American Journal of Epidemiology Copyright 1996 by The Johns Hopkins University School of Hygiene and Public Health All

l rights reserved

Vol. 143, No. 6 Printed in U.S.A.

Breast Cancer and Lactation History in Mexican Women

Isabelle Romieu,1 Mauricio Hernandez-Avila,2 Eduardo Lazcano,2 Lizbeth Lopez,2 and Ricardo Romero-Jaime3 The authors conducted a case-control study in Mexico City between September 1990 and December 1992 to determine whether a dose-response relation could be observed between duration of lactation and the risk of breast cancer. Cases, women aged 20-75 years, were identified through six hospitals in Mexico City (n = 349) and were interviewed to obtain data on risk factors for breast cancer, including a detailed history of lactation. Controls (n = 1,005) were selected from the general population using the Mexican national sampling frame. Parous women who had ever lactated had a reduction in breast cancer risk (age-adjusted odds ratio (OR) = 0.39, 95% confidence interval (Cl) 0.25-0.62). A small decreasing trend of breast cancer risk in relation to duration of lactation (p < 0.001) was observed. Compared with parous women who had never breast-fed, women who had breast-fed for 12-24 months had an age-adjusted odds ratio of 0.47 (95% Cl 0.27-0.83). A stronger protective effect was observed with lactation duration for the first live birth among pre- and postmenopausal women (for 4-12 months of lactation, OR = 0.56 (95 percent Cl 0.32-0.96) and OR = 0.48 (95 percent Cl 0.29-0.81) in pre- and postmenopausal women, respectively). Adjusting for potentially confounding factors modified these results only slightly. The declining trend in fertility and lactation among Mexican women could lead to a major epidemic of breast cancer such as that observed in Western countries. Am J Epidemiol 1996;143:543-52. breast feeding; breast neoplasms; lactation

Downloaded from http://aje.oxfordjournals.org/ by guest on December 13, 2011

Results from recent case-control studies suggest a significant protective effect of prolonged lactation on breast cancer that is independent of parity, essentially among premenopausal women (1-12). Two cohort studies (13, 14) and various case-control studies (1520) have failed to find such an inverse association. However, these studies were largely confined to postmenopausal women, and most failed to examine the effect of total duration of lactation (3, 11) or included few women with prolonged lactation. Until recently, prolonged lactation has been the norm among Mexican women. However, to date there are no data on the potential impacts of such a practice on breast cancer incidence. Therefore, as part of an ongoing multicenter case-control study, we examined
Received for publication June 29,1994, and in final form June 30, 1995. Abbreviations: Cl, confidence interval; OR, odds ratio; SES, socioeconomic status. 1 Centra Pan-Americano de Ecologia Humana y Salud, Organizacion Pan-Americana de la Salud, Metepec, Mexico State, Mexico. 2 Centra de Investigaciones en Salud Publica, Instituto Nacional de Salud Publica, Cuernavaca, Morelos, Mexico. 3 Instituto Mexicano del Seguro Social, Mexico City, Distrito Federal, Mexico. Reprint requests to Dr. Mauricio Hernandez-Avila, Instituto Nacional de Salud Publica, Av. Universidad 655, Col. Santa Maria Ahuacatitlan, C.P. 62508 Cuernavaca, Morelos, Mexico.

the relation between breast cancer and lactation history, as well as whether a dose-response relation between lactation duration and breast cancer could be observed in a population of women among whom prolonged lactation is highly prevalent.
MATERIALS AND METHODS Cases

We conducted a case-control study of breast cancer, diet, and different reproductive factors in Mexico City between September 1990 and December 1992. Newly incident cases of breast cancer were identified among women younger than 75 years of age who had been residents of the metropolitan area of Mexico City for at least 2 years. Cases were identified from six hospitals belonging to the two major health providers in Mexico City: the Social Security system, which provides health care to the formally employed and their families, and the Ministry of Health hospitals, which provide health care to the population that is not in the formal sector or that is unemployed. Cases from the private sector were not included, because the hospital participation rate was low. The hospitals that participated in our study provided medical care to 80 percent of cases reported to the Mexico City Tumor Registry.
543

544

Romieu et al.

