Hormone replacement therapy and cancer risk: A systematic analysis from a network of case-control studies

Authors

  • Esteve Fernandez,

    Corresponding author
    1. Cancer Prevention and Control Unit, Institut Català d'Oncologia, L'Hospitalet (Barcelona), Catalonia, Spain
    2. Istituto di Ricerche Farmacologiche “Mario Negri,” Milan, Italy
    3. Department of Public Health, University of Barcelona, Catalonia, Spain
    • Cancer Prevention and Control Unit, Institut Català d'Oncologia, Av. Gran Via s/n Km 2.7, 08907 L'Hospitalet (Barcelona), Spain
    Search for more papers by this author
    • Fax: +34-93-2607956

  • Silvano Gallus,

    1. Istituto di Ricerche Farmacologiche “Mario Negri,” Milan, Italy
    2. Servizio di Epidemiologia, Centro di referimento Oncologico, Aviano, Italy
    Search for more papers by this author
  • Cristina Bosetti,

    1. Istituto di Ricerche Farmacologiche “Mario Negri,” Milan, Italy
    Search for more papers by this author
  • Silvia Franceschi,

    1. Field and Intervention Studies Unit, International Cancer for Research on Cancer, Lyon, France
    Search for more papers by this author
  • Eva Negri,

    1. Istituto di Ricerche Farmacologiche “Mario Negri,” Milan, Italy
    Search for more papers by this author
  • Carlo La Vecchia

    1. Istituto di Ricerche Farmacologiche “Mario Negri,” Milan, Italy
    2. Istituto di Statistica Medica e Biometria, Università di Milano, Milan, Italy
    Search for more papers by this author

Abstract

To provide comprehensive and quantitative information on the benefits and risks of hormone replacement therapy (HRT) on several cancer sites, we systematically examined the relation between HRT use and the risk of various cancers in women aged 45–79 by using data from a framework of case-control studies conducted in Italy between 1983 and 1999. The overall data set included the following incident, histologically confirmed neoplasms: oral cavity, pharynx, larynx and esophagus (n = 253), stomach (n = 258), colon (n = 886), rectum (n = 488), liver (n = 105), gallbladder (n = 31), pancreas (n = 122), breast (n = 4,713), endometrium (n = 704), ovary (n = 1,614), urinary bladder (n = 106), kidney (n = 102), thyroid (n = 65), Hodgkin's disease (n = 26), non-Hodgkin's lymphomas (n = 145), multiple myeloma (n = 65) and sarcomas (n = 78). The control group comprised 6,976 women aged 45–79 years, admitted for a wide spectrum of acute, nonneoplastic conditions. Odds ratios (OR) and the corresponding 95% confidence intervals (CI) for use of HRT were derived from multiple logistic regression equations. There was an inverse association between ever use of HRT and colon (OR = 0.7), rectum (OR = 0.5) and liver cancer (OR = 0.2), with a consistent pattern of protection for duration of use. An excess risk was found for gallbladder (OR = 3.2), breast (OR = 1.1), endometrial (OR = 3.0) and urinary bladder cancer (OR = 2.0). These data from a southern European population add some useful information on the risk-benefit assessment of HRT among postmenopausal women. © 2003 Wiley-Liss, Inc.

Hormone replacement therapy (HRT) is highly effective to relieve short-term menopause-related symptoms and prevent postmenopausal osteoporosis.1 The use of estrogens increases the risk of endometrial cancer, breast cancer, and venous thromboembolism and that of ovarian and gallbladder cancer,2, 3, 4, 5, 6, 7, 8 whereas a protective effect on colorectal cancer risk has been reported.9, 10 With respect to other cancer sites, there are some indications that HRT is associated with a reduced risk of cervical cancer and stomach cancer and an increased risk of renal cell cancer, whereas there is no consistent effect on the risk of liver, malignant melanoma, or thyroid cancer.6, 11 Notwithstanding the widespread use of HRT, relatively few studies have focused on its effect on cancers other than the breast, endometrium, ovary, and colorectum.6, 11

To provide comprehensive and quantitative information on the benefits and risks of HRT on several cancer sites, we systematically examined the relation between HRT use and the risk of various cancers by using data from an integrated series of hospital-based case-control studies conducted in Italy.

