Lactation and breast cancer risk

Transcription

1 International Epidemiological Association 1999 Printed in Great Britain International Journal of Epidemiology 1999;28: Lactation and breast cancer risk H Furberg, a B Newman, a P Moorman b and R Millikan a Background Data from the Carolina Breast Cancer Study, a population-based, case-control study of breast cancer in African-American and white women residents of North Carolina, were evaluated to determine whether specific aspects of lactation are associated with a reduction in the risk of breast cancer. Methods Analyses included 751 parous cases and 742 parous controls frequency-matched on age and race. Information on lactation, reproductive history, lifestyle characteristics and family history were obtained through a personal interview. Results When women who breastfed were compared to those who never breastfed, odds ratios and 95% confidence intervals of 0.8 ( ) and 0.7 ( ) were found for women years and years, respectively. Similar inverse associations were observed for each of three categories of lifetime duration (1 3, 4 12, 13+ months). The inverse associations persisted and did not vary when number of children breastfed, ages at first and last lactation and lactational amenorrhoea were examined. Conclusions Our findings suggest that any lactation, regardless of duration or timing, is associated with a slight reduction in the risk of breast cancer among younger and older parous women. Keywords Breast cancer, lactation, breastfeeding, risk factors, epidemiology, female Accepted 12 October 1998 The hypothesis that lactation reduces risk of breast cancer has been investigated in numerous epidemiological studies, but results remain inconclusive In 1970, a large, international, case-control study reported no effect of lactation on breast cancer risk independent of age at first full-term pregnancy. 1 Regarded as definitive, these results led many researchers to ignore the effect of breastfeeding. In 1985, however, Byers et al. observed an independent protective effect of lactation among premenopausal women. 2 Several recent case-control studies have reported similar lower risks associated with breastfeeding limited mainly to premenopausal women, 3 14 but two cohort studies 15,16 and other case-control studies have reported no association. Previous studies examined the effects of various aspects of lactation on risk of breast cancer, including the role of ever having breastfed, duration of breastfeeding, number of children breastfed, and most recently, age at first lactation. Many found a lower risk of breast cancer with increasing duration of lactation. 2,4 6,8 12,13 The results of studies that examined number of children breastfed have been inconsistent. 3,9,11,17 Three studies reported that women who began breastfeeding at younger ages experienced greater reductions in risk than women who began lactating at later ages. 3,6,14 The effects of age at last lactation and amenorrhoea during lactation have not been previously addressed. a Department of Epidemiology, University of North Carolina, CB# 7400, McGavran-Greenberg Hall, Chapel Hill, NC , USA. b Department of Epidemiology, Yale University. The present study used data from the Carolina Breast Cancer Study (CBCS), a population-based, case-control study of breast cancer in African-American and white women residents of North Carolina, to examine whether breastfeeding or specific aspects of lactation reduce the risk of breast cancer. Breastfeeding characteristics include lifetime duration of lactation, number of children breastfed, ages at first and last lactation, months without menstrual periods while breastfeeding, and the use of lactation suppressants. Material and Methods The Carolina Breast Cancer Study is a population-based, casecontrol study of breast cancer in African-American and white women residents of a 24-county area of central and eastern North Carolina. 23 Both cases and controls were sampled using a modification of randomized recruitment. 24 All women between the ages of 20 and 74 years who were diagnosed with a first, invasive breast cancer between 1 May 1993 and 31 May 1996, were eligible as cases for this study and were identified using the North Carolina Central Cancer Registry s Rapid Ascertainment System. 25 Sampling probabilities ensured approximately equal sample sizes in the four age-race groups: younger (20 49 years) African-American women, older (50 74 years) African-American women, younger white women and older white women; women of other races comprised 2% of the study population and were included with white women. Of the 1153 eligible and locatable cases, physicians refused to allow contact for 73 breast cancer patients (6%), and an 396