Cases were selected among women attending the gynecologic clinic of the hospital for biopsy of a breast lump. A total of 506 women agreed to participate and completed an interview. All women for whom the biopsy confirmed the diagnosis of breast cancer were included as cases in the study ( = 349). The participation rate among the cases was 95 percent.
Controls

Controls were an age-stratified random sample of residents of the Mexico City metropolitan area, identified from 3,694 households randomly selected from the National Household Sampling Frame (21). The selected households were visited by interviewers, who obtained a listing of women living there. Women aged 25-80 years who had been residents of the Mexico City area for at least 2 years were considered eligible controls. Limited numbers for each age group were established a priori to match the age distribution of cancer cases. Of the households visited, there was no eligible control in 28.4 percent, and the eligible control refused to be interviewed in 14.6 percent. Only one eligible control per household was included in the study. Interviewers administered a questionnaire that asked about sociodemographic variables, potential risk factors for breast cancer (including lifestyle habits), and reproductive and gynecologic history; also included was a food frequency questionnaire. Information on lactation history included duration of lactation for each live birth, identification of breastfeeding as bilateral or unilateral, and reason for having breast-fed unilaterally. Cases referred for a breast problem were interviewed at the hospital before the breast cancer diagnosis to avoid influencing their responses to the questionnaire due to knowledge of their disease. In fact, 40 percent (n = 151) of the women interviewed as potential cases were diagnosed as having a benign breast disease. Controls were interviewed in their homes.
Analysis

fined as 12 consecutive months of amenorrhea without an obvious cause. The effect of lactation on the risk of breast cancer was studied, examining the effect of ever having lactated, total duration of lactation, and duration of lactation for the first and second liveborn children. An index of socioeconomic status (SES) was developed as suggested by Bronfman et al. (24) by combining five variables (number of persons living in the house, number of rooms in the house (excluding the kitchen and bathroom), availability of drinking water, sanitary conditions, and education of the head of the family). Odds ratios and 95 percent confidence intervals were obtained from logistic regression models to evaluate the association between breast cancer and lactation and other potential risk factors, including age (in 5-year groups), SES (low, medium, or high), age at menarche (^11, 12, 13, 14, 15, or >16 years), age at first full-term pregnancy (<19, 20-29, or >30 years), parity (0, 1-2, 3-4, or S5), menopausal status (yes/ no), and family history of breast cancer (yes/no). Multivariate logistic regression was used to control for potentially confounding factors (25). Testing for linear trend in logit risks with increasing exposure was carried out using the likelihood ratio test under the assumption of a linear relation (25). To test for the existence of a trend associated with duration of lactation, we included a dichotomous variable for ever having lactated and a continuous variable for duration of lactation (0, 1-3, 4-12, 13-24, 25-36, 37-60, or ^ 6 1 months) simultaneously in logistic regression models.
RESULTS

Downloaded from http://aje.oxfordjournals.org/ by guest on December 13, 2011

The SAS (22) and Stata (23) systems were used for database management and statistical analysis. Univariate statistics were calculated for all variables. Variables measured on a continuous scale were explored in terms of their original distribution as well as in categories. Analyses were performed for all women and for preand postmenopausal women separately. Information on type of menopause (natural or surgical) was obtained by questionnaire. Natural menopause was de-

The associations between breast cancer and selected risk factors are presented in tables 1 and 2. SES was positively related to breast cancer risk among pre- and postmenopausal women. Late age at first birth increased the risk of breast cancer. When the birth occurred after age 29 years, women had an approximately twofold increase in the risk of breast cancer in comparison with women who had their first child at age 19 years or younger (age-adjusted odds ratio (OR) = 1.91, 95 percent confidence interval (CI) 1.12-3.24). This effect was stronger among postmenopausal women. Multiparous women had a lower risk of breast cancer than nulliparous women. We observed a decreasing trend in the risk of breast cancer with an increasing number of liveborn children (test for trend: p < 0.001). Family history of breast cancer was a strong risk factor among both pre- and postmenopausal women. Fourteen percent of c<>ses and 3.7 percent of controls had never had children and were excluded from the
Am J Epidemiol Vol. 143, No. 6, 1996

Breast Cancer and Lactation History

TABLE 1. Odds ratios for breast cancer according to selected risk factors, Mexico City, Mexico, 1990-1992*
Risk factor

545

No. of cases

No. of controls

Ageadjusted ORt

95% Clf

Multivariate*

g 5 % c |

Socioeconomic status Low Medium. High p for trend Age (years) at menarche <11 12 13
14 15 >16

68 149 132

259 523 223

1.00 1.15 2.48

0.82-1.60 1.74-3.52 <0.001

1.00 0.89 1.84

0.63-1.27 1.26-2.70 0.001

50 73 83
71

45 17

114 238 224 192 151 53

1.38 0.89 1.00 1.02 0.78 0.87

0.90-2.12 0.61-1.30 0.70-1.49 0.51-1.19 0.47-1.60 0.193

1.25 0.82 1.00 1.02 0.82 0.93

0.80-1.98 0.55-1.22 0.68-1.53 0.53-1.28 0.48-1.77 0.387 Downloaded from http://aje.oxfordjournals.org/ by guest on December 13, 2011

