Oral contraceptives and colorectal tumors

Oral contraceptives and colorectal tumors

ORIGINAL RESEARCH ARTICLE Oral Contraceptives and Colorectal Tumors A Review of Epidemiologic Studies Silvia Franceschi* and Carlo La Vecchia†‡ Over...

107KB Sizes 0 Downloads 42 Views

ORIGINAL RESEARCH ARTICLE

Oral Contraceptives and Colorectal Tumors A Review of Epidemiologic Studies Silvia Franceschi* and Carlo La Vecchia†‡

Over the last two decades, the mortality rates of colorectal cancer in many developed countries have declined in women but not in men. One of the explanations of this difference between the genders may be the favorable influence of the spread of exogenous female hormone use (ie, oral contraceptives [OC] and hormone replacement therapy). Reduced risk in ever-users of OC was found in three of four cohort studies available on this topic, and was significant in the one based on colorectal cancer mortality. The fourth one showed no difference. Of 11 case-control studies (or groups of studies), none showed significantly elevated risk. Five reported lowered colorectal cancer risk among ever-users, with a significant inverse association in the largest investigation available. Recent OC use, more than long-duration use, seemed to be related to some risk reduction. One cohort study and three case-control investigations suggested that OC use was not related to the risk of colorectal adenomatous polyps. Thus, at present, it seems that OC use does not increase and may even have a favorable effect on either incidence or mortality of colorectal cancer. CONTRACEPTION 1998;58:335–343 © 1998 Elsevier Science Inc. All rights reserved. KEY WORDS:

oral contraceptives, colorectal cancer, female

hormones

Evidence is strong for a substantial contribution of dietary habits, sedentary life, and genetics. Similar incidences between the genders are seen for colon cancer, whereas a male predominance is found for rectal cancer. The female:male ratio of colon cancer incidence is relatively higher at premenopausal age, further suggesting an influence of some biologic correlate of gender.1 Interestingly, over the last two decades, mortality rates in many developed countries have declined in women but not in men.2 The more favorable pattern in women may be due to earlier or greater dietary improvements than in men. A role of the exogenous female hormones (ie, oral contraceptives [OC] and hormone replacement therapy) is also possible and evidence on this from casecontrol and cohort investigations is accumulating rapidly. The present report will systematically review studies where the association between OC use and colorectal neoplasm has been evaluated. Colorectal cancer and adenomatous polyps will be considered separately. Wherever data are available, possible difference between subsites (ie, colon and rectum and, within colon, right and left colon) will be evaluated. In addition to ever-use, data on duration and recent use of OC will be examined to help find an interpretation.

Introduction

C

olorectal cancer is the most frequent cancer in nonsmokers (genders combined) in Western countries.1,2 International trends in the incidence and mortality of colorectal cancer are heterogeneous. Broadly, colorectal cancer rates correlate with the adoption of a Western lifestyle, and, depending on each country’s stage of transition, colorectal cancer incidence has been increasing (often rapidly), leveling, or declining in the last two or three decades. *Servizio di Epidemiologia-Centro di Riferimento Oncologico, Aviano; †Istituto di Ricerche Farmacologiche “Mario Negri,” Milano; and ‡Istituto di Statistica Medica e Biometria, Universita´ degli Studi di Milano; Milano, Italy Name and address for correspondence: Silvia Franceschi, M.D., Servizio di Epidemiologia, Centro di Riferimento Oncologico, Via Pedemontana Occ., 33081 Aviano (PN), Italy; Tel.: 139-434-659354; Fax: 139-434-659222; e-mail: [email protected] Submitted for publication July 28, 1998 Revised November 13, 1998 Accepted for publication November 16, 1998

© 1998 Elsevier Science Inc. All rights reserved. 655 Avenue of the Americas, New York, NY 10010

Subjects and Methods Articles considered were epidemiologic studies on colorectal neoplasms (ie, cancer, polyps, and adenomas) published as full papers in English up to September 1998 and including information on OC use. They were identified by reviewing reference lists in relevant papers, manual and computer (Medline, Cancerlit, Current Contents, Embase, etc.) searches in the literature, and discussions with colleagues. Search strategy included a range of synonyms of neoplasms, tumors, or cancer of colon or rectum, and of exogenous female hormones, oral contraceptives, estroprogestins, etc. No restriction by study size or study quality indicators was set, but studies were eligible only if information on hormone(s) was obtained from each woman and OC were accurately distinguishable from hormone replacement therapy. Thus, the ISSN 0010-7824/98/$19.00 PII S0010-7824(98)00128-0

