis of survival outcome with adjustment for prognostic factors—age (entered as a continuous variable), deprivation category (as three separate categories), and tumour size (as three separate categories) but not nodal involvement—using Cox's proportional hazards model.5 An additional analysis was carried out that included nodal involvement as a prognostic factor with the categories node negative (based on four or more nodes sampled), node negative (based on one to three nodes sampled), and node positive. This took account of concern about the comparability of nodal information for patients cared for by specialists and non-specialists. Tumour size and the state of the nodes were used in preference to conventional staging as they provide a more precise measure of prognosis for an individual patient.6
Table 1 shows the five and 10 year survival rates together with the relative hazard ratios for specialist: non-specialist care. Survival of all patients aged under 75 years was 9% higher at five years and 8% higher at 10 years among those cared for by specialists compared with non-specialists. When the results were adjusted for the prognostic factors of age, deprivation category, and tumour size, a 17% reduction in the risk of dying (relative hazard ratio 0.83) was apparent for those cared for by specialist surgeons. These results remained virtually the same (relative hazard ratio 0.84) when nodal involvement was included as an additional prognostic factor.
Table 2 shows five year survival rates by tumour size, extent of nodal involvement, nodal status, and number of affected nodes found in patients cared for by specialist and non-specialist surgeons. The benefit of specialist care was apparent regardless of the size of the tumour, the extent of nodal involvement, nodal status, and the number of affected nodes found. This benefit was most obvious among the intermediate group of patients—that is, those with tumour sizes between 20 and 39 mm and those with affected nodes who had fewer than half the nodes affected. Those patients with information missing on both tumour size and nodal status were a group with a poor prognosis whether cared for by specialists or non-specialists.
Table 3 shows improvements in survival at all ages for those treated by specialist surgeons. The most clear difference occurred among those aged 50 to 64 years and the smallest among those between 65 and 74 years. The differences in survival were also consistent across the deprivation categories. The affluent were more likely to survive five years than the deprived, whether they were treated by a specialist or a non-specialist.
Table 4 shows the distribution of prognostic factors examined among patients cared for by specialists and non-specialists. The distribution of tumour size, where stated, was the same for specialists and non-specialists. Although specialists treated a slightly higher proportion of patients whose nodes were unaffected, when node status was reported there was a difference in the reporting, this being unavailable in 17% of patients cared for by specialists and 39% of those cared for by non-specialists. There was no significant trend in the proportion of cases from different socioeconomic groups seen by specialists and non-specialists.
Table 5 shows the number of nodes sampled for those women who were cared for by specialists and those cared for by non-specialists. Only 10% of women cared for by specialists had three or fewer nodes sampled whereas this figure rose to 38% for those cared for by non-specialists. These figures remained constant over the time period of the study.
The main finding in this paper is that in a geographically defined population (1.5 million men and women) the survival advantage of specialist care was 9% better at five years and 8% better at 10 years. This benefit was clear for large as well as small tumours, for patients with no nodes affected as well as extensive involvement, for each age group studied, and for the deprived as well as the affluent. The size of the improvement in survival is of the same order of magnitude as that found for systemic treatment in early breast cancer.7 We do not have evidence on reasons for the differences we have observed as information on the methods of treatment used was not obtained. The results of our own previous studies8 9 and others10 11 in various cancers suggest that the benefit associated with care provided by specialists may be due not only to surgery but to the use of additional therapies.
The strength of this study is that it is population based, comprising virtually all (98%) histologically verified cases in a geographically defined area, and is not subject to the biases associated with a case series based on one or a group of hospitals. This study has been conducted on a confidential basis. We therefore cannot name the hospitals studied, the specialist surgeons, nor the catchment areas. Also, there has been a considerable change in surgical staff since this study closed in June 1988.
The main problem in interpreting our results is ensuring that cases cared for by specialists and non-specialists are comparable. The personal and pathological prognostic factors available to us in patients with breast cancer were age, socioeconomic status, tumour size, extent of nodal involvement, and histological grade. There was no difference in the distribution of tumour size among patients seen by specialists and non-specialists. Comparison of the distribution of nodal involvement between specialists and non-specialists presented two problems. Firstly, information on nodal status was less precisely recorded and more often missing among patients cared for by non-specialists. Patients with data on both nodal status and tumour size missing had a poorer prognosis (five year survival rate 48%). Some of these presumably had advanced disease so had a diagnostic biopsy only. The information was unavailable to be recorded for these cases. Secondly, non-specialists were more likely to sample fewer nodes (table 5). This meant there was an increased likelihood of patients cared for by non-specialists being classified node negative when, had more nodes been sampled, they may have been node positive. Indeed, an analysis of survival among node negative patients, which included the number of nodes sampled as a predictive factor of mortality, showed an increased risk of dying when fewer than four nodes were sampled. Thus, two survival analyses, one with and one without nodal involvement as a prognostic factor, were carried out (table 1). The two analyses gave virtually identical answers for the magnitude of the benefit (relative hazard ratio 0.83 when nodal information was ignored and 0.84 when it was included). Thus our method of classifying nodal involvement as node negative (four or more nodes sampled), node negative (fewer than four nodes sampled), and node positive was a reasonable reflection of the distribution of nodal status.
Information on histological grade was available for only 412 patients. Its distribution was the same among patients cared for by specialists and non-specialists. Analysis of survival that included grade in these patients made no change to the size of the difference observed between specialists and non-specialists.
No data on other pathological prognostic factors were routinely collected over the time period of the study. There were no significant differences in the personal prognostic factors (age, socioeconomic status) of patients cared for by specialists and non-specialists. Thus, the benefit associated with specialist care was likely to be a true reflection of a better level of care provided by specialists rather than a reflection of the type of patients being seen by specialists.
A case has been made by the King's Fund consensus development conference that management of women with breast cancer should be carried out through specialist units,12 although there was no direct evidence then, or since, that survival was improved. Several studies have examined outcome in relation to other factors such as type of hospital,10 11 volume of patients,13 and use of chemotherapy13 14 and have observed a survival advantage associated with the more “specialist” approach (that is, treatment at a teaching hospital, larger workload, and greater use of adjuvant therapy). We believe that this paper is the first attempt to measure directly the benefit of being cared for by specialists.
How do we define specialist? This report is of a retrospective study based on cases diagnosed between 1980 and mid-1988 when specialisation was not an issue. It is difficult now to apply strict criteria for the definition of what then constituted a specialist because the pattern of care itself was also changing over the time period. The most important point to stress is that the specialists in this study were chosen without prior knowledge of the survival outcome of patients in their care but their special interest in breast cancer was known. In retrospect, specialist interest was characterised by their setting up dedicated breast clinics, having a defined association with pathologists and oncologists, and organising and facilitating clinical trials as well as maintaining a separate records' system of all breast cancer cases in their unit's care. Some evidence of the nature of the specialist approach lay in the complete correspondence of the specialists' own records and the relevant pathology records in accurately defining their own cases. Not all patients in the specialists' care group were operated on by the specialists themselves, which may indicate that training of junior staff by specialists is important.