Main

Following water and tea, coffee is the third most consumed beverage in the world (Bushman, 1998; La Vecchia and Tavani, 2007). A recent meta-analysis reported an inverse association of coffee intake with endometrial cancer (EC) risk (Je and Giovannucci, 2011). The presence of antioxidants and other chemopreventive compounds in coffee may explain its anticarcinogenic effect (Vivani, 1993; Cavin et al, 2002). However, it is not clear whether coffee per se, caffeine or other methylxanthines (e.g., theophylline and theobromine) are most relevant. Also unexplored is whether there is heterogeneity by Type I vs Type II EC, which may have different aetiologies (Bokhman, 1983; Doll et al, 2008; Mendivil et al, 2009). The aim of the present study was to evaluate the association of coffee consumption (with and without caffeine) and other sources of methylxanthines with risk of Type I vs Type II EC, overall and stratified on body mass index (BMI), smoking history and hormone therapy (HT) use.

Materials and methods

Details regarding the Iowa Women’s Health Study (IWHS) have been published (Folsom et al, 1990). In brief, 41 836 women aged 55–69 years completed a self-administered survey at enrolment in 1986. The baseline survey included a 126-item semiquantitative food-frequency questionnaire (FFQ) (Willett et al, 1988), which included the average intake in the past year of the following items: caffeinated coffee; decaffeinated coffee; tea (excluding herbal teas); regular and sugar-free carbonated beverages with caffeine; chocolate; chocolate bars; and brownies. The FFQ was reliable and valid in this population (Munger et al, 1992).

Incident EC cases were identified through 2005 via annual linkage with the Iowa Cancer Registry. Cancer data were coded according to the International Classification of Diseases for Oncology (Fritz et al, 2000). Type I or Type II were classified based on registry codes (see Table 1 footnote) as described previously (Uccella et al, 2011); there was no central pathology review. Deaths were ascertained by follow-up surveys, annual linkage with Iowa death certificates and linkage to the National Death Index.

Table 1 Association of coffee and sources of caffeine and methylxanthines with risk of Type I and Type II endometrial cancera, Iowa Women’s Health Study, 1986–2005
Full size table

Women with history of cancer before baseline, except non-melanoma skin cancer (n=3830); hysterectomy before baseline (n=14 350); extreme dietary intake (<600 or >5000 kcal per day) or incomplete FFQ questionnaires (⩾30 blank food items) (n=3096); or who were not postmenopausal at baseline (n=569) were excluded from the present analysis (not mutually exclusive), yielding a final sample size of 23 356 study participants.

Each woman accumulated person-years of follow-up from baseline to date of EC diagnosis, move from Iowa, death or administrative censoring on 31 December 2005. Relative risks (RR) and 95% confidence intervals (95% CIs) were estimated using Cox proportional hazards regression, and modelling age was used as the time variable (Korn et al, 1997). All Cox model attributes included as covariates are listed in corresponding table footnotes, and were selected a priori based on their suspected or known associations with endometrial cancer. Separate analyses were carried out for Type I and Type II EC. Tests for trend were carried out by ordering the intake quartiles from lowest to highest and including the resulting variable as a 1 d.f. linear term in the Cox regression models.

We formally determined if risk ratios for the exposure variables differed by type of EC using a competing risk form of Cox proportional hazards regression (Lunn and McNeil, 1995). We also examined associations between exposure variables and subtype-specific EC risk within strata defined by BMI, smoking status and use of HT. All statistical tests were two-sided, and analyses were carried out using SAS (SAS Institute Inc., Cary, NC, USA) and R software systems.

Results

At study baseline, there were 23 356 women in the at-risk cohort, of whom 5218 (22.3%) were obese (BMI ⩾30 kg m−2) and 6843 (29.3%) drank 4+ cups per day of coffee (caffeinated or decaffeinated). The correlation of coffee intake with EC risk factors is shown in Table 2.

Table 2 Correlation of coffee intake with selected endometrial cancer risk factors, Iowa Women’s Health Study (1986)
Full size table

During the 20-year follow-up period, we identified a total of 542 incident cases of EC, 471 Type I and 71 Type II. The mean age at diagnosis of Type I EC was 71.8 years (range, 57.2–89.5 years) and Type II EC was 72.8 years (range, 60.2–89.3 years).