336

Franceschi and La Vecchia

Contraception 1998;58:335–343

Table 1. Cohort studies on oral contraceptives (OC) and colorectal cancer

Reference

Country

Population (followup); no. with cancer

Relative risk (RR) (95% confidence interval), ever- vs. never-users Colonrectum

Colon

Rectum

Duration of use

Recent use

Adjustment and comments

RR $ 151 Age, parity, social after OC class and 5 1.00 smoking. Prevalence of OC use 5 49%. 0.84 0.64 0.76 Significant Not shown Age, BMI, (0.69–1.02) (0.40–1.02) (0.49–1.18) trend (RR exercise, cancer for $ 8 yr family history, use 5 0.60; aspirin, and 0.40–0.89) alcohol, meat intake, menstrual factors. Prevalence of OC use 5 32%, mostly past use. 0.5 Not shown Not shown Age, height, (p 5 0.47) parity, caloric intake, and vitamin intake. Prevalence of OC use 5 19%. 1.0 No trend Not shown Age only. (0.75–1.4) (RR for $ Adjustment for 5 yr use 5 education and 1.1; 0.66–1.8) BMI left RR unaltered. Prevalence of OC use 5 17%.

Beral et al.5

U.K.; RCGP OC Study

46,000 0.57 (25 years); (0.35–0.94) 170 deaths

Chute et al.6 Martinez et al.7

U.S.; Nurses Health Study

89,448 (12 years); 501

Bostick et al.8

Iowa, U.S.

35,215 (4 years); 212

Troisi et al.9

U.S.; BCDDP

57,528 (10 years); 203

No significant trend

RCGP, Royal College of General Practitioners; BCDDP, Breast Cancer Detection Demonstration Project; BMI, body mass index.

present review did not include one record-linkage cohort study,3 which reported no association of OC use with colorectal cancer, and one case-control study, 4 which showed a negative association with the use of any female hormone. A total of 19 studies or groups of studies was reviewed, including four cohort5–9 and 11 case-control investigations10 –23 on colorectal cancer and one cohort and three case control studies24 –27 on colorectal adenomas. Of these studies, only one5 had not yet been published. When more than one publication was available from the same study, only the most up-to-date results were reported in detail. For each study, data extraction and evaluation included relevant information on the study design, number of subjects (or person-years), age range, prevalence of OC use, and control of confounding factors. Primary analysis concerned the comparison of everusers vs. never-users of OC, but the influence of duration of use and recent use was assessed wherever possible. Also, the combination of cancers of the

colon and rectum was, in most instances, the primary outcome. A few investigations, however, were restricted to colon cancer, whereas a few others allowed separate assessment of colon and rectum or right and left colon (Tables 1 and 2). Colorectal adenomatous polyps, or adenomas, were also evaluated. No studies were excluded for purported weaknesses of design or data quality, thus allowing the reader a full evaluation of the available evidence. The measure of association considered was the relative risk (RR) and the corresponding statistical significance (generally 95% confidence interval [CI]). All given RR derived from original articles. Relative risks adjusting for as many potential confounding factors reported as possible (multivariate) were given, and major factors for which allowance had been made were mentioned. In the 1980s and 1990s the relationship between OC use and the risk of colorectal neoplasms was generally assessed as a secondary analysis in the framework of the study of reproductive and men-

Contraception 1998;58:335–343

strual factors in female colorectal cancer. Only recently7,16 has the issue started being addressed systematically. Because of the present heterogeneity of RR, particularly from case-control investigations, and the fast accumulation of new data, we attempted to provide a complete descriptive review, but not to compute a combined RR from available studies.