p for trend Age (years) at first birth

<20

20-29
>29

104 171 25

446 469 52

1.00 1.64 1.91

1.24-2.17 1.12-3.24 < 0.001

1.00 1.57 1.58

p for trend Parity 0 1-2 3-^1

5

1.17-2.11 0.90-2.75 < 0.005

48 73 103 125

37 200 312
455

1.00 0.41 0.33 0.20

p for trend Menopausal status No Yes Family history of breast cancer No Yes

0.24-0.70 0.20-0.55 0.12-0.32 <0.001

1.00 0.41 0.36 0.25

0.24-0.72 0.21-0.60 0.15-0.41 <0.001

129 220

526 479

1.00 0.86

0.57-1.31

1.00 0.99

0.63-1.55

335 14

988

17

1.00 2.38

1.14-4.93

1.00 2.42

1.13-5.17

* Totals may vary because of missing data. t OR, odds ratio; Cl, confidence interval. $ Each OR was adjusted for all other variables in the table.

analysis of the relation between breast cancer and lactation. Among parous women who had been diagnosed with breast cancer, 37 percent were premenopausal (n = 112) and 63 percent were postmenopausal (n = 189). Parous women who had ever lactated had a breast cancer risk of 0.47 (95 percent CI 0.30-0.73) compared with parous women who had never lactated. Adjusting by age, the protective effect of lactation increased slightly (OR = 0.39, 95 percent CI 0.250.62).,This effect was present among both premenopausal women (age-adjusted OR = 0.48,95 percent CI 0.26-0.89) and postmenopausal women (age-adjusted OR = 0.29, 95 percent CI 0.14-0.59) (tables 3 and 4). Duration of lactation, even short-term, was also associated with a reduction in breast cancer risk.
Am J Epidemiol Vol. 143, No. 6, 1996

Women who had breast-fed for up to 3 months had an age-adjusted odds ratio of 0.48 (95 percent CI 0.230.96) compared with parous women who had never breast-fed. Women who had breast-fed for 37-60 months had an age-adjusted odds ratio of 0.27 (95 percent CI 0.15-0.49), and women who breast-fed for more than 60 months had an odds ratio of 0.23 (95 percent CI 0.13-0.39). After adjustment for SES, parity, age at first full-term pregnancy, family history of breast cancer, and menopausal status, results remained similar; we also observed a significant decreasing trend in the risk of breast cancer associated with total duration of lactation (test for trend: p < 0.0005) (table 3). This protective effect was more evident for postmenopausal women. Among premenopausal women, a long duration of lactation did not substantially modify

TABLE 2. Premenopausal women No. of cases No. of controls 95% Clt 95% Cl No. of controls 95% Cl Multtvariatef OR No. of cases Ageadjusted ORt Ageadjusted OR Multlvariatef OR Postmenopausal women

Odds ratios for breast cancer according to selected risk factors, by menopausai status, Mexico City, Mexico, 1990-1992*

Risk (actor

95% Cl

Socioeconomic status Low Medium High p for trend

29 48 52
0.39-1.10 1.02-2.97 0.27-0.85 0.68-2.23 0.203 < 0.001 0.013 1.06-2.49 1.95-4.95

113 288 124

39 101 80

1.00 0.65 1.75

1.00 0.48 1.23

146 235 98

1.00 1.63 3.11

1.00 1.24 2.28

0.78-1.95 1.37-3.78 0.001

0.59-2.31 0.41-1.47 0.49-1.78 0.32-1.43 0.22-2.13 0.227

0.68-2.23 0.60-1.53 0.61-1.62 0.48-1.38 0.44-2.00 0.484

0.76-2.69 0.44-1.50 0.55-1.89 0.35-1.49 0.25-2.09 0.221 1.17 0.77 1.00 0.93 0.67 0.70

0.63-2.24 0.52-1.45 1.23 0.96 1.00 1.00 0.81 0.94 1.19 0.87 1.00 1.06 0.95 1.09 0.63-1.77 0.54-1.66 0.49-2.43 0.967

Downloaded from http://aje.oxfordjournals.org/ by guest on December 13, 2011

Age (years) at menarche 511 12 13 14 15 16 p for trend

26 26 28 27 15 5 69 134 104 99 75 26
1.43 0.81 1.00 1.02 0.72 0.72

24 47 57 44 31 12

45 104 120 92 77 27

40
64 271

217
25
<0.001 <0.001 1.00 1.22 1.43 0.79-1.90 0.59-3.44 1.00 1.27 1.34 0.79-2.02 0.53-3.34