2 LACTATION AND BREAST CANCER RISK 397 additional 191 (17%) declined to participate. Thus, information from 890 cases was available for analysis, for an overall response rate of 77%. Participation rates for the four age-race groups were 83%, 81%, 74% and 68% for younger white and African- American cases and older white and African-American cases, respectively. Controls were drawn from a North Carolina Division of Motor Vehicles list for women aged years, and a US Health Care Financing Administration list for women aged years. Sampling probabilities for controls ensured approximate frequency-matching to cases by race and 5-year age groups. Of the 1245 eligible and locatable women, 404 (32%) declined to participate. Thus, information from 841 controls was available for analysis, for an overall response rate of 68%. Participation rates for the four age-race groups were 73%, 72%, 67% and 59% for younger white and African-American controls and older white and African-American controls, respectively. Information was collected by personal interviews and focused on established and potential risk factors for breast cancer, including family history of cancer, menstrual and reproductive history, and sociodemographic and lifestyle characteristics. Lactation history was obtained for each live birth, including the duration, number of months without periods while lactating, and use of lactation suppressants. These variables were summed across all pregnancies for analyses. Statistical analyses were restricted to parous women (751 cases [84%] and 743 controls [88%]). Analyses were performed separately for younger women (20 49 years), older women (50 74 years), and all women combined. The younger and older age ranges were used because of difficulties in assigning menopausal status unambiguously when surgery or hormone therapy is involved. This approximation reasonably characterized the older age-group, as 96% of both cases and controls reported being postmenopausal. However, approximately 31% of the cases and 32% of the controls in the younger age group were postmenopausal, primarily as a result of a surgical procedure. Thus, to control for residual confounding, menopausal status was included in all models involving younger women. Effect modification by menopausal status was also examined among younger women. Participants who were still menstruating were considered premenopausal, while women who reported reaching menopause naturally, surgically, or through radiation or chemotherapy for a condition other than breast cancer, were regarded as postmenopausal. Women who reported taking hormone replacement therapy were considered postmenopausal. There were 425 parous cases and 371 parous controls aged years, and 326 parous cases and 372 parous controls aged years. Sample sizes differed in some analyses due to occasional missing data. Odds ratios (OR) and 95% confidence intervals (CI) were calculated using unconditional logistic regression models that examined the association between various aspects of lactation and breast cancer after adjustment for other covariates including: age (in 5-year groups), race, parity (number of pregnancies longer than 6 months), age at last full-term pregnancy, current body mass index (kg/m 2 ), history of breast or ovarian cancer in a first-degree relative, and menopausal status (when relevant). Analyses were performed in SAS using Proc Genmod, 26 which permits the use of an offset term to take into account the sampling design. Results Selected characteristics of parous cases and controls are presented in Table 1. Cases had a younger mean age at menarche, older age at first full-term pregnancy, fewer children, lower body Table 1 Selected characteristics of parous cases and controls, years old, Carolina Breast Cancer Study Covariates Cases n = 751 (%) Controls n = 743 (%) Race African-American 307 (40.9) 309 (41.6) White 444 (59.1) 434 (58.4) Age at menarche (years) (22.9) 147 (19.9) (27.3) 200 (27.0) (28.3) 203 (27.4) (21.5) 190 (25.7) Age at first full-term pregnancy (years) (32.9) 211 (34.1) (36.8) 231 (37.9) (18.6) 115 (18.8) (11.7) 63 (9.2) Age at last full-term pregnancy (years) (29.6) 195 (26.4) (28.3) 246 (33.2) (26.3) 193 (26.1) (15.8) 106 (14.3) Parity (22.6) 139 (19.8) (36.6) 253 (36.1) (20.5) 147 (21.0) (20.3) 162 (23.1) Education High school 371 (49.4) 376 (50.6) High school 380 (50.6) 367 (49.4) Body mass index (kg/m 2 ) (22.3) 152 (20.6) (34.0) 222 (30.2) (18.0) 154 (20.9) (25.7) 208 (28.3) Waist-hip ratio (40.9) 324 (48.5) (38.9) 242 (36.2) (20.2) 102 (15.3) Cigarette smoker Ever 354 (48.5) 329 (46.9) Never 376 (51.5) 372 (53.1) Alcohol drinker Ever 513 (70.3) 490 (70.0) Never 217 (29.7) 210 (30.0) Family history of breast or ovarian cancer Yes 127 (17.4) 94 (13.1) No 603 (82.6) 624 (86.9) History of breast biopsy Ever 127 (17.0) 125 (16.9) Never 621 (83.0) 616 (83.1)