Results Cohort Studies Of three early prospective studies on OC (ie, Oxford Family Planning Association, the Walnut Creek Contraceptive Drug Study, and the Royal College of General Practitioners’ Oral Contraception Study), only the latter5 has provided information on the association of OC with colorectal cancer risk. Beral et al.5 reported on a 25-year follow-up of 46,000 women who were recruited in 1968 – 69 by their general practitioners throughout Britain. The multivariate RR of death from colorectal cancer was 0.57 (95% CI: 0.35– 0.94) (Table 1). The trend in risk by duration of use was not significant. The RR for death from all cancers combined in ever-users was 0.95 (95% CI: 0.82–1.09). Two reports on the association of colorectal cancer with OC have been published based on data from the Nurses’ Health Study.6,7 This cohort was assembled in 1976, when questionnaires were mailed to all female married registered nurses, ages 30 –55 years, residing in 11 large states of the United States. The analysis by Chute et al.6 was extended by Martinez et al.7 in 1997 for more than 1 million person-years. An inverse association for ever-use of OC emerged (multivariate RR 5 0.84, 95% CI: 0.69 –1.02), and it was significant (RR 5 0.60, 95% CI: 0.40 – 0.89) in women who had used OC for longer periods (Table 1). The results for colon cancer were similar to those for rectal cancer. A prospective cohort study of 35,215 women, ages 55– 69 years, in possession of a driver’s licence and without a history of cancer, was carried out in Iowa from 1986 –1990.8 Ever-use was associated with a multivariate RR of 0.96 (95% CI: 0.67–1.39) (Table 1). Among 57,529 women in the United States 31–90 years of age who volunteered for a nationwide breast cancer screening program from 1973–1980, 154 pathologically confirmed cases of colon cancer and 49 cases of rectal cancer were observed for up to 10 years of follow-up9; OC use, reflecting mainly past use, was unrelated to colorectal cancer risk (RR 5 1.0; 95% CI: 0.75–1.4).

Oral Contraceptives and Colorectal Tumors

337

Case-Control Studies A case-control study was conducted in 1976 –77 in Washington State, on 143 white women, ages 45–74 years, and 707 white women in the same age group drawn from a population survey in the same area.10 The difference in OC use between women with cancer and controls was not significant (age-adjusted RR for ,5 and $5 year-use 5 1.3, 95% CI: 0.5–3.1, and 2.0, 95% CI: 0.7–5.2, respectively). The RR were 1.0 for colon cancer and 2.6 (p 5 0.09) for rectal cancer (Table 2). Between 1979 and 1980, Potter and McMichael11 conducted a case-control study in Adelaide, Australia, on 155 women with colorectal cancer (of 212 eligible cases) and 311 control women selected from the local electoral roll. The multivariate RR of colon cancer was 0.5 (95% CI: 0.3–1.2) for ever- vs. never-users. The RR for rectal cancer was 0.7 (95% CI: 0.3–1.8). There was a trend of decreasing risk with increasing duration of use (Table 2). Ninety women with colorectal cancers and 208 controls (ie, spouses of colorectal cancer patients) were interviewed between 1980 and 1983 in Chicago (ie, 63% of initially contacted subjects).12 The crude RR in ever-users of OC was 0.6 (95% CI: 0.28 –1.34), but only nine women with cancer and 32 control women were reported to have ever used OC. A case-control study was conducted by Kune et al.13 in Melbourne, Australia, between 1980 and 1981. It included all local women with incident colorectal cancer (108 colon and 82 rectum) and 200 agematched female controls (ie, a random sample of local population). Multivariate RR among ever-users of OC were 1.17 (95% CI: 0.59 –2.29) for colon and 2.04 (95% CI: 1.00 – 4.14) for rectal cancer. For .9-year duration of OC use, however, RR were 0.68 for colon and 0.87 for rectal cancer (Table 2). Three Italian case-control investigations provided data on OC use and colorectal cancer risk. One,14 carried out in Northeastern Italy, included very few OC users (one woman with cancer and nine controls), with a crude RR of 0.2 (95% CI: 0.03–2.0) (Table 2). A larger case-control study from Milan, Northern Italy, was reported by Negri et al.15 and Fernandez et al.16 A reduced risk of colorectal cancer was observed among women who had ever used OC (multivariate RR 5 0.58, 95% CI: 0.36 – 0.92). The RR was 0.52 (95% CI: 0.27–1.02) for $2 year duration of use.16 Protection seemed to last over time. Talamini et al.17 showed results from a subsequent independent, but similar, investigation carried out in six Italian areas. Use of OC did not seem influential on colon cancer (multivariate RR 5 1.0, 95% CI: 0.69 –1.46), but inversely associated with the risk of rectal cancer (RR 5 0.56, 95% CI: 0.31–1.01). A pooled analyses of

Country

Weiss et al.10

Washington State, U.S.