64
107

229
198

17

27

1.00 1.96 2.20

1.36-2.83 1.13^.30 <0.001

1.00 18 .1 1.64

1.24-2.66 0.81-3.32 <0.001

Age (years) at first birth <20 20-29 >29 p for trend

Parity 0 1-2

17 39 45 28 13 151 214 147

1.00 0.21 0.13 0.09 0.09-0.50 0.06-0.32 0.03-0.21 <0.001 1.00 0.23 0.14 0.10

0.09-0.57 0.06-0.35 0.40-0.26 <0.001

31 34 58 97

24 49 98 308

1.00 0.59 0.53 0.26

0.29-1.19 0.28-1.00 0.14-0.48 <0.001

1.00 0.58 0.56 0.33

0.28-1.20 0.29-1.08 0.18-0.61 <0.001

p for trend

6'
121 8
1.00 5.06

Family history of breast cancer No Yes

520 6

1.70-15.1

1.00 6.46

1.96-21.2

214 6

468 11

1.00 1.21

0.44-3.34

1.00 1.19

0.41-3.44

* Totals may vary because of missing data. t OR, odds ratio; Cl, confidence interval. i Each OR was adjusted for all other variables in the table.

co

CO

Breast Cancer and Lactation History

TABLE 3. Odds ratios for breast cancer according to lactation experience, Mexico City, Mexico, 1990-1992*
Risk factor No. of cases No. of controls Ageadjusted ORf

547

95% Clf

Muttivariate* OR 1.00 0.54

5 / 9

CI

r|

Ever lactated No Yes Total duration of lactation (months)

0

36 265

58 905

1.00 0.39

0.25-0.62

0.33-0.89

1-3 4-12 13-24 25-36 37-60

>60

36 16 59 42 39
34 75

58 67 170 132
88 141 307

1.00 0.48 0.59 0.47 0.60 0.27 0.23

p for trend No. of children breastfed

0 1-2 3-4 36

0.23-0.96 0.34-1.00 0.27-0.83 0.33-1.08 0.15-0.49 0.13-0.39 <0.001

1.00 0.47 0.61 0.52 0.72 0.33 0.31

0.23-0.95 0.35-1.05 0.29-0.92 0.39-1.32 0.18-0.62 0.16-0.57 <0.001 Downloaded from http://aje.oxfordjournals.org/ by guest on December 13, 2011

76 80
108

>4 p for trend Duration of lactation with first live birth (months)
0

58 231 256 399

1.00 0.55 0.45 0.28

0.33-0.92 0.27-0.75 0.17-0.46 <0.001

1.00 0.57 0.52 0.39

0.34-0.96 0.31-0.87 0.21-0.70 < 0.001

1-3 4-12 >12 p for trend Duration of lactation with second live birth (months) 0 1-3 4-12
>12

66 56 140

39

133 189 430 215

1.00 0.70 0.53 0.24

0.45-1.08 0.37-0.77 0.15-0.39 < 0.001

1.00 0.68 0.62 0.28

0.43-1.06 0.42-0.91 0.17-0.47 < 0.001

90 43 100 43

257

123 351 167

1.00 1.09 0.67 0.53

p for trend

0.71-1.68 0.48-0.94 0.34-0.82 < 0.001

1.00 1.00 0.75 0.65

0.64-1.56 0.53-1.08 0.41-1.04 0.069

* Totals may vary because of missing data. No parous cases or controls were excluded from the analyses of the relation between breast cancer and lactation. t OR, odds ratio; Cl, confidence interval. t Adjusted for age, socioeconomic status, family history of breast cancer, menopausal status, age at first birth, and parity.

the results obtained for lactation up to 3 months in duration (table 4, figure 1). Long-term lactation was protective against breast cancer when lactation with the first live birth was examined. Women who had breast-fed their first child for 4-12 months had an age-adjusted breast cancer risk of 0.53 (95 percent CI 0.37-0.77). We observed a significant decreasing trend in the risk of breast cancer with increasing duration of lactation among pre- and postmenopausal women (tests for trend: p < 0.001 and p < 0.001, respectively). Results remained similar after adjustment for potentially confounding factors. Duration of lactation with the second liveborn child appeared to have a smaller impact on risk and was
Am J Epidemiol Vol. 143, No. 6, 1996

significant only among postmenopausal women (tables 3 and 4). We conducted additional analysis to determine the impact of lactation duration in months after accounting for whether a woman had ever breast-fed. We observed a decreasing trend in breast cancer risk with additional lactation time (/3 = 0.89, 95 percent CI 0.79-0.99) that was significant among postmenopausal women (|3 = 0.85, 95 percent CI 0.74-0.98). When lactation duration with the first child was considered, we observed a strong protective effect of additional lactation time (j3 = 0.74, 95 percent CI 0.62-0.87) for all women. This effect was observed separately for premenopausal (/3 = 0.79, 95 percent CI

TABLE 4.
Premenopausal women
No. of cases No. of controls

nenopausal status, Mexico City, Mexico, 1990-1992* Odds ratios for breast cancer according to lactation experience, by r
Postmenopausal women 95% Cl 95% Cl No. of cases No. of controls Ageadjusted OR Multivariat ' OR 1.00