3 398 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY Table 2 Descriptive characteristics of lactation among parous cases and controls, Carolina Breast Cancer Study Variable Cases (%) Controls (%) Cases (%) Controls (%) Cases (%) Controls (%) Lactation Ever 155 (36.5) 151 (40.7) 155 (47.5) 205 (55.1) 310 (41.3) 356 (47.9) Never 270 (63.5) 220 (59.3) 171 (52.5) 167 (44.9) 441 (58.7) 387 (52.1) Lifetime duration lactation (months) Mean and SD 11.7± ± ± ± ± ± months 349 (34.0) 44 (32.6) 41 (28.5) 55 (30.7) 90 (31.3) 99 (31.5) 4 12 months 52 (36.1) 51 (37.8) 43 (29.8) 64 (35.8) 95 (32.9) 115 (36.6) 13 months 43 (29.9) 40 (29.6) 60 (41.7) 60 (33.5) 103 (35.8) 100 (31.9) Age at first lactation (years) Mean and SD 26.9± ± ± ± ± ± years 10 (6.9) 17 (12.6) 56 (39.2) 66 (37.1) 66 (23.0) 83 (26.5) years 36 (25.0) 23 (17.0) 53 (37.0) 70 (39.3) 89 (31.0) 93 (29.7) years 55 (38.2) 58 (43.0) 18 (12.6) 27 (15.2) 73 (25.4) 85 (27.2) 30 years 43 (29.9) 37 (27.4) 16 (11.2) 15 (8.4) 59 (20.6) 52 (16.6) Age at last lactation (years) Mean and SD 29.4± ± ± ± ± ± years 9 (6.3) 11 (8.2) 21 (14.6) 36 (20.1) 30 (10.4) 47 (14.9) years 17 (11.8) 15 (11.1) 41 (28.5) 60 (33.5) 58 (20.2) 75 (23.9) years 42 (29.2) 40 (29.6) 38 (26.4) 30 (16.8) 80 (27.8) 70 (22.3) years 50 (34.7) 44 (32.6) 25 (17.4) 35 (19.6) 75 (26.0) 79 (25.2) 35 years 26 (18.0) 25 (18.5) 19 (13.1) 18 (10.0) 45 (15.6) 43 (13.7) Number of periods missed while lactating (months) Mean and SD 5.6± ± ± ± ± ±9.3 0 months 36 (25.7) 42 (31.8) 46 (33.6) 75 (43.6) 82 (29.6) 117 (38.5) 1 3 months 43 (30.7) 32 (24.2) 33 (24.1) 39 (22.7) 76 (27.4) 71 (23.4) 4 6 months 19 (13.6) 14 (10.6) 16 (11.7) 19 (11.0) 35 (12.6) 33 (10.8) 7 12 months 29 (20.7) 27 (20.5) 16 (11.7) 18 (10.5) 45 (16.3) 45 (14.8) 13 months 13 (9.3) 17 (12.9) 26 (18.9) 21 (12.2) 39 (14.1) 38 (12.5) Lactation suppressant use Ever 242 (58.5) 207 (58.3) 165 (52.2) 179 (51.7) 407 (55.8) 386 (55.1) Never 172 (41.5) 148 (41.7) 151 (47.8) 167 (48.3) 323 (44.2) 315 (44.9) mass index, higher waist-to-hip ratio, and were more likely to have a history of breast or ovarian cancer in a mother or sister. Cases and controls were similar with respect to education level, drinking and smoking status, and history of breast biopsy. These patterns were generally consistent when younger and older parous women were examined separately. Lactation characteristics for cases and controls, stratified by age are described in Table 2. Fewer cases reported ever breastfeeding than controls across all age groups. Among women who breastfed, the mean lifetime duration was slightly longer for cases than controls (13.7 months versus 12.9 months, respectively, for all women). Older women reported lactating longer than younger women. Cases and controls were similar with respect to ages at first and last lactation, though older women had a mean age at first lactation of approximately 27 years, while older women began lactating earlier, on average, than younger women (22 years versus 27 years respectively). Younger women also reported a later age at last lactation (29 years) relative to older women (26 years). Cases and controls breastfed the same number of children, on average, but younger women breastfed fewer children than older women (two and three children, respectively). Younger cases and controls each reported missing approximately six menstrual cycles while lactating, while for older women, cases reported missing more months of periods while lactating than controls (eight for cases and five for controls). Ever having breastfed was inversely associated with risk of breast cancer (Table 3). When parous women who breastfed were compared to those who did not, OR (95% CI) of 0.8 ( ), 0.7 ( ) and 0.7 ( ) were found for younger, older and all women, respectively. A stronger inverse association was not observed for women who breastfed longer, when total lifetime duration of lactation was divided into three groups (1 3, 4 12, 13+ months). Results were unchanged when different coding schemes were used, including using months of lactation as a continuous variable (data not shown). When OR