Potter and Adelaide, McMichael11 Australia

Furner et al.12

Chicago, Illinois, U.S.

Kune et al.13

Melbourne, Australia

Franceschi et al.14

Northeastern Italy

Negri et al.15 Fernandez et al.16 Talamini et al.17 Fernandez et al.18

Italy

143:707 (population)

Relative risk (RR) (95% confidence interval), ever- vs. never users Colonrectum #5 yr 1.3 (0.5–3.1) .5 yr 2.0 (0.7–5.2)

155:311 (population)

90:208 (spouses)

Colon

Right colon

Left colon

Rectum

0.2 (0.03–2.0)

1536:3110 (hospital)

0.64 (0.49–0.85)

Adjustment and comments

2.6 (p 5 0.09)

No significant trend

Not shown

0.5 (0.3–1.2)

0.7 (0.3–1.8)

Inverse trend (RR for .2 yr use 5 0.20, 0.04– 1.00)

Not shown

Not shown

Not shown

No effect (RR for .9 yr use:0.68 for colon and 0.87 for rectum, not significant) Not shown

Not shown

Inverse trend for colon but not rectal cancer

Age, education, Stronger cancer family protection history, BMI, from HRT, parity, recent menopause; use (,10 calorie intake. yr: 0.44, 0.27–0.72) Prevalence of OC use among controls: 7%. RR for colon site not specified 5 0.52 (0.30–0.89).

1.17 (0.59–2.29)

89:148 (hospital)

Recent use

1.0

0.6 (0.28–1.34)

190:200 (population)

Duration of use

0.63 (0.45–0.87)

2.04 (1.00–4.14)

0.71 (0.34–1.50)

0.83 (0.54–1.28)

0.66 (0.43–1.01)

Not shown

Age. Prevalence of OC use was about 40% (25% for $5 yr use). Reproductive variables (diet was uninfluential). Prevalence of OC use among controls: 18%. Unadjusted. Prevalence of OC use among controls: 18%. Parity and age at 1st child. Prevalence of OC use among controls: 12%.

Franceschi and La Vecchia

Reference

Cases: Controls (type of controls)

338

Table 2. Case-control studies on oral contraceptives (OC) and colorectal cancer

Unadjusted. Only 1 case and 9 controls had ever used OC.

Contraception 1998;58:335–343

Contraception 1998;58:335–343

Table 2 Continued

Reference

Country

Cases: Controls (type of controls)

Relative risk (RR) (95% confidence interval), ever- vs. never users Colonrectum

Colon

Right colon

Left colon

1.02 (0.59–1.75) 1.06 (0.39–2.89)

1.42 (0.62–3.30) 1.30 (0.31–5.46)

0.70 (0.32–1.53) 0.98 (0.20–4.82)

Rectum

Duration of use

327:327 (neighbors)

,5 yr

Wu-Williams et al.20

North America and China

395:1112 (neighbors)

North America China

Jacobs et al.21 Washington, Seattle, U.S.

193:194 (population)

1.16 (0.70–1.90)

Kampman et al.22

The Netherlands

102:123 (population)

0.97 (0.46–2.03)

Not shown

Kampman et al.23

U.S.; KPMC

894:1,120 (KPMC members)

0.86 (0.67–1.10)

Not shown

$5 yr

1.2 (p 5 0.67) 0.55 (p 5 0.27)

No effect

0.4 (p 5 0.04) 0.7 (p 5 0.34)

1.23 (0.65–2.32)

BMI, body mass index; KPMC, Kaiser Permanente Medical Care; HRT, hormone replacement therapy.

1.13 (0.62–2.06)

No trend

No trend

Adjustment and comments

Not shown Cancer family history, parity, exercise, fat, alcohol, and calcium intake. Prevalence of OC use among controls: 19%. Not shown Unadjusted (but unaltered by exercise, saturated fat, and years in the U.S.). Prevalence of OC use among controls: 16% in North America and 12% in China. Not shown Age and vitamin intake. Prevalence of OC use among controls: 59%. Not shown Age, urbanization, cholecystectomy, socioeconomic level, colon cancer, family history, and dietary habits. Prevalence of OC use among controls: 48%. Not shown Age, cancer family history, aspirin, calorie intake, HRT, and exercise. Prevalence of OC use among controls: 25%.

