Abortion and breast cancer: a case-control record linkage study
- Unit of Health-Care Epidemiology, Department of Public Health, University of Oxford, Institute of Health Sciences, Old Road, Oxford OX3 7LF, UK
- Dr Goldacre ( )
- Accepted 21 November 2000
There is controversy about whether interruption of pregnancy, particularly if it is induced rather than spontaneous, increases the risk of breast cancer. Individual studies, and reviews summarising them, have given conflicting results.1 2 Recent guidelines from the UK Royal College of Obstetricians and Gynaecologists (http://www.rcog.org.uk) state that the evidence is inconclusive but that, when only those studies least susceptible to bias are considered, induced abortion does not seem to increase risk.
Most studies of this association have been case-control interview studies. An important and much discussed consideration is whether such studies are inherently subject to reporting bias—that women with breast cancer may be more likely than control women to tell the interviewer if they have had an induced abortion when questioned about their reproductive history.3-5 If there are systematic reporting biases in interview studies, neither pooling of data across studies in meta-analysis nor further similar studies will eliminate their effects. The overall odds ratio calculated in Brind's meta-analysis relied exclusively on data from 21 case-control studies.1 By contrast, prospective cohort studies and studies based on linkage of independent records cannot be influenced in this way by reporting bias. However, only three such studies have been published.6-8
We undertook a nested case-control study using linked, anonymised data from the Oxford record linkage dataset. Our methods for studying disease associations in this way are described in detail elsewhere.9 Data were analysed from 1968–1998 (that is, after the Abortion Act of 1967 which liberalised abortion) and comprise statistical abstracts of records of National Health Service (NHS) hospital admissions (including day cases) and death certificates. “Cases” were records of women with breast cancer (n = 28 616). “Controls” were records of women with a wide range of other, mainly minor, medical and surgical admissions (n = 325 456). Records were analysed, comparing cases and controls, to identify prior hospital admissions coded as induced abortion, spontaneous abortion, or abortion unspecified as induced or spontaneous. It was common for abortions to be coded without qualification of whether they were induced or spontaneous. Cases and controls were stratified by age (in five year bands), year of occurrence of case or control event, and place of residence. Limited stratification for social class was possible although the recording of social class data was incomplete.
Overall, previous abortion was slightly less common in women with breast cancer than in controls (table 1). We wondered if the lower rate might result from confounding with social class; but it was found in each social class stratum. The ratio of observed to expected cases did not increase over time either for all abortions or for induced abortions, as would be expected if abortions were a cause of breast cancer. When induced abortions were compared with other abortions, the ratios of observed to expected cases were not significantly different overall (table 1) or at individual time periods.
None of the cohort or record linkage studies have shown a significant increase in breast cancer risk after exposure to induced abortion.6-8 Like the Swedish linkage study,6 we found a slightly lower than expected rate of abortion before breast cancer. The most probable explanation for this is confounding with other reproductive or lifestyle variables outside the scope of our study. For example, if women who have abortions have more pregnancies and/or earlier pregnancies than other women, a slightly reduced risk of breast cancer might follow. Our data on abortions are substantially incomplete because they only include women admitted to hospital, only include those in the care of the National Health Service, and only in the time and area covered by the study. However, our use of control groups that are closely matched for these factors means that the relative rates of occurrence, comparing cases and controls, should be unbiased in these respects. Furthermore, our study includes appreciable numbers of women with an interval of 15 years or more between abortion and cancer; and the confidence intervals are small. The fact that the relative rates did not increase with increasing time intervals between abortion and breast cancer indicates that interruption of pregnancy, whether spontaneous or induced, does not increase the risk of breast cancer.
Funding: The Unit of Health-Care Epidemiology is funded by the South East Regional Office of the National Health Service Executive.
Conflicts of interest: none.