4 LACTATION AND BREAST CANCER RISK 399 Table 3 Odds ratios (OR) and 95% confidence intervals (CI) for breast cancer according to lactation experience among parous cases and controls, Carolina Breast Cancer Study Variable OR a 95% CI OR b 95% CI OR a 95% CI Lactation Ever lactated Lifetime duration of lactation 1 3 months months months No. of children breastfed 1 child children children children a Adjusted for age, race, parity, age at first full-term pregnancy, family history of breast or ovarian cancer, body mass index and menopausal status. b Adjusted for age, race, parity, age at first full-term pregnancy, family history of breast or ovarian cancer and body mass index. Table 4 Odds ratios (OR) and 95% confidence intervals (CI) of breast cancer according to timing of lactation among parous cases and controls, Carolina Breast Cancer Study Variable OR a 95% CI OR b 95% CI OR a 95% CI Age at first lactation 19 years years 0.8 c years d years Age at last lactation 19 years years 0.7 c years years e e years a Adjusted for age, race, parity, age at first full-term pregnancy, family history of breast or ovarian cancer, body mass index and menopausal status. b Adjusted for age, race, parity, age at first full-term pregnancy, family history of breast or ovarian cancer and body mass index. c OR comparing women 24 years versus those who never lactated. d OR comparing women 25 years versus those who never lactated. e OR comparing women 30 years versus those who never lactated. were examined in relation to number of children breastfed, lower risks were generally observed, but a dose-response relationship was not seen. Timing of lactation also did not appear to influence breast cancer risk (Table 4). Since the average age at first lactation varied between younger and older women, the exposure variables were categorized differently. Similar estimates were found for each age at first lactation range considered, with OR ranging from 0.7 to 0.8 among the younger women, 0.6 to 0.8 among older women, and 0.6 to 0.8 for women combined. Most, though not all, CI included 1.0. Residual confounding due to age at first full-term pregnancy could not be addressed, because it was highly correlated with age at first lactation. Small sample sizes resulted in unstable estimates when the effect of age at first lactation was examined using stratified analyses to control for age at first birth (results not shown). Assessment of age at last

5 400 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY Table 5 Odds ratios (OR) and 95% confidence intervals (CI) for breast cancer according to lactational amenorrhoea and lactation suppressant use among parous cases and controls, Carolina Breast Cancer Study Variable OR a 95% CI OR b 95% CI OR a 95% CI No. of periods missed while lactating Never lactated None months months months Lactation suppressant use Never Ever a Adjusted for age, race, parity, age at first full-term pregnancy, family history of breast or ovarian cancer, body mass index and menopausal status. b Adjusted for age, race, parity, age at first full-term pregnancy, family history of breast or ovarian cancer and body mass index. lactation produced similar results. Adjusting for age at last full-term pregnancy did not significantly change the observed associations; however, among younger women, this could not be examined since the two age variables were highly correlated (r = 0.95). The association between lactational amenorrhoea and lactation suppressant use and risk of breast cancer are summarized in Table 5. Compared to those who never breastfed, parous women years of age who breastfed but missed no menstrual cycles had an OR of 0.6. The OR for women who missed 1 6 months of menstrual cycles, 7 12 months, and more than 13 months were 0.8, 0.9 and 0.7. OR estimates for lactational amenorrhoea and breast cancer among the younger and older women also did not vary substantially: virtually all OR were less than 1.0, with overlapping CI. Lifetime duration of lactation did not appear to modify the effect of lactational amenorrhoea, as similar OR were found among women who breastfed for shorter ( 6 months) and longer ( 7 months) durations (data not shown). The use of lactation suppressants following pregnancy was reported by 58% of younger cases and controls and by 52% of older participants. It was not significantly associated with breast cancer risk in any age group. Discussion Our results suggest that breastfeeding is associated with a slight reduction in the risk of breast cancer among both younger and older parous women. As expected, the inclusion of nulliparous women in the never lactated reference category resulted in stronger inverse associations (data not shown). Similar inverse relationships have been demonstrated in previous studies. 