Risk factor

Ageadjusteci ORt

95% Clt

Multivariatet

17
43 95

15
439 1.00 0.29

Ever lactated No Yes 466

1.00 0.48

0.26-0.89

1.00 0.58

0.30-1.11

19 170 0.14-0.59

0.43

0.20-0.92

Total duration of lactation (months) 0 17

1.00 1.00 0.44 8 34 14 17 9 13 55 127 90 49 61 84

43

15

1-3

4-12 13-24 25-36 37-60 0.72 0.39 0.80 0.28 0.26

22 25 62

19 8 25 28

>60

p for trend

0.17-1.15 0.36-1.45 0.17-0.87 0.36-1.79 0.11-0.71 0.11-0.63 <0.001 0.40 0.77 0.40 0.96 0.39 0.40
12 43 42 39 80 223

0.15-1.05 0.38-1.56 0.17-0.92 0.41-2.25 0.14-1.06 0.14-1.13 <0.001

1.00 0.60 0.45 0.52 0.43 0.23 0.19

0.19-1.88 0.19-1.04 0.22-1.20 0.18-1.02 0.10-0.53 0.01-0.41 < 0.001

1.00 0.65 0.46 0.59 0.54 0.29 0.27

0.20-2.05 0.19-1.11 0.25-1.41 0.22-1.32 0.12-0.68 0.11-0.64 <0.001

No. of children breastfed 17

36 38 20 43 179 157 122 19 40 42 88 15 52 99

Downloaded from http://aje.oxfordjournals.org/ by guest on December 13, 2011

0 1-2 3-4 >4

1.00 0.56 0.57 0.29 0.28-1.11 0.29-1.12 0.13-0.63 0.004 0.28-1.12 0.31-1.30 0.16-1.07 0.355

1.00 0.56 0.64 0.42

277

1.00 0.58 0.33 0.23

1.00 0.61 0.36 0.30 0.26-1.30 0.15-0.71 0.11-0.49 <0.001

p for trend

0.27-1.38 0.16-0.79 0.14-0.64 0.008

Duration of lactation with first live birth (months)

0 1-3 30 31 43 8 92 140 204 76

4-12

>12

1.00 0.72 0.56 0.24 0.40-1.29 0.32-0.96 0.10-0.58 <0.001

1.00 0.75 0.66 0.32

36 25 97 31

41 49 226 139

1.00 0.66 0.48 0.23

0.34-1.29 0.29-0.81 0.12-0.42 < 0.001

1.00 0.58 0.55 0.26 0.42-1.36 0.38-1.17 0.13-0.78 0.042

p for trend

0.29-1.17 0.32-0.96 0.13-0.50 <0.001

s! 37 167

C D

Duration of lactation with second live birth (months) 53

g21 30 13 86 155 57

0 1-3

<

4-12

>12

1.00 1.10 0.78 0.88

143

p for trend

0.60-2.00 0.45-1.33 0.43-1.80 0.444

1.00 1.12 0.93 1.12

22 70 30

90 37 196 110

1.00 1.10 0.58 0.41

0.58-2.09 0.38-0.91 0.24-0.70 < 0.001

0.61-2.07 0.53-1.63 0.52-2.41 0.767

1.00 0.89 0.62 0.49

0.46-1.73 0.38-0.99 0.27-0.88 0.012

* Totals may vary because of missing data. No parous cases or controls were excluded fromi the analyses of the relation between breast cancer and lactation. t OR, odds ratio; Cl, confidence intervaI. | Adjusted for age, socioeconomic status, family history of breast cancer, age at first birth, and parity.

6,1996

Breast Cancer and Lactation History

549

.5
post L .4
_

o
,AA A

J? .3
(0 (0

!&

5O

tot

.1

* /
pre

$ 0 1^S aSj^^^. * *"|;^^, K^MAIMMI - * ? V^_ B3FSKS *

% X

* ^ A .

?
100

^ !

r11 ^u>.

'

200

300

Downloaded from http://aje.oxfordjournals.org/ by guest on December 13, 2011

Cumulative lactation (months)

FIGURE 1. Risk of developing breast cancer among pre- and postmenopausal women in relation to lifetime accumulated months of breastfeeding, Mexico City, Mexico, 1990-1992.