Oral Contraceptives and Colorectal Tumors

Peters et al.19 Los Angeles, California, U.S.

Recent use

339

340

Franceschi and La Vecchia

the two studies from Milan16 and all Italy17 involved 994 colon cancers, 542 rectal cancers, and 3110 hospital controls.18 It provided a multivariate RR of 0.63 (95% CI: 0.45– 0.87) for colon cancer and 0.66 (95% CI: 0.43–1.01) for rectal cancer in ever-users of OC (Table 2). Increasing duration of OC use was related to decreasing risk of colon but not rectal cancer. The inverse association with colorectal cancer was slightly stronger for recent use (RR for ,10 years since last use 5 0.44, 95% CI: 0.27– 0.72). Peters et al.19 conducted a population-based casecontrol study in Los Angeles between 1983 and 1986. A total of 327 white women with colon cancer (of 472 eligible cases) and 327 individually matched neighborhood controls were interviewed. The multivariate RR was 1.06 (95% CI: 0.39 –2.89) for $5 years of use (Table 2). Such estimate was based, however, on very few long-term users (six women with cancer and seven controls). A population-based case-control study was conducted among Chinese women in western North America and China between 1981 and 1986, using a common protocol.20 It included 395 women with colorectal cancer (189 from North America and 206 from China) and 1112 age-matched controls (494 and 618, respectively). Use of OC was associated with a decreased risk of rectal cancer (crude RR 5 0.4, p 5 0.04 in North America, and RR 5 0.7, p 5 0.34 in China), but not of colon cancer (RR 5 1.2, p 5 0.67 and RR 5 0.55, p 5 0.27, respectively) (Table 2). Jacobs et al.21 conducted a case-control study among women ages 30 – 62 years in Seattle between 1985 and 1989. It included 193 women with incident colon cancer (of 295 eligible cases) and 194 controls (of 227 eligible controls) selected by random-digit dialing. The multivariate RR for ever-use of OC was 1.16 (95% CI: 0.70 –1.90) (Table 2). A case-control study from the Netherlands22 included 102 women with colon cancer (ie, 60% of those invited to participate) and 123 women serving as a control group (ie, 57% of those invited to participate). Control women were recruited randomly by the general practitioners of the participating cases and were frequency matched for age and residential area. The multivariate RR among ever-users of OC was 0.97 (95% CI: 0.46 –2.03). Kampman et al.23 conducted a case-control study between 1992 and 1995 in the United States among women ages 30 –79 years who were members of the Kaiser Permanente Medical Care Program. It included 894 cases of colon cancer (of 1521 eligible cases) and 1120 controls (63% of those who were contacted). No significant association was found with OC use, which was reported by 25% of control subjects (multivariate RR 5 0.86, 95% CI: 0.67–1.10) (Table 2).

Contraception 1998;58:335–343

Adenomatous Polyps In recent years, the spread of colorectal cancer screening has allowed accumulation of data on the relationship between OC use and benign colorectal neoplasms (ie, polyps and adenomas). The risk of adenomatous polyps of the distal colorectum was assessed in 26,983 participants in the Nurses’ Health Study (United States).24 During 14 years of follow-up, 982 polyps were found. The multivariate RR among ever-users of OC was 1.0 (95% CI: 0.8 –1.1), and no risk trend by duration of OC use was observed (Table 3). Jacobson et al.25 studied colorectal adenomatous polyps between 1986 and 1988 in New York. Cases (n 5 128) were cancer-free women ages 35– 84 years who had an adenoma detected at the index colonoscopy. Controls (n 5 283) were cancer-free women with a normal index colonoscopy at the same institution as cases. A higher frequency of ever-use of OC among controls (19/126) than cases of adenomatous polyps (41/196) was found (RR 5 0.6, 95% CI: 0.3–1.1). Potter et al.26 undertook a case-control study in Minnesota between 1991 and 1994. Cases were 219 women, ages 30 –74 years, with colonoscopy proven, pathology confirmed adenomatous polyps of colon and rectum. Two control groups were selected: 439 women without polyps at colonoscopy and 247 ageand zip code-matched community controls (response rates 5 65% among the three groups). Similar proportions of OC ever-users were found in cases and controls (RR for $5-year use vs. colonoscopy-negative controls 5 0.84, 95% CI: 0.51–1.39; RR vs. community controls 5 1.05, 95% CI: 0.59 –1.84). Peipins et al.27 carried out a study in North Carolina between 1988 and 1991 on 115 women with colorectal adenomatous polyps and 232 control women (ie, women who had undergone colonoscopy, but were found to be free from colorectal adenomas). All study subjects were older than 30 years. Ever-use of OC was reported by 25% of women with cancer and 32% of controls (RR 5 0.91, 95% CI: 0.45–1.82). The RR for $5 year-duration of use was 1.02 (95% CI: 0.40 –2.58)