MATERIAL AND METHODS

The data were derived from a network of case-control studies conducted between 1983 and 1999, whose general design has been already described.12 Trained interviewers identified and questioned patients admitted to teaching and general hospitals in the area under surveillance for selected cancers and for a wide spectrum of other acute, nonneoplastic conditions. All interviews were conducted during the hospital stay. On average, less than 3% of eligible cases and controls refused to be interviewed. The same study design, criteria of enrollment of cases and controls, and interview setting were adopted for all the diseases studied, and all the questionnaires contained a basic structured section, including sociodemographic characteristics, anthropometric measures and environmental exposures (including smoking and alcohol consumption), and frequency of consumption of selected dietary items. For women, information was also collected on menstrual and reproductive factors, and on use of oral contraceptives, menopausal HRT, and female hormones for other indications. The time of each episode of use was registered, together with the brand name, whenever available.

The cases were women between 45 and 79 years who were admitted to the National Cancer Institute and the Ospedale Maggiore, which includes the 4 largest teaching and general hopitals of the Greater Milan area (northern Italy) and the Oncologic Reference Hospital of the province of Pordenone (northeastern Italy) with histologically confirmed incident (i.e., diagnosed within the year prior to interview) cancers of the oral cavity, pharynx, larynx and esophagus (n = 253), stomach (n = 258), colon (n = 886), rectum (n = 488), liver (n = 105), gallbladder (n = 31), pancreas (n = 122), breast (n = 4,713), endometrium (n = 704), ovary (n = 1,614), urinary bladder (n = 106), kidney (n = 102), thyroid (n = 65), Hodgkin's disease (n = 26), non-Hodgkin's lymphomas (n = 145), multiple myeloma (n = 65) and sarcomas (n = 78).

The control group comprised 6,976 women aged 45 to 79 years admitted for a wide spectrum of acute conditions in the same hospitals where cases were identified. For admission diagnoses, specific exclusions were made for malignant tumors, digestive tract diseases, or any disorder related to alcohol or tobacco consumption, or which might have induced a long-term modification of diet. Overall, controls were admitted for traumatic conditions (21%), nontraumatic orthopedic disorders (25%), acute surgical conditions (35%), and other miscellaneous diseases (19%), such as ear, nose, and throat, skin or dental disorders. For endometrial and ovary cancers, controls with hysterectomy or bilateral ooforectomy were excluded.

Odds ratios (OR) and corresponding 95% confidence intervals (CI) for HRT use were derived from unconditional multiple logistic regression equations, fitted by the method of maximum likelihood.13 The potential confounders included in the regression equations were age, center, year of interview, education, body mass index, status/type of menopause, age at menopause, smoking status, alcohol consumption and, for colon, rectum, liver, breast, endometrium and ovary, ever use of oral contraceptives. For breast, endometrium, ovary, gallbladder and colorectal cancer, further adjustment for other reproductive factors (i.e., parity) was done but no meaningful differences emerged.

RESULTS

The distribution of cases according to ever use and duration of HRT use is shown in Table I. The frequency of exposure among cases ranged from 2.8% (among liver cancer) to 16.1% (among gallbladder cancer), and the overall frequency of ever use among controls was 7.6%.

Table I. Distribution of Cases of Selected Cancers and of Controls According to Different Measures of Hormone Replacement Therapy Use, Italy, 1983–99
 Never useEver useDuration of use1
< 2 years≥ 2 years
  • 1

    In some strata, sum does not add up to the total because of some missing values.

