Risk of breast cancer in relation to the interval since last full term pregnancyBMJ 1994; 308 doi: https://doi.org/10.1136/bmj.308.6945.1672 (Published 25 June 1994) Cite this as: BMJ 1994;308:1672
- P Cumming,
- J L Stanford,
- J R Daling,
- N S Weiss,
- B McKnight
- University of Washington School of Public Health and Community Medicine, Seattle, Washington 98112, USA
- Correspondence to: Dr Peter Cummings, Harborview Injury Prevention and Research Center, ZX-10, Seattle, WA 98104-2499, USA.
Objective: To examine whether the risk of breast cancer is increased by a recent term pregnancy.
Design: Population based case-control study.
Setting: Eight areas in the United States.
Subjects: Cases were 2279 multiparous women residents of eight areas aged 25-49 who were diagnosed as having breast cancer during 1980-2. Controls were 2357 multiparous women selected from the same areas by random digit dialling.
Main outcome measure: Relative risk of developing breast cancer according to the time interval since last full term pregnancy. Results - The distribution of intervals since the last term pregnancy was similar in cases and controls. Adjusted for age, parity, and age at first term pregnancy, the odds ratios observed for categories of years since the last full term pregnancy were: 0-2 years, odds ratio 1.16 (95% confidence interval 0.84 to 1.59); 3-6 years, odds ratio 1.21 (0.95 to 1.54); 7-9 years, odds ratio 1.04 (0.84 to 1.38); >=10 years, odds ratio 1.00 (reference).
Conclusions: Among multiparous women aged 25-49 years there was no association between the risk of breast cancer and the time interval since the last full term pregnancy.
The risk of breast cancer is thought to be reduced by a past history of pregnancy
Previous evidence suggests that a recent pregnancy (within three years) may transiently increase the risk of breast cancer
A controlled study of over 2000 multiparous women aged 25-49 with newly diagnosed breast cancer has found no association between risk of breast cancer and time since last pregnancy
Though it is generally agreed that a woman's risk of breast cancer is reduced by a history of pregnancy,1,2 some studies suggest this effect may be modified by age.3,4 Forty years ago Logan noted that married, parous women aged 35 or over had a lower breast cancer mortality than married nulliparous women or single women of corresponding age.3 However, below the age of 35 the direction of this association was reversed. A possible explanation is that pregnancy has two effects on breast tissue.1 Firstly, there is a short term deleterious effect, which increases the risk of cancer; and, secondly, there is a long term protective effect. To test this, several studies have examined the risk of breast cancer in relation to the time interval since the last term pregnancy, with conflicting findings.*RF 5-8* Because the question is unsettled we analysed data to test the hypothesis that a recent term pregnancy is associated with an increased risk of breast cancer.
Subjects and methods
We used data from the cancer and steroid hormone study, a large population based case-control study of risk factors for breast cancer.*RF 9-12* Cases were women aged 20-54 years who had primary breast cancer diagnosed during 1980-2. They were identified by eight population based tumour registries in the United States. Controls were women selected during the same time period from the same regions by random digit dialling. Controls were frequency matched to the age distribution of the cases. Study participants were interviewed in the home. Further details have been published.*RF 9-13*
A total of 5896 cases and 5698 controls were identified, of whom 4742 (80.4%) and 4754 (83.4%) respectively were interviewed.13 Twelve cases were removed from the data file because it was unknown if they had used oral contraceptives, and 66 controls were removed because they had a history of breast cancer. We excluded women under the age of 25 because there were only 16 cases in this age group and therefore reliable estimates of risk would not be possible. To allow direct comparison of our results with previous studies5,6 women aged over 49 were excluded. Only 32 women aged over 49 (12 cases, 20 controls) had reported a term pregnancy in the previous 10 years. Nulliparous women provide no information about the risk associated with recency of term pregnancies, so they were excluded. A term pregnancy was defined as any pregnancy which lasted over six months. Women with no pregnancy lasting over six months, with any pregnancy for which the length was unknown, and with unknown age at first term pregnancy were excluded (table I).
Age and age at first term pregnancy affect the risk of breast cancer,1,2 and adjustment was made for those variables. For primiparous women, current age minus age at first term pregnancy equals the time since last pregnancy, so that the effect of time since last term pregnancy cannot be assessed independently. For example, if a woman had her first and only term pregnancy at age 30 and developed breast cancer at age 35 the time since her last pregnancy is five years. Within each stratum of age and age at first term pregnancy there is no variation in years to last term pregnancy for primiparous women. Therefore, primiparous women were excluded.
Analysis - Logistic regression was used to control for confounding by other exposures.14 Time since last pregnancy was categorised as <3, 3-6, 7-9, and >=10 years to conform to previous reports.5,6 Other variables examined included those listed in table II, as well as months of breast feeding, age at menarche, menopausal status, family history of breast cancer, body mass index, marital status, race, education, income, geographic region, history of benign breast disease, number of spontaneous and induced abortions, past use of oral contraceptives, duration of oral contraceptive use, history of using oestrogens, past sterilising surgery, smoking history, and current pregnancy.