0.72, 95 percent 0.61-1.00) and postmenopausal (E j CI 0.58-0.89) women. There was also a significant decreasing trend in breast cancer risk in relation to number of children breast-fed (test for trend: p < 0.001). Similar results were observed among pre- and postmenopausal women, although this protective effect was more pronounced among postmenopausal women (tables 3 and 4). SES appeared to be a strong risk factor for breast cancer in our population, and women who did not breast-feed were more likely to be of high SES (10.3 percent in the high SES group vs. 5.4 percent in the low SES group). Therefore, we repeated our analyses, stratifying data by SES. The observed odds ratios for ever lactating were 0.43 (95 percent CI 0.14-1.30) for women of low SES, 0.47 (95 percent CI 0.23-0.97) for women of medium SES, and 0.56 (95 percent CI 0.26-1.18) for women of high SES. We calculated the total number of months of lactation that took place during the first 6 months after delivery for each woman, to explore the impact of this period of lactation (the period when inhibition of ovulation and estrogen is thought to occur (26)). Controlling for known breast cancer risk factors, women with a lifetime accumulation of >12 months' lactation during the 6 months after delivery had a twofold decrease in the risk of breast cancer when compared with parous women who had never breast-fed. Results were similar for pre- and postmenopausal women (OR = 0.48, 95 percent CI 0.23-1.03, and OR - 0.39, 95 percent CI 0.18-0.86, respectively).
Am J Epidemiol Vol. 143, No. 6, 1996

DISCUSSION In this study, we found an inverse relation between lactation duration and breast cancer risk. The observed trend between duration of lactation and risk of breast cancer was small and was significant only among postmenopausal women. We observed that most of the protective effect was associated with lactation duration with the first live birth. This association was present among both pre- and postmenopausal women and persisted after potential confounders such as SES and parity were accounted for. Some authors have suggested that lactation with the first child and perhaps the second child may be of special significance (3, 8). Some limitations of our study require special consideration. Selection bias may have been a problem in our study to the extent that women with certain characteristics may seek care for a breast problem and be overrepresented in our case series. However, the bias would be introduced only to the extent that these characteristics are associated with breastfeeding patterns. We think it is unlikely that our results are explained by selection bias, for two major reasons. First, we observed that when we used as a control group women who visited a clinic for breast biopsy but were diagnosed as having benign breast disease, results were similar (data not shown). Second, we believe it is unlikely that a woman would be included in the study as a breast cancer case based on her lactation history, given that the information was collected before the

550

Romieu et al.

diagnosis of breast cancer and the interviewers were not aware of the hypothesis being tested. SES could act as a confounder or an effect modifier in that it was associated with risk of breast cancer, frequency of breastfeeding, and parity. However, our results were affected little when we stratified women according to their socioeconomic level. Because cases were interviewed at the clinics and controls in their homes, differences may have been introduced in responses by participants. However, the observation that results were similar when hospital controls were analyzed suggests that this source of bias is unlikely as an explanation for our findings. If this bias did occur, lactation histories would probably be more complete among breast cancer cases, leading to underestimation of the real protective effect of lactation. Our results are concordant with those of recent studies that have reported a protective effect of lactation on breast cancer risk (1-12). Some previous studies found a protective effect among parous women regardless of menopausal status (1, 7-9), whereas others found this effect in premenopausal women only (2, 10-12). In these studies, a significant dose-response effect was observed, although the minimal duration necessary to achieve a significant reduction in risk varied considerably (11). Earlier case-control studies (15-18) and two recent cohort studies (13, 14) did not find a protective effect of lactation. It has been suggested that failure to examine the relation between lactation and breast cancer risk according to menopausal status, as well as potential misclassification of data on lactation history, may explain discordant findings between studies (11). In our study, a large number of women had breastfed for an extended period (mean = 46 months, standard deviation 54 months), which resulted in high power for studying a dose-response effect of lactation on breast cancer risk. We observed a significant protective effect for as few as 3 months of lactation and a major protective effect after 36 or more months of lactation. However, among premenopausal women, we did not observe a significant decreasing trend in breast cancer risk in relation to total duration of lactation, although we did observe a strong protective effect when lactation duration with the first child was considered. Several factors may explain this. First, in our population, parity and duration of lactation were strongly correlated (r = 0.71), and therefore the specific effects of these factors on breast cancer risk were difficult to disentangle. Duration of lactation with the first child was less correlated with parity (r = 0.37), and its effect could be more accurately determined.