Oral cavity, esophagus, pharynx and larynx2371686
Stomach2451385
Colon832543320
Rectum46622157
Liver102312
Gallbladder26522
Pancreas1101292
Breast4,354359241115
Endometrium617876126
Ovary1,4981166946
Bladder8917116
Kidney891393
Thyroid61431
Hodgkin's disease23321
Non-Hodgkin's lymphomas13213112
Multiple myeloma62302
Sarcomas681073
Controls6,439537344180

Figure 1 and Table II give the OR and 95% CI of various neoplasms for ever vs. never use of HRT. There was an inverse relation between ever use of HRT and colon (OR = 0.7; 95% CI = 0.5–0.9), rectal (OR = 0.5; 95% CI = 0.3–0.8) and liver cancer (OR = 0.2; 95% CI = 0.1–0.8). A nonsignificant reduction in risk was observed for cancers of the oral cavity, esophagus, pharynx and larynx, as well as for thyroid cancer, Hodgkins's, non-Hodgkin's lymphomas and myelomas. A nonsignificant increased risk of pancreatic cancer, kidney cancer and sarcomas for women who had ever used HRT was present. An excess risk was found for gallbladder (OR = 3.2; 95% CI = 1.1–9.3), breast (OR = 1.1; 95% CI = 1.0–1.3), endometrial (OR = 3.0; 95% CI = 2.2–4.0) and urinary bladder cancer (OR = 2.0; 95% CI = 1.1–3.5).

Figure 1.

Odds ratio and 95% confidence intervals of selected cancers according to hormone replacement therapy use. Estimates from multiple logistic regression equations including terms for age, study center, year of interview, education, smoking, drinking, type of menopause, age at menopause and body mass index. The reference category is never use of hormone replacement therapy.

Table II. Odds Ratios1 and 95% Confidence Intervals of Selected Cancers According to Different Measures of Hormone Replacement Therapy Use, Italy, 1983–99
 Ever use OR (95% CI)Duration of useChi-square trend
< 2 years OR (95% CI)≥ 2 years OR (95% CI)
  • 1

    Estimates from multiple logistic regression equations including terms for age, study center, year of interview, education, smoking, drinking, type of menopause, age at menopause and body mass index. The reference category is never use of hormone replacement therapy.

  • 2

    p < 0.05.

  • 3

    Estimates further adjusted for oral contraceptive use.

Oral cavity, esophagus, pharynx and larynx0.7 (0.4–1.2)0.5 (0.2–1.1)0.8 (0.4–1.9)1.76
Stomach0.7 (0.4–1.3)0.7 (0.4–1.5)0.8 (0.3–1.9)0.82
Colon30.7 (0.5–0.9)0.7 (0.5–1.0)0.7 (0.4–1.1)6.062
Rectum30.5 (0.3–0.8)0.5 (0.3–0.9)0.5 (0.2–1.0)8.672
Liver30.2 (0.1–0.8)0.2 (0.0–1.1)0.3 (0.1–1.7)4.452
Gallbladder3.2 (1.1–9.3)2.3 (0.5–10.5)3.0 (0.6–15.0)2.47
Pancreas1.6 (0.9–3.1)2.0 (1.0–4.1)0.7 (0.2–3.1)0.43
Breast31.1 (1.0–1.3)1.2 (1.0–1.4)1.1 (0.9–1.4)2.60
Endometrium33.0 (2.2–4.0)3.3 (2.3–4.6)2.5 (1.4–4.3)38.292
Ovary31.0 (0.8–1.3)0.9 (0.7–1.2)1.2 (0.8–1.7)0.10
Bladder2.0 (1.1–3.5)1.9 (1.0–3.8)2.2 (0.9–5.6)5.702
Kidney1.3 (0.7–2.4)1.3 (0.6–2.8)0.9 (0.3–3.2)0.06
Thyroid0.8 (0.3–2.3)1.0 (0.3–3.4)0.5 (0.1–4.0)0.26
Hodgkin's disease0.8 (0.2–3.1)0.8 (0.2–3.8)1.2 (0.1–10.7)0.01
Non-Hodgkin's lymphomas0.7 (0.3–1.3)0.9 (0.4–1.8)0.2 (0.0–1.6)2.18
Multiple myeloma0.5 (0.1–1.6) 
Sarcomas1.1 (0.5–2.7)1.8 (0.7–4.6)0.5 (0.1–2.1)0.06

With reference to duration of use (Table II), a pattern of greater protection for longer duration was present for cancers of the liver, colon, and rectum (p for trend < 0.05). A pattern of increasing risk with duration was also present for endometrium and bladder cancer, whereas a nonsignificant trend was present for cancer of the gallbladder, breast and ovary. The pattern of cancer risk was consistent across age strata and no appreciable effect modification was found.