5,6,8 10,12 14 The inverse association persisted, and did not vary appreciably, when specific aspects of lactation were examined in this study, including total lifetime duration, number of children breastfed, ages at first and last lactation, and number of months without periods while breastfeeding. In contrast, prior investigations have reported lower risks of breast cancer with increasing duration of lactation, primarily among premenopausal women. 2,4 6,8 12 In their study of Chinese women, Yuan et al. reported the strongest inverse association among women who breastfed for 3 years or more. 13 In our US study population, both cases and controls tended to breastfeed for relatively short durations. The median lifetime duration of lactation was 8 months for parous cases and controls, while the mean was 13 months. Thus, if extended breastfeeding is necessary to detect a dose-response relationship, this study had limited power to do so. This reflects one potential difficulty encountered by researchers evaluating the role of lactation duration in countries where women do not breastfeed for extended durations. Results from previous studies regarding the relationship between number of children breastfed and breast cancer provide conflicting results. 3,9,11,17 Our findings are consistent with those of Yoo et al. 11 and Negri et al. 17 who found no relationship with number of children breastfed. To address the timing of lactation, ages at first and last lactation were examined. Contrary to three previous case-control studies, 3,6,14 we did not find that younger age at first lactation further reduced risk of breast cancer among premenopausal or postmenopausal women. Since most women in the study breastfed their first child, it was not possible to assess the independent effect of age at first birth and age at first lactation. Age at first lactation and age at last lactation also were highly correlated in these data, producing similar OR estimates for associations with breast cancer. In 1993, the United Kingdom National Case-Control Study Group published results on the relationship between duration of amenorrhoea after a full-term pregnancy and breast cancer risk among women younger than 36 years. 9 After adjustment for a number of covariates, including duration of lactation, duration of postpartum amenorrhoea was not significantly associated with breast cancer. We also found that missing more menstrual cycles specifically during lactation did not decrease breast cancer risk. All durations of amenorrhoea showed modest inverse associations with breast cancer. These findings are contrary to expectation based on the model proposed by Pike et al. which hypothesizes a beneficial effect of decreased exposure to ovarian hormones during amenorrhoea. 30 The lack of an association between lactation suppressant use and breast cancer risk in all age groups is consistent with previous studies. 3,6,9,21

6 LACTATION AND BREAST CANCER RISK 401 Several hypotheses have been proposed to explain the inverse association between breastfeeding and breast cancer risk. One mechanism relates to a woman s reduced exposure to ovarian hormones during lactation. Since oestrogen and progesterone may play a role in breast cancer risk by increasing mitotic activity of breast cells, 29,30 reduced exposure to ovarian hormones during lactation might confer a protective effect. 2,5,6,29 Our findings did not support this theory, since neither longer durations of lactation nor longer periods of lactational amenorrhoea produced greater reductions in risk. However, as noted above, the variability in lifetime duration of lactation in this study may have been insufficient to test the relationship with breastfeeding duration. Another hypothesis suggests that a woman s overall body burden of potential carcinogens is reduced through elimination in breast milk, thereby reducing risk of breast cancer. 5,31 Lactation might also prevent breast cancer through apoptosis, whereby cells are disintegrated into membrane-bound particles that are then phagocytosed by other cells. 32 After a woman stops lactating, her breasts undergo involution, as cells are resorbed into the body. 33 Resorption of initiated cells might result in a decreased risk of breast cancer. The first theory would predict a stronger inverse association among women who breastfed for longer durations, while the second theory suggests a greater protective effect for women who breastfed multiple children. Neither of these effects was seen in our data. Moreover, these theories suggest that later age at last lactation may confer greater protection, since body burden of both potential carcinogens and initiated cells might be expected to increase with age. Once again, our findings did not show this. A fourth hypothesis suggests that the protective effect of lactation occurs through structural changes in breast tissue. Lactation causes epithelial cells to differentiate, potentially reducing their subsequent susceptibility to carcinogenesis. 