Second, the timing of lactation with the first child may be more important, because during the first pregnancy and lactation, the mammary cells undergo maturation, which may be protective against the carcinogenic process. In our population, women were more likely to lactate at an early age (mean = 21 years, standard deviation 5 years). It has been suggested that this is a protective factor which could act due to both the timing of exposure and the latency effect (12). Newcomb et al. (11) reported that no association of risk and lactation duration was observed for women who had lactated less than 10 years before diagnosis. This could partly explain the strong protective trend observed among premenopausal women in relation to lactation duration with the first child, given the relatively young age of the premenopausal breast cancer cases in our population (median age, 39 years). However, because data were sparse, we could not fully explore the modifying effect of age at first lactation. Different hypotheses may explain the protective effect of lactation. Direct action on the pituitary gland or on ovarian activity would change the woman's postpartum hormonal status, which would reduce estrogen levels through suppression of ovulation. As a result, total cumulative exposure to estrogen would be reduced, which could inhibit initiation or growth of breast cancer (3, 27). The strong protective effect on breast cancer risk that we observed using cumulative months of lactation in the early postpartum period (^6 months) tends to support this hypothesis. Another hypothesis to explain the protective effect of lactation is that it may "flush out" carcinogens (3). Experimental studies that have evaluated the induction of breast tumors in mice have documented more tumors in nipple-excised breasts than in normal breasts. The obstruction of the nipple resulted in prolonged exposure to carcinogens (28-30), whereas on the suckled side, the carcinogen was excreted in the milk. Some authors have also observed that exogenous chemicals are secreted into the breast ducts of nonlactating women, suggesting that normal physiologic secretory activity may result in exposure of breast epithelium to ingested or inhaled environmental carcinogens (31). Furthermore, comparisons of patients who had breast-fed unilaterally with patients who had breast-fed from both breasts indicate a significant increased risk of cancer in the unsuckled breast (32). Because breastfeeding is highly prevalent in Mexico, the reason for not breastfeeding could be associated with an underlying risk factor for breast cancer. Byers et al. (2) have suggested that successful lactation could be seen as an indicator of a normally balanced endocrine system, and unsuccessful lactation, conversely, as a correlate of an underlying hormonal
Am J Epidemiol Vol. 143, No. 6, 1996

Downloaded from http://aje.oxfordjournals.org/ by guest on December 13, 2011

Breast Cancer and Lactation History

551

imbalance that might cause increased subsequent risk of breast cancer. However, even if parous women who had not breast-fed had an underlying risk of breast cancer, this would not explain the dose-response relation that was observed between duration of lactation and breast cancer. During the interview, we did not collect information on the use of hormones to suppress lactation and reasons for not breastfeeding. However, in Mexico, hormonal treatments to suppress lactation are seldom used, and lactation is promoted throughout the public health sector. Published articles on studies conducted in Mexico mention that the most frequent reasons for not breastfeeding reported by mothers are: a perception of insufficient milk, a lack of confidence, a lack of knowledge regarding the nutritional value of maternal milk, rejection by the infant, a lack of postpartum breastfeeding guidance, the practice of providing artificial milk to new mothers in some health institutions, and health problems of the mother or the child (33). Lactation appears to play an important role in protecting women from breast cancer. Because of changes in women's lifestyles, lactation is decreasing in Mexico, and young women tend not to breast-feed or to shorten the duration of lactation (34). In association with the strong decline in the fertility rate, this decreasing trend in lactation may be responsible for the increasing incidence of breast cancer observed in the Mexican population. Although a better understanding of the physiologic changes that occur during lactation is necessary to determine the main mechanism by which lactation may protect women from breast cancer, breastfeeding is certainly a cost-effective intervention that should be promoted to protect the health of both mother and infant.

ACKNOWLEDGMENTS

This work was supported by funding from the American Institute for Cancer Research (United States), the Health Ministry of Mexico, the Pan-American Health Organization, and the Division of Environmental Health and Injury Control, Centers for Disease Control and Prevention, Atlanta, Georgia.

REFERENCES 1. Lubin JH, Burns PE, Blot WJ, et al. Risk factors for breast cancer in women from North Alberta, Canada, as related to age at diagnosis. J Natl Cancer Inst 1982;68:211-17. 2. Byers T, Graham S, Rzepka T, et al. Lactation and breast cancer: evidence for a negative association in premenopausal women. Am J Epidemiol 1985;121:664-74. 3. McTiernan A, Thomas DB. Evidence for a protective effect of lactation on risk of breast cancer in younger women: results Am J Epidemiol Vol. 143, No. 6, 1996