Discussion The present review shows that the association of OC use with colorectal cancer risk has been evaluated in several studies, which provide no evidence of an unfavorable effect or, in a few datasets, suggestions of a favorable one. Reduced risk in ever-users was found in three of four cohort studies on the topic, and was significant in the one based on colorectal cancer mortality. Eleven case-control studies (or groups of studies) also

Contraception 1998;58:335–343

Table 3. Studies on oral contraceptives (OC) and colorectal adenomatous polyps

Reference

Country

Population (follow-up), no. with polyps or cases:controls (type of controls)

Cohort studies Platz et al.24 U.S.; Nurses 26,983 Health (14 years) Study 982 Case-control studies Jacobson et al.25

Relative risk (RR) (95% confidence interval), ever- vs. never-users Colonrectum 1.0 (0.8–1.1)

128:283 (colonoscopy negative)

0.6 (0.3–1.1)

Potter et al.26

Minnesota, U.S.

219 vs. 439 (colonoscopy negative) and 247 (population)

Peipins et al.27

North Carolina, U.S.

115:232 (colonoscopy negative)

,5 yr 1.25 (0.82–1.91) $5 yr 0.84 (0.51–1.39) ,5 yr 1.42 (0.88–2.31) $5 yr 1.05 (0.59–1.84) 0.91 (0.45–1.82)

Left colon 1.0 (0.8–1.1)

Rectum

Duration of use

Recent use

Adjustment and comments

No trend (RR No for $6 yr difference use 5 1.0, 0.7–1.4)

Age, history of endoscopy, parity, body mass index, physical activity, and dietary habits.

Not shown

Not shown

No trend

Not shown

Age, education, smoking, and body mass index. Prevalence of OC use among controls: 26%. Adjustment not reported in detail. Prevalence of OC use: 49% among community controls.

No trend (RR Not shown for $5 yr use 5 1.02, 0.40–2.58)

Age, education, body mass index, and dietary habits. Prevalence of OC use among controls: 30%.

Oral Contraceptives and Colorectal Tumors

U.S. 3 states

Right Colon colon

341

342

Franceschi and La Vecchia

contributed to the issue. None showed significantly elevated risk. Five of them reported lowered colorectal cancer risk among ever-users of OC, with a significant inverse association in the largest investigation available. This was a pooled analysis of two Italian studies18 where allowance for a broad range of potential confounding factors, including energy intake, cancer family history, and menstrual and reproductive factors, was possibile. Most studies on colorectal cancer, however, dealt with relatively small proportions of OC users (lower, for instance, than those of hormone replacement therapy users),28 due to the relatively advanced age of the study subjects. Ever-use of OC is a crude measure of exposure and an evaluation of timing of exposure is essential. Longer duration of use was associated with the lowest risk in the Nurses’ Health Study7 and in one casecontrol investigation.11 Several studies, particularly the earliest ones, included too few cases of colorectal cancer or OC users to evaluate risk by duration, but the lack of a trend in the largest case-control study available18 is of notice. Time since last use could be assessed in even fewer studies.5,18 An attenuation of the apparent protection 5–15 years after cessation of use was suggested. Data on colorectal adenomatous polyps and OC use started accumulating later and are scantier than those on colorectal cancer. The only cohort investigation7,24 that showed data for benign as well as malignant neoplasms of the colon-rectum suggested an inverse association only for cancer. It is conceivable that agents acting on late-stage carcinogenesis may not influence the risk of early tumors such as colorectal adenoma,24 but more data are needed to help interpretation. Long-standing hypotheses about the biologic mechanisms of estrogens in colorectal cancer have focused on variations in bile acid metabolism in response to female hormones.29 Secondary bile acids are believed to initiate or promote malignant changes in the colonic epithelium, and exogenous estrogens (which decrease secondary bile acid production) could therefore be protective, particularly at the right side of the colon. No consistent differences between the right and left colon and even between colon and rectum, however, emerged in available studies. A biologic heterogeneity of colorectal tumors (eg, selective inverse association of OC use with p53-negative colorectal tumors)30 has also been suggested, but needs confirmation. More recent hypotheses have focused on steroid hormone receptors in normal as well as colorectal tumor cells (eg, methylation-associated loss of expression after estrogen withdrawal31). Negative confounding is also possible if OC users tend to have a healthier lifestyle (ie, higher intake of