5,29 Since we observed that any lactation reduced breast cancer risk, irrespective of timing or duration, a permanent change during first occurrence is most compatible with our results. The lack of a linear relationship between number of children breastfed and breast cancer also lends support to this hypothesis. In order to refine our understanding of the relationship between breastfeeding and breast cancer, future epidemiological studies should make distinctions between women who choose not to breastfeed, and those who are unable to, as inability to breastfeed might signify a hormonal imbalance or structural problem of the breast. If women who are unable to breastfeed successfully are at an increased risk of breast cancer, then failure to exclude them from the reference category of never lactated would falsely indicate a protective effect of lactation. Yang et al. found that breast cancer risk was particularly elevated for women who tried to breastfeed but were unsuccessful. 10 It has been suggested that more detailed breastfeeding histories should be collected in future studies, including frequency of breastfeeding and supplementation with formula, so that women can be divided into more homogeneous exposure groups for analyses. 34 While obtaining this additional information might be useful in theory, it would be difficult to accomplish using case-control designs, in which lactation experience is collected retrospectively, often many years after a woman breastfed. Some limitations should be considered when interpreting our results. In this study, overall response rates differed by case-control status, age and race of the participant. Women who declined the home interview were asked to complete an abbreviated telephone interview (unpublished findings). No questions about breastfeeding were asked, but we were able to compare the demographic characteristics and breast cancer risk factors of these women and those who consented to the full interview. Women who completed only the partial interview were slightly older, had less education and were less likely to have ever been married. Nevertheless, they reported similar numbers of children, on average, and age at first full-term pregnancy. For most characteristics assessed, differences between those fully participating and those agreeing to a partial interview were modest and in the same direction for cases and controls, minimizing the concern about selection bias influencing the study s results. Recall bias was unlikely to have influenced the results in this study. Breastfeeding is a defined event in a woman s life, and it has been demonstrated that women tend to remember accurately pregnancy-related events. 28 To further aid recall, the lactation questions were asked in the context of each pregnancy. Although the data were collected by nurses who knew the case status of participants, a standardized questionnaire was used, and lactation history was not the main focus of the interview. Stuver et al. suggested that studies reporting no association between breastfeeding and breast cancer among postmenopausal women were likely to have been influenced by non-differential misclassification due to poorer recall among older women. 22 This was not supported by our findings. The exposure most likely to be misclassified due to faulty recall is lactational amenorrhoea. While older women may have more difficulty recalling their reproductive history than younger women, there is little reason to suspect systematic differences in the way that cases and controls report the number of menstrual cycles missed while breastfeeding. In summary, the results of this population-based study suggest that breastfeeding, regardless of duration or timing, is associated with a lower risk of breast cancer. Considering the conflicting results in the literature, it is premature to advise women to breastfeed solely to reduce their breast cancer risk. However, it is important to consider that no studies have found a harmful effect of breastfeeding, and the benefits for the baby s health overwhelmingly supports its promotion. 35 Acknowledgements Supported by grant: National Cancer Institute SPORE in Breast Cancer 1P50-CA References 1 MacMahon B, Lin TM, Lowe CR et al. Lactation and cancer of the breast. A summary of an international study. Bull World Health Organ 1970;42: Byers T, Graham S, Rzepka T, Marshall J. Lactation and breast cancer: evidence for a negative association in premenopausal women. Am J Epidemiol 1985;121: Brinton LA, Potischman NA, Swanson CA et al. Breastfeeding and breast cancer risk. Cancer Causes Control 1995;6:

