Intended for healthcare professionals

Education And Debate

Ovarian ablation in breast cancer, 1896 to 1998: milestones along hierarchy of evidence from case report to Cochrane review

BMJ 1998; 317 doi: https://doi.org/10.1136/bmj.317.7167.1246 (Published 31 October 1998) Cite this as: BMJ 1998;317:1246
  1. Michael J Clarke, overviews' coordinator (mike.clarke{at}ctsu.ox.ac.uk)
  1. Clinical Trial Service Unit, Radcliffe Infirmary, Oxford OX2 6HE

    One of the earliest randomised trials of a treatment for cancer was of ovarian ablation for breast cancer. Apart from surgical removal of the tumour, this is the oldest treatment still used in breast cancer. Its use was first reported in the Lancet in 1896,1 but not until the 1940s was it first assessed in randomised trials and it was nearly a century before the Early Breast Cancer Trialists' Collaborative Group provided a reliable estimate of its effect on resectable breast cancer.2 Now, 50 years after its completion, the first randomised trial of ovarian ablation has been included in the first of the collaborative group's overviews to be prepared as a Cochrane review.3 Here I describe this journey along a hierarchy of evidence and show why the research is still relevant.

    Summary points

    The hierarchy of evidence needed to determine treatment efficacy ranges from case reports, through case series and comparison with historical groups, to randomised trials and systematic reviews

    Over the past 100 years the investigation of ovarian ablation for women with breast cancer has progressed along this hierarchy

    The first case report was published in 1896 and the first randomised trial began in 1948

    Not until 1992 did a systematic review show this treatment to be clearly effective

    Hierarchy of evidence

    David Byar described an eight point hierarchy for the quality of evidence needed to determine a treatment's efficacy. This leads from anecdotal case reports, through case series and comparison with historical groups, to randomised trials and systematic reviews.4 Olkin added a ninth point: a meta-analysis in which individual patient data from each relevant trial are collected and analysed.5 He described this as being in the “realms of luxury,” but several examples now exist.6

    Occasionally, the evidence that a treatment is effective becomes apparent without needing to progress far along the hierarchy of evidence. A randomised trial is not necessary to show that transfusing blood into someone with severe haemorrhage is beneficial or that attempting to restart the stopped heart after a heart attack will increase the likelihood of survival. Unfortunately, such clearly effective treatments are uncommon. The differences between treatments on major endpoints such as death are usually much less dramatic but may still be important. For example, the widespread use of a treatment which reduced 10 year mortality among women with breast cancer by “only” 5-10% could mean that, worldwide, several tens of thousands more women would survive for at least this length of time. Thus, although the percentage difference might not be large, the potential global impact could be.

    It would clearly be worth knowing about such a treatment, with the knowledge based on reliable evidence. It should come from as far along the hierarchy of evidence as possible, ideally from large scale randomised evidence. This can come from systematic reviews of past randomised trials; new, large randomised trials; and ultimately the combination of the two. The process should be continuous. Reviews often raise new questions to be addressed by future trials, which should then be combined in updated reviews. Reviews should be systematic and up to date. This underpins the work of the Cochrane Collaboration, which prepares and maintains systematic reviews on all aspects of health care.

    From case reports to case series

    Perhaps the earliest suggestion that the removal of a woman's ovaries might help treat her breast cancer was by Schinzinger in 1889.7 However, George Beatson, a surgeon in Glasgow, was the first to publish a case report. He postulated an association between the hormonal action of the ovaries and the proliferation of breast cells and tested this in June 1895. A 33 year old woman had noticed a small lump on her left breast when breast feeding her first child. This lump got larger after the birth of her second baby and she went to the Glasgow Royal Infirmary. A 12 cm tumour was removed but the cancer was already far advanced. She was referred to the Glasgow Cancer Hospital, and Beatson removed both her ovaries. He reported the case to the Edinburgh Medico-Chirurgical Society and published details in the Lancet in July 1896.1 The woman survived for nearly four years before dying of recurrent disease.8

    Beatson also described other cases treated successfully and concluded that ovarian ablation was beneficial, providing the women were selected carefully.8 In 1905, Lett reported on 99 women, of whom 24 experienced a “very marked improvement.”9 And when Schinzinger returned to the subject in the same year he was able to summarise the results of several case series.10 However, ovarian ablation by surgery declined with the arrival of radiotherapy as an alternative means of stopping ovarian function.11

    Non-randomised comparative studies

    Many early case series contained descriptive comparisons with what might have been expected if ovarian ablation had not been used. The next step along the hierarchy of evidence was taken when investigators made direct comparisons between groups of patients with advanced breast cancer treated with and without ovarian ablation. These studies were usually based on historical comparisons. Investigations of ovarian ablation for women with cancer that was localised to the breast also began to be reported. These studies produced apparently conflicting conclusions. In 1939, Taylor compared 47 women who received ovarian irradiation with 50 women who did not. He concluded that irradiation conferred no advantage.12 Another analysis compared 60 women who were surgically ablated and 44 who received ovarian irradiation with 576 who had radical mastectomy alone. This study was more promising for ovarian ablation.13 However, McWhirter's analysis of nearly 800 women found almost no difference, and his department abandoned the adjuvant use of ovarian irradiation.14 Further case series and non-randomised comparative studies were subsequently reported.15 But this period also coincided with the emergence of the randomised trial as a means of studying efficacy of treatments.

    Randomised trials

    The first randomised trial of ovarian ablation in breast cancer was started by Ralston Paterson at the Christie Hospital, Manchester, in 1948. Initially, women were allocated to ovarian irradiation or control using shuffled, sealed envelopes. These were prepared in advance and the next one would be opened to determine each woman's treatment. This concealed method of allocation was later changed so that a woman's month of birth was used to determine her treatment in order to accommodate an additional investigation of local radiotherapy.16Unfortunately, such an allocation method causes problems since prior knowledge of the treatment to be allocated may affect whether a patient is entered into the study and lead to biased results.