Contraception 1998;58:335–343

vegetables, more exercise, etc.), as it has been reported for hormone replacement therapy (prevention bias).32 However, the two largest studies7,18 that showed inverse associations with colorectal cancer had the possibility to stratify and adjust results for a very large spectrum of risk factors. Surveillance bias (ie, more frequent pelvic examination and colorectal cancer screening in women who had used OC) may lead to a lower risk of fatal colorectal cancer, as seen in Beral et al.5 It would, however, increase artifactually RR estimates in studies of adenomatous polyps and early cancer. Information bias (ie, more accurate report of OC use by control women than by those with colorectal cancer) seems an unlikely explanation for the present findings. Selection bias is always difficult to rule out completely, especially because of relatively low participation rates in many of the studies assessed here. Some type of publication bias may have facilitated the recent disclosure and publication of inverse associations of OC use with colorectal neoplasms, particularly as some evidence of a favorable effect had emerged for hormone replacement therapy.28,33,34 In the near future, exogenous female hormones will be carefully evaluated, or reevaluated, in studies available on colorectal neoplasms, thus allowing the computation of a reliable overall (combined) RR. Were a favorable effect of OC use demonstrated, it would weigh greatly on any risk/benefit assessment for OC.

Acknowledgment This work was conducted with the contribution of the Italian Association for Research on Cancer. We thank Mrs. A. Redivo for editorial assistance.

References 1. McMichael AJ, Giles GG. Colorectal cancer. Cancer Surv 1994;19:77–98. 2. La Vecchia C, Negri E, Levi F, et al. Cancer mortality in Europe: effects of age, cohort of birth and period of death. Eur J Cancer 1998;34:118 – 41. 3. Risch HA, Howe GR. Menopausal hormone use and colorectal cancer in Saskatchewan: a record linkage cohort study. Cancer Epidemiol Biomarkers Prev 1995; 4:21– 8. 4. Gerhardsson de Verdier M, London S. Reproductive factors, exogenous female hormones, and colorectal cancer by subsite. Cancer Causes Control 1992;3:355– 60. 5. Beral V, Hermon C, Clifford K, et al. Mortality in relation to oral contraceptive use: 25–year follow-up of women in the Royal College of General Practitioners’ Oral Contraception Study. Br Med J (in press). 6. Chute CG, Willett WC, Colditz GA, et al. A prospective study of reproductive history and exogenous estrogens on the risk of colorectal cancer in women. Epidemiology 1991;2:201–7.