There was an inverse association of caffeinated coffee consumption with risk of Type I EC after multivariate adjustment (RR=0.65 for 4+ cups per day compared with ⩽1 cup per month; P-trend=0.033), but there were no statistically significant trends with intake of total coffee, decaffeinated coffee, tea, colas or other sources of methylxanthines, although the highest intake of total coffee and decaffeinated coffee did have RRs <0.8 (Table 1). Compared with women who did not drink either caffeinated or decaffeinated coffee, those who drank 4+ cups per day of caffeinated coffee only (RR=0.73; 95% CI: 0.52–1.02) or 1+ cups per day of both types of coffee (RR=0.69; 95% CI: 0.47–1.01) had lower EC risk, whereas the association was weaker and not statistically significant for women who drank 4+ cups per day of decaffeinated coffee only (RR=0.81; 95% CI: 0.52–1.27). Caffeine intake showed a suggestive inverse associated with risk (RR=0.80 for >385 mg per day compared with <29.7 mg per day; P-trend=0.059). In contrast, coffee and other sources of methylxanthines were not associated with risk of Type II EC.

We next examined coffee intake with risk of Type I EC within strata defined by BMI (30+ vs <30 kg m−2), smoking history (ever/never) and HT use (ever/never); the sample size was too small to conduct these analyses for risk of Type II EC. As shown in Table 3, the inverse associations for total and caffeinated coffee, caffeine and perhaps decaffeinated coffee were only observed among obese women and not among women with a BMI <30 kg m−2. There was no striking or consistent heterogeneity in the associations for coffee or caffeine intake when stratified on smoking status (Supplementary Table 1) or HT use (Supplementary Table 2).

Table 3 Association of coffee and caffeine with risk of Type I endometrial cancer, stratified by BMI, Iowa Women’s Health Study, 1986–2005
Full size table

Discussion

Coffee consumption was most strongly associated with a lower risk of Type I EC among obese postmenopausal women, and these associations were generally stronger and statistically significant for caffeinated relative to decaffeinated coffee intake. There were no statistically significant associations of coffee consumption with Type I EC among non-obese women or for Type II EC. Tea, cola and chocolate intake were not associated with risk of Type I or Type II EC.

A recently updated meta-analysis of 6 cohort and 10 case–control studies (Je and Giovannucci, 2011) reported a pooled RR of 0.71 (95% CI: 0.62–0.81) for the risk of EC for the highest vs lowest categories of coffee intake, with the strongest inverse association observed in Japanese studies (RR=0.40; 95% CI: 0.25–0.63), intermediate for North American studies (RR=0.69; 95% CI: 0.60–0.79) and weakest but still evident for European studies (RR=0.79; 95% CI: 0.63–0.99). Consistent with our results, four recent studies found an inverse association of coffee with EC, particularly among women with BMI ⩾30 kg m−2 (Friberg et al, 2009; Giri et al, 2011; Gunter et al, 2011; Je et al, 2011). For the first time, we extend this association specifically to Type I EC and to coffee but not other common sources of methylxanthines, which were not addressed by these prior studies.

The exact mechanisms involved in any putative beneficial effect of coffee on EC remain largely unknown. Coffee is a major source of caffeine, and this methylxanthine may increase levels of circulating sex-hormone-binding globulin, thus reducing the concentrations of bioavailable sex-steroid hormones, in particular free oestradiol, and consequently modifying the hormonal milieu leading to downregulation of endometrial hyperproliferation (Ferrini and Barrett-Connor, 1996; Nagata et al, 1998). However, coffee, irrespective of caffeine content, also contains additional compounds with antioxidant activities. These compounds vary widely depending on the type of coffee, roasting and preparation, and many have been found to inhibit the proliferation of tumour cells in vitro (Vivani, 1993; Cavin et al, 2002).

An intriguing hypothesis suggests that coffee may be an insulin sensitiser (Wu et al, 2005; Huxley et al, 2009; Loopstra-Masters et al, 2011). Coffee (both caffeinated and decaffeinated) and caffeine intake were inversely associated with levels of circulating C-peptide, a marker of insulin secretion and resistance, and this association was much stronger in overweight and obese women (Wu et al, 2005).

An inverse association with coffee was not observed for Type II EC, although our analysis was limited by the relatively small number of Type II cases and by the absence of central pathology review. Type I and Type II EC may have different aetiologic pathways and distinct risk factors (Uccella et al, 2011). From a molecular point of view, Type II EC is often associated with p53 mutations, which commonly lead to DNA derangements, chromosomal instability and a more aggressive clinical behaviour (Doll et al, 2008). Conversely, alterations of p53 have been reported in only a small proportion of Type I tumours and, when they occur, they are usually a late event (Doll et al, 2008). Apoptosis of rapidly growing cells induced by caffeine in vitro is dependent on the presence of a functional p53 product, so when p53 is mutated cellular growth is not inhibited by caffeine (He et al, 2003).

In conclusion, our results suggest that coffee consumption, perhaps in part related to caffeine, may be relevant for chemoprevention of Type I EC, particularly among obese women.