    The first trial to include several hundred patients was started in 1957 by Roar Nissen-Meyer at the Norwegian Radium Hospital in Oslo. He minimised the possibility of foreknowledge of the allocated treatment by ensuring that allocation was done by a telephone call to an independent person. He included 336 premenopausal women over six years.15During this time five similar trials began in the United States and Canada. These included the National Surgical Adjuvant Breast Project B-03 trial; this group represented the first formal collaborative effort to use randomised trials to assess treatments for breast cancer.


    Embedded Image

    In the early 1960s some people felt that sufficient evidence had accumulated on the effect of ovarian ablation on recurrence. An answer in the Any Questions? section of the BMJ in 1963 noted that contemporary “adequately controlled statistical evidence” meant that “there is no doubt” that ovarian ablation would reduce the incidence of recurrence. Survival was not mentioned. It recommended that the decision to use ovarian ablation should usually be made by the surgeon and the family doctor alone because of the difficulty they might experience in discussing this treatment with the woman or her husband.17

    However, considerable uncertainty remained about the benefits of ovarian ablation. Another series of randomised trials started in the 1970s. By the early 1980s about 3500 women had been randomised into 13 trials worldwide in the 35 years since the Christie Hospital trial began. Taken separately, the results from these trials did not resolve whether ovarian ablation might be a worthwhile treatment in resectable breast cancer. The next step along the hierarchy of evidence was needed: a systematic review.

    Systematic review of randomised trials

    The Early Breast Cancer Trialists' Collaborative Group was established in the early 1980s to bring together all randomised trials of hormonal therapy or chemotherapy for resectable breast cancer. Limited data were sought on every woman randomised into trials that began before 1985. After central data processing and analysis the results from these trials were combined and reviews for tamoxifen and chemotherapy were published in 1988.18More detailed analyses and the findings for ovarian ablation appeared two years later. This report contained data from six ovarian ablation trials and noted “although moderately promising… more follow-up of the existing trials is needed, and more patients need to be randomized in the future with respect to some surgical, radiotherapeutic or hormonal form of ablation of ovarian function.”19

    The second cycle of the collaborative group's overview sought further follow up from the hormonal therapy and chemotherapy trials and brought trials of local therapy (radiotherapy and surgery) into the process. The reviews for hormonal therapy, immunotherapy, and chemotherapy were published in 1992.2 Ten trials of ovarian ablation were included and a significant benefit was shown not only in the delay of recurrence but also in overall survival. The paper was accompanied by an editorial calling for a re-examination of this “clearly effective” treatment.20

    This re-examination has taken place in several ways. The overview underwent its third cycle, with a further five years of follow up information being sought for the trials included previously. Data also became available for the first time on two additional trials, including the original Christie Hospital study. The findings of the third cycle were published in 1996,21 and an associated Cochrane review was published this year.3

    In total, data were collected on 3456 women, 2102 of whom were aged under 50 at randomisation. Although this is a relatively small number of women, many had died (1130 (54%)) because most of the trials were several decades old. The estimates of the effects of ovarian ablation on long term survival were therefore reasonably reliable. Among women who were under 50 at randomisation, 47.6% of those allocated ovarian ablation had died by 15 years after primary treatment compared with 53.9% of the controls. This 6.3% (SD 2.3) absolute improvement in survival is highly significant (log rank P=0.001) and the annual rate of death beyond the 15th year continues to be smaller among those allocated to ablation. 3 21 Though the difference is moderate, it is statistically reliable and clinically important: six deaths were avoided in the first 15 years for every 100 women treated. It provides solid evidence of the effect of ovarian ablation on survival against which other factors can be considered—not only loss of fertility but also possible side effects such as acute menopausal symptoms and long term effects of early menopause. It also shows the importance of hormonal therapy for younger women with breast cancer and, thereby, reinforces the recent finding that the hormone blocking drug tamoxifen could prevent an additional 20 000 deaths a year worldwide if used more widely for women with breast cancer.22

    Hierarchy of evidence on ovarian ablation for breast cancer

    View this table:

    Randomised trials of ovarian suppression by drugs started in the late 1980s to investigate whether the benefit seen with ovarian ablation by surgery or radiotherapy could be obtained with this form of treatment. These studies, which are either continuing or have recently closed, have recruited more premenopausal women than are currently available for analysis in the overview. In addition, studies of ovarian ablation by radiotherapy or surgery continue. The largest of these, in China, has randomised more than 3000 women in the six years since it began. These trials will contribute important new data to the next cycle of the collaborative group's overview, which will begin in 1999. They show the renewed importance of ovarian ablation as a possible treatment for women with breast cancer.

    Conclusion

    Obtaining the best evidence to help guide treatment should be a fundamental aim of those engaged in clinical research. They must strive to ensure that their findings are as reliable as possible. Achieving this goal often requires a journey along a hierarchy of evidence. This typically begins with promising case reports and progresses through non-randomised comparisons to randomised trials and systematic reviews or meta-analyses of these trials. Occasionally, it can also lead to collaborative overviews in which individual patient data from each trial are used in a centrally conducted analysis. The treatment of breast cancer by ovarian ablation provides an example of such a journey.

    Acknowledgments

    I thank the thousands of women who took part in the randomised trials discussed here and the thousands more who continue to take part and Kay Dickersin, Liz MacKinnon, Andy Oxman, Richard Peto, and Dave Sackett for their helpful comments.

    Footnotes

    • Competing financial interest None declared.

    References

    View Abstract