DISCUSSION

In this comprehensive analysis of HRT in relation to several cancer sites, an inverse relation between ever use of HRT and risk of colon, rectum and liver cancers was observed, whereas increased risks for gallbladder, breast, endometrium and bladder cancers were present.

There is substantial evidence supporting that HRT use is inversely associated with colorectal cancer. A significant 20% to 30% reduction of risk for ever use was consistent across the different studies, with a greater protection for longer duration of use.14, 15 A healthy user effect or prevention bias has been suggested,16 since HRT users may differ from nonusers in ways that influence colorectal cancer risk. However, the results from randomized clinical trials showed a decreased risk of colorectal cancer after several years of follow-up, supporting that the inverse association was real.4 Different biological mechanisms have been postulated. Female hormones confer a protection against colorectal cancer, possibly as a result of changes in bile synthesis, which lead to reduced concentration of bile acids in the colon.17 Estrogens inhibit the growth of colon cancer cells in vitro,18 and estrogen receptors have been identified in normal and neoplastic colon epithelial cells.19 The estrogen receptor (ER) gene might play a tumor suppressor role, since the hypermethylation of the promotor region of the ER gene results in a reduced expression and deregulated growth in colonic mucosa.20 Estrogens may also reduce serum insulin-like growth factor-I (IGF-I) levels,21, 22, 23, 24 a mitogen that may play an important role in the pathophysiology of colorectal and other cancers.25, 26, 27

There are scarce epidemiological data available on the association between HRT and liver cancer.6, 11, 28 In a Swedish cohort study based on prescription information, the relative risk of primary liver cancer was 0.7.29 No association between conjugated estrogen and other estrogen use and hepatocellular carcinoma was observed in a case-control study in Los Angeles County (RR = 1.1 for ever use).30 The use of female hormone preparations was not associated with the risk of liver cancer (RR = 1.29) in the Japanese cohort of atomic bomb survivors.31 In a previous analysis of this Italian case-control study, which included 82 liver cancer cases (with 1 exposed case to HRT),32 a nonsignificant decreased risk was observed (OR = 0.2), while in the present updated report, the OR and the pattern in risk are significant. These results, however, are based on small numbers. Estrogen receptors may be involved in the inhibition of malignant transformation of preneoplastic liver cells, and experimental studies have shown that exogenous estrogens can suppress chemical hepatocarcinogenesis.33 Moreover, the liver is the major source of IGF-I, and deregulation of the IGF axis has been identified in the development of hepatocellular carcinoma.34 Thus, downregulation of IGF-I by HRT may exert a protective effect against hepatocellular carcinoma, as also hypothesized for colorectal cancer.

Analytical studies to date have provided scanty information on the relation between HRT and the risk of gallbladder cancer.35. In a case-control study conducted in Los Angeles County, including 105 histologically confirmed gallbladder cases, no association was found,36 whereas the pooled analysis from 5 case-control studies included in the SEARCH program37 showed a nonsignificant reduction in risk based in 10 exposed cases and 126 exposed controls (OR = 0.5). Gallstones are the main risk factor for gallbladder cancer,35, 37 and gallstones are associated with HRT.38, 39, 40 Thus, the pathogenic link between HRT and gallbladder is colelithiasis. Some kind of surveillance bias is possible since HRT stimulates gallbladder motility, making gallstones symptomatic, which in turn may determine an early diagnosis of gallbladder cancer in subjects with clinical cholelithiasis.8

The relationship between HRT and endometrial cancer has been firmly established in experimental and epidemiological studies since the 1970s.11 The RR of endometrial cancer is 2 to 3 times greater in ever than in never users, is directly related to the duration of use, and is inversely related with time since last use.2, 6, 11, 41, 42, 43, 44 The present findings, therefore, confirm the available evidence.