from a case-control study. Am J Epidemiol 1986; 124:353-8. 4. Katsouyanni K, Trichopoulos D, Boyle P, et al. Diet and breast cancer: a case-control study in Greece. Int J Cancer 1986;38:815-20. 5. Rosero-Bixby L, Oberle MW, Lee NC. Reproductive history and breast cancer in a population of high fertility, Costa Rica, 1984-85. Int J Cancer 1987;40:747-54. 6. Tao SC, Yu MC, Ross RK, et al. Risk factors for breast cancer in Chinese women of Beijing. Int J Cancer 1988;42:495-8. 7. Yuan JM, Yu MC, Ross RK, et al. Risk factors for breast cancer in Chinese women of Shanghai. Cancer Res 1988;48: 1949-53. 8. Siskind V, Schofield F, Rice D, et al. Breast cancer and breastfeeding: results from an Australian case-control study. Am J Epidemiol 1989;130:229-36. 9. Layde PM, Webster LA, Baughman AL, et al. The independent association of parity, age at first full term pregnancy, and duration of breast feeding with the risk of breast cancer. J Clin Epidemiol 1989;42:963-73. 10. Yoo K-Y, Tajima K, Kuroishi T, et al. Independent protective effect of lactation against breast cancer: a case-control study in Japan. Am J Epidemiol 1992;135:726-33. 11. Newcomb PA, Storer BE, Longnecker MP, et al. Cancer of the breast in relation to lactation history. N Engl J Med 1994;330: 81-7. 12. Yang CP, Weiss NS, Band PR, et al. History of lactation and breast cancer risk. Am J Epidemiol 1993;138:1050-6. 13. Kvale G, Heuch I. Lactation and cancer risk: Is there a relation specific to breast cancer? J Epidemiol Community Health 1987;42:30-7. 14. London SJ, Colditz GA, Stampfer MJ, et al. Lactation and risk of breast cancer in a cohort of US women. Am J Epidemiol 1990;132:17-26. 15. 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;42:185-94. 16. Kalache A, Vessey MP, McPherson K. Lactation and breast cancer. Br Med J 1980,280:223-4. 17. MacMahon B, Purde M, Cramer D, et al. Association of breast cancer risk with age at first and subsequent births: a study in the population of the Estonian Republic. J Natl Cancer Inst 1982;69:1035-8. 18. Brinton LA, Hoover R, Fraumeni JF. Reproductive factors in the aetiology of breast cancer. Br J Cancer 1983;69:1035-8. 19. Duffy SW, Roberts MM, Elton RA. Risk factors for breast cancer: relevance to screening. J Epidemiol Community Health 1983;37:127-31. 20. Brignone G, Cusimano R, Dardanoni G, et al. A case-control study on breast cancer risk factors in a Southern European population. Int J Epidemiol 1987;16:356-61. 21. Gutierrez G, Sepulveda J, Tapia R, et al. Encuesta Nacional Seroepidemioldgica. I. Diseno conceptual y metodologia. (In Spanish). Salud Publica de Mexico 1988;30:836-52. 22. SAS Institute, Inc. SAS procedures guide. Release 6.03 ed. Cary, NC: SAS Institute, Inc, 1988. 23. Stata Corporation. 1993 Stata reference manual. Release 3.1, 6th ed. College Station, TX: Stata Corporation, 1993. 24. Bronfman M, Guiscafre H, Castro V, et al. Estrategias para mejorar los patrones terapeuticos utilizados en diarrea aguda en unidades de atencion medica primaria. II. Medicion de la desigualdad: una estrategia metodologica, analisis de las caracteristicas socioeconomicas de la muestra. (In Spanish). Arch Invest Med (Mex) 1988; 19:351-60. 25. Breslow NE, Day NE. Statistical methods in cancer research. Vol 1. The analysis of case-control studies. Lyon, France: International Agency for Research on Cancer, 1980. (IARC scientific publication no. 32). 26. Gray HG, Campbell OM, Apelo R, et al. Risk of ovulation during lactation. Lancet 1990;l:25-9. 27. Henderson BE, Ross RK, Pike MC, et al. Endogenous hormones as a major factor of human cancer. Cancer Res 1982; 42:3232-9.

Downloaded from http://aje.oxfordjournals.org/ by guest on December 13, 2011

552

Romieu et al.
menopausal status and cerumen type with breast fluid secretion in nonlactating women, as determined by nipple aspiration. J Natl Cancer Inst 1975;54:829-34. 32. Ing R, Ho JHC, Petrakis NL. Unilateral breast-feeding and breast cancer. Lancet 1977;2:124-7. 33. Ysunza A. El abandono de la lactancia materna en Mexico. H Su causalidad. (In Spanish). La Rev Invest Clin (Mex) Suppl 1986;38:97-102. 34. Ysunza A. El abandono de la lactancia matema en Mexico. I. Tendencias recientes. (In Spanish). La Rev Invest Clin (Mex) Suppl 1986;38:41-6.

28. Kelsey JL, Berkowitz GS. Breast cancer epidemiology. Cancer Res 1988;48:5615-23. 29. Marchant J. Chemical induction of breast tumors in mice of the C57BI strain: the influence of pseudo-pregnancy, pregnancy and lactation on induction by methylcholanthrene. Br J Cancer 1961;15:568-73. 30. Brancifion C, Bonser GM, Cashera F. Chemically induced mammary tumours following unilateral excision of the nipples in pseudo-pregnant and lactating breeding BALB/c mice. Br J Cancer 1962,16:232-7. 31. Petrakis NL, Mason L, Lee R, et al. Association of race, age,

Downloaded from http://aje.oxfordjournals.org/ by guest on December 13, 2011

Am J Epidemiol

Vol. 143, No. 6, 1996