Contraception 1998;58:335–343

7. Martinez ME, Grodstein F, Giovannucci E, et al. A prospective study of reproductive factors, oral contraceptive use, and risk of colorectal cancer. Cancer Epidemiol Biomarkers Prev 1997;6:1–5. 8. Bostick RM, Potter JD, Kushi LH, et al. Sugar, meat, and fat intake, and non-dietary risk factors for colon cancer incidence in Iowa women (United States). Cancer Causes Control 1994;5:38 –52. 9. Troisi R, Schairer C, Chow W-H, et al. Reproductive factors, oral contraceptive use, and risk of colorectal cancer. Epidemiology 199;8:75–9. 10. Weiss NS, Daling JR, Chow WH. Incidence of cancer of the large bowel in women in relation to reproductive and hormonal factors. J Natl Cancer Inst 1981;67:57– 60. 11. Potter JD, McMichael AJ. Large bowel cancer in women in relation to reproductive and hormonal factors: a case-control study. J Natl Cancer Inst 1983;71:703–9. 12. Furner SE, Davis FG, Nelson RL, Haenszel W. A casecontrol study of large bowel cancer and hormone exposure in women. Cancer Res 1989;49:4936 – 40. 13. Kune GA, Kune S, Watson LF. Oral contraceptive use does not protect against large bowel cancer. Contraception 1990;41:19 –25. 14. Franceschi S, Bidoli E, Talamini R, et al. Colorectal cancer in northeastern Italy: reproductive, menstrual and female hormone-related factors. Eur J Cancer 199; 27:604 – 8. 15. Negri E, La Vecchia C, Parazzini F, et al. Reproductive and menstrual factors and risk of colorectal cancer. Cancer Res 1989;49:7158 – 61. 16. Fernandez E, La Vecchia C, D’Avanzo B, et al. Oral contraceptives, hormone replacement therapy and the risk of colorectal cancer. Br J Cancer 1996;73:1431–5. 17. Talamini R, Franceschi S, Dal Maso L, et al. The influence of reproductive and hormonal factors on the risk of colon and rectal cancer in women. Eur J Cancer 1998;34:1070 – 6. 18. Fernandez E, La Vecchia C, Franceschi S, et al. Oral contraceptive use and risk of colorectal cancer. Epidemiology 1998;9:295–300. 19. Peters RK, Pike MC, Chang WWL, Mack TM. Reproductive factors and colon cancers. Br J Cancer 1990;61: 741– 8. 20. Wu-Williams AH, Lee M, Whittemore AS, et al. Reproductive factors and colorectal cancer risk among Chinese females. Cancer Res 1991;51:2307–11. 21. Jacobs EJ, White E, Weiss NS. Exogenous hormones,

Oral Contraceptives and Colorectal Tumors

22. 23.

24.

25.

26.

27.

28. 29. 30.

31. 32. 33. 34.

343

reproductive history, and colon cancer (Seattle, Washington, USA). Cancer Causes Control 1994;5:359 – 66. Kampman E, Bijl AJ, Kok C, van’t Veer P. Reproductive and hormonal factors in male and female colon cancer. Eur J Cancer Prev 1994;3:329 –36. Kampman E, Potter JD, Slattery ML, et al. Hormone replacement therapy, reproductive history, and colon cancer: a multicenter, case-control study in the United States. Cancer Causes Control 1997;8:146 –58. Platz EA, Martinez ME, Grodstein F, et al. Parity and other reproductive factors and risk of adenomatous polyps of the distal colorectum (United States). Cancer Causes Control 1997;8:849 –903. Jacobson JS, Neugut AI, Garbowski GD, et al. Reproductive risk factors for colorectal adenomatous polyps (New York City, NY, United States). Cancer Causes Control 1995;6:513– 8. Potter JD, Bostick RM, Grandits GA, et al. Hormone replacement therapy is associated with lower risk of adenomatous polyps of the large bowel. The Minnesota Cancer Prevention Research Unit Case-Control Study. Cancer Epidemiol Biomarkers Prev 1996;5:779 – 84. Peipins LA, Newman B, Sandler RS. Reproductive history, use of exogenous hormones, and risk of colorectal adenomas. Cancer Epidemiol Biomarkers Prev 1997;6: 671–5. Grodstein F, Martinez E, Platz EA, et al. Postmenopausal hormone use and risk of colorectal cancer and adenoma. Ann Intern Med 1998;128:705–12. McMichael AJ, Potter JD. Parity and death from colon cancer in women: a case-control study. Community Health Stud 1984;8:19 –25. Freedman AN, Michalek AM, Troisi R, et al. Oral contraceptives, reproductive factors and p53 gene expression in colorectal cancer. Carcinogenesis 1997;8: 855– 6. Issa JP, Ottaviano YL, Celano P, et al. Methylation of the estrogen receptor CpG island links aging and neoplasia in human colon. Nature Genet 1994;7:536 – 40. Barrett-Connor E. Postmenopausal estrogen and prevention bias. Ann Intern Med 1991;6:455– 6. La Vecchia C, Franceschi S. Reproductive factors and colorectal cancer. Cancer Causes Control 1991;2: 193.200. Franceschi S, La Vecchia C. Colorectal cancer and hormone replacement therapy: an unexpected finding. Eur J Cancer Prev 1998;7:427–38.