Ever use of HRT has also been associated to the risk of breast cancer in women.28, 45 A pooled analysis of individual data of 51 epidemiological studies from 21 different countries showed a significant 35% excess risk of breast cancer for ever HRT use.46 The Women's Health Initiative Randomized Controlled Trial has recently confirmed that women with a uterus at randomization who were using combined estrogen plus progestin had a significant 26% increased incidence rate of developing breast cancer.2

With reference to HRT and bladder cancer risk, the available literature is scanty.6, 11 Although risk factors for bladder cancer have been investigated,47, 48 only the above-mentioned Swedish cohort29 has evaluated the role of HRT on bladder cancer risk. In that study, a total of 58 cases of bladder cancer were observed vs. 67.7 expected, corresponding to a nonsignificant RR of 0.9. The role of sexual hormones in bladder cancer risk is controversial. Whereas a study from Iowa49 found a protective association with menstrual and reproductive factors, results from this Italian case-control study showed no consistent association with parity and late age at birth.50 Detection bias is possible, since HRT users are referred to the medical system more frequently for menopausal bleeding,28 but the risk was not higher in recent users.

An excess risk of ovarian cancer has been established among ever HRT users.7, 51, 52 In our study, a nonsignificant 20% excess risk was found among women with 2 or more years of HRT use. The biological mechanisms that support this epidemiological observation are related to the rise in estradiol and decrease in gonadotropins levels.53 Moreover, estrogen receptors are present in human ovarian cells, and estrogens stimulate the proliferation of normal ovarian epithelial cells and the malignant differentiation of ovarian cellular lines.54

The interest and importance of this comprehensive analysis are given by the possibility of assessing the pattern of risk for various cancer sites. The overall pattern is unlikely to be due only to random or systematic error since the pattern was consistent across cancer sites that share common etiologic characteristics. Among the strengths of this study, there is the comparable catchment area of cases and controls, and the almost complete participation, which are reassuring against selection bias. Thus, the control series represents the person-time distribution of exposure in the source population. For each women contributing time to the source population, the time that she is eligible to be selected as a control is the same time during which she is eligible to become a case if the disease should occur. Selection bias due to overrepresentation of women with traumatic conditions in the control group (who could be more frequently HRT nonusers than other types of controls) is unlikely, since the analysis excluding the controls with trauma or other orthopedic conditions did not meaningfully alter the results. Cases and controls were interviewed in the same setting, which enabled us to obtain comparable information. Due to the low prevalence of HRT in this population, and the low incidence of some of the cancer sites studied, the statistical power is however limited. The study power and the information collected were also inadequate to address the issue of different types of administration and of estrogen-only vs. progestin-combined therapy.

Information bias due to a systematic underreporting of HRT is unlikely, since the potential association between HRT and cancer risk was unknown to the interviewers as well as to the patients. Differential reporting of HRT according to the period of study could also have biased the association observed. However, a similar protection was present when separate analysis according to study period was performed and the results were adjusted for the year of interview. Furthermore, although the frequency of use of HRT in Italy has increased over the last decade, the correlates of use have not changed appreciably.55 Finally, the choice of hospital controls has advantages in relation to the reliability and validity of drug use information, since cases and controls should be similarly sensitized toward various aspects of their past medical history, and reproducibility of the information was satisfactory.56, 57, 58 With reference to confounding, the regression equations included major potential confounders, including use of oral contraceptives for cancers of the colon, rectum, liver, gallbladder, breast, endometrium and ovary.

In summary, this study shows a consistent protective effect of ever use of HRT against colorectal and liver cancers and an increased risk for gallbladder, breast, endometrium and bladder cancers. In addition to the results already published from observational studies and U.S. clinical trials2, 40, 42 this data from a southern European population could be of help in the risk-benefit assessment of HRT among postmenopausal women.

Acknowledgements

The authors thank Dr. Carmen Rodríguez for useful comments to a preliminary version of this article.

Ancillary