The distribution of years since last term pregnancy was similar in cases and controls (table II). The adjusted relative risk estimates for breast cancer in multiparous women differed little with time since last term pregnancy (table III). When time since last term pregnancy was treated as a continuous linear variable the relative risk estimate for each additional year since last term pregnancy was 0.9968 (95% confidence interval 0.9774 to 1.017). These estimates were adjusted for age, age squared, parity (classified as 2, 3, 4, 5, and >=6), and age at first term pregnancy (linear). Further adjustment for other variables (described above) had no important influence on these estimates.
There was no important association between time since last term pregnancy and breast cancer incidence within individual categories of parity or age at first term pregnancy. The relative risk estimates related to comparatively recent time intervals since last pregnancy were slightly higher for younger than for older women (table IV).
The first three years after the most recent term pregnancy were subdivided and adjusted relative risk estimates calculated (retaining >=10 years as the reference category). Results were: <1 year, odds ratio 0.91 (95% confidence interval 0.57 to 1.47); 1 year, odds ratio 1.17 (0.76 to 1.81); 2 years, odds ratio 1.36 (0.88 to 2.09). Thus there was no appreciably increased risk in any of the first three years after a term pregnancy.
Among multiparous women aged 25-49 there was no evidence that a woman who had had a recent term pregnancy was at increased risk of breast cancer as compared with a woman of the same parity whose last delivery had occurred earlier in life. A limitation of our analysis was that the association between breast cancer and the time interval since last pregnancy could not fully be separated from other measures of pregnancy and time, such as age at first term pregnancy. If the risk of breast cancer is related to a recent first term pregnancy our analysis was unable to detect it. No analysis can separate the effect of a first pregnancy from the joint effects of age and age at the time of the pregnancy.
The data we used were population based and response rates were high, reducing the likelihood of selection bias. Nevertheless, response rates could have differed with interval since last term pregnancy. Possibly women with a recent pregnancy might be more likely to be at home compared with other women and would be overrepresented among controls contacted by telephone. This bias would tend to lower the relative risk estimates associated with recent pregnancy. This bias, if present, was probably small in our series because the study called all telephone numbers on at least five occasions and at different times both on weekdays and at weekends.13
Our failure to find an association between the risk of breast cancer and the interval after pregnancy agrees with the results of two population based case-control studies. A Scandinavian group found no association between years since last pregnancy and the risk of breast cancer.7 Using nulliparous women as the reference population, the group obtained a relative risk of 0.6 (95% confidence interval 0.2 to 2.0) for the first year, 0.9 (0.4 to 2.0) for 1-4 years, and 0.8 (0.4 to 1.5) for >=5 years. Similar results were reported from Norway: the relative risk was 1.2 (95% confidence interval 0.9 to 1.5) for <=5 years since last pregnancy, 1.0 (0.8 to 1.2) for 6-10 years, and 1.0 (reference group) for 11-15 years.8 Though both studies included nulliparous and primiparous women, making it harder to interpret results, it seems unlikely that these inclusions could have masked a raised relative risk in the period soon after delivery.
Our findings conflict with two reports of hospital based case-control studies. An Italian study of 573 multiparous women with breast cancer diagnosed before the age of 50 reported an association between time since last pregnancy and the risk of breast cancer.5 The relative risk for 0-2 years was 2.7 (95% confidence interval 1.3 to 5.4), for 3-6 years 1.8 (1.1 to 2.9), for 7-9 years 1.4 (1.0 to 1.9), and for >=10 years 1.0 (reference group). Similar results were reported by British workers, who analysed data from 422 multiparous women with breast cancer diagnosed before the age of 50.6 They found a relative risk for 0-2 years after the last pregnancy of 2.9 (95% confidence interval 1.3 to 6.5), for 3-6 years 1.4 (0.8 to 2.6), for 7-9 years 0.8 (0.5 to 1.3), and for >=10 years 1.0 (reference).
Further subdividing the three years immediately after the last term pregnancy, the British group noted that the relative risk during the first year was 20.5, that the relative risk during the first year was 20.5, which fell to 1.4 during the next two years.6 The authors suggested that women with a recent pregnancy might avoid attending hospital and therefore be underrepresented in the control group. If their suggestion is correct this selection bias might explain why the hospital based studies seem to show a transient increase in the relative risk of breast cancer related to a recent term pregnancy while the three population based studies found virtually no association.
This research was supported in part by grants T32CA09168, RO1CA01364, and R35CA39779.
We acknowledge the contributors to the cancer and steroid hormone study. That study was designed and coordinated by the division of reproductive health, Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control. The principal investigator was George L Rubin, project director Phyllis A Wingo, and project associates Nancy C Lee, Michele G Mandel, and Herbert B Peterson. Data collection center principal investigators were: Raymond Greenberg (Atlanta); J Wister Meigs and W Douglas Thompson (Connecticut); G Marie Swanson (Detroit); Elaine Smith (Iowa); Charles Key and Dorothy Pathak (New Mexico); Donald Austin (San Francisco); David Thomas (Seattle); and Joseph Lyon and Dee West (Utah). Pathology review principal investigators were: Fred Gorstein, Robert McDivitt, and Stanley J Robboy. Project consultants were: Lonnie Burnett, Robert Hoover, Peter M Layde, Howard W Ory, James J Schlesselman, David Schottenfeld, Bruce Stadel, Linda A Webster, and Colin White. Pathology consultants were: Walter Bauer, William Christopherson, Deborah Gersell, Robert Kurman, Allen Paris, and Frank Vellios.