7 402 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY 4 Levin ML, Sheehe PR, Graham S, Glidewell O. Lactation and menstrual function as related to cancer of the breast. Am J Public Health 1964;54: McTiernan A, Thomas DB. Evidence for a protective effect of lactation on risk of breast cancer in younger women: results from a casecontrol study. Am J Epidemiol 1986;124: Newcomb PA, Storer BE, Longnecker MP et al. Cancer of the breast in relation to lactation history. N Engl J Med 1994;330: Siskind V, Schofield F, Rice D, Bain C. Breast cancer and breastfeeding: results from an Australian case-control study. Am J Epidemiol 1989; 130: Romieu I, Hernadez-Avila M, Lazcano E, Lopez L, Romero-Jaime R. Breast cancer and lactation history in Mexican women. Am J Epidemiol 1996;143: United Kingdom National Case-Control Study Group. Breast feeding and risk of breast cancer in young women. Br Med J 1993;307: Yang CP, Weiss NS, Band PR, Gallagher RP, White E, Daling JR. History of lactation and breast cancer risk. Am J Epidemiol 1993;138: Yoo KY, Tajima K, Kuroishi T et al. Independent protective effect of lactation against breast cancer: a case-control study in Japan. Am J Epidemiol 1992;135: Katsouyanni K, Lipworth L, Trichopoulou A, Samoli E, Stuver S, Trichopoulos D. A case-control study of lactation and cancer of the breast. Br J Cancer 1996;73: Yuan J-M, Yu MC, Ross RK et al. Risk factors for breast cancer in Chinese women in Shanghai. Cancer Res 1988;48: Freudenheim JL, Marshall JR, Vena JE et al. Lactation history and breast cancer risk. Am J Epidemiol 1997;146: Kvale G, Heuch I. Lactation and cancer risk: is there a relation specific to breast cancer? J Epidemiol Comm Health 1987;42: London SJ, Colditz GA, Stampfer MJ et al. Lactation and risk of breast cancer in a cohort of US women. Am J Epidemiol 1990;132: Negri E, Braga C, LaVecchia C, Levi F, Talamini R, Franceschi S. Lactation and risk of breast cancer in an Italian population. Int J Cancer 1996;67: Adami HO, Bergstrom R, Lund E et al. Absence of association between reproductive variables and the risk of breast cancer in young women in Sweden and Norway. Br J Cancer 1990;62: Brinton LA, Hoover R, Fraumeni JF Jr. Reproductive factors in the aetiology of breast cancer. Br J Cancer 1983;47: Kalache A, Vessey MP, McPherson K. Lactation and breast cancer. Br Med J 1980;280: Thomas DB, Noonan EA, the WHO Collaborative Study of Neoplasia and Steroid Contraceptives. Breast cancer and prolonged lactation. Int J Epidemiol 1993;22: Stuver SO, Hsieh C-C, Butone E, Trichopoulos D. The association between lactation and breast cancer in an international case-control study: a re-analysis by menopausal status. Int J Cancer 1997;71: Newman B, Moorman PG, Millikan R et al. The Carolina Breast Cancer Study: Integrating population-based epidemiology and molecular biology. Breast Cancer Res Treat 1995;35: Weinberg CR, Sandler DP. Randomized recruitment in case-control studies. Am J Epidemiol 1991;143: Aldrich TE, Vann D, Moorman PG, Newman B. Rapid reporting of cancer incidence in a population-based study of breast cancer: one contructive use of a central cancer registry. Breast Cancer Res Treat 1995;35: SAS Institue, Inc. SAS Procedures Guide. Release Cary, NC: SAS Institue, Inc., Kelsey JL, John EM. Lactation and the risk of breast cancer. N Engl J Med 1994;330: Paganini-Hill A, Ross RK. Reliability of recall of drug usage and other health-related information. Am J Epidemiol 1982;116: Petrakis NL, Wrensch MR, Ernster VL et al. Influence of pregnancy and lactation on serum and breast fluid estrogen levels: implications for breast cancer risk. Int J Cancer 1987;40: Pike MC, Spicer DV, Dahmoush L et al. Estrogens, progestogens, and normal breast cell proliferation, and breast cancer risk. Epidemiol Rev 1993;15: Heeschen WH. Harmful substances in human breast milk. Arch Gynecol 1985;238: Thomas CL (ed.). Taber s Cyclopedic Medical Dictionary. Philadelphia: FA Davis Company, 1989;17: Lund LR, Romer J, Thomaset N et al. Two distinct phases of apoptosis in mammary gland involution: proteinase-independent and -dependent pathways. Development 1996;122: Fenster DL. Letter to the Editor related to Newcomb et al. article. N Engl J Med 1994;330: Statement of the Standing Committee on Nutrition of the British Pediatric Association. Is breastfeeding beneficial in the UK? Arch Dis Child 1994;71: