Association of Menopausal Hormone Therapy with Risk of Pancreatic Cancer: A Systematic Review and Meta-analysis of Cohort Studies

Globally, the age-standardized incidence rate of pancreatic cancer increased from 5.2 per 100,000 in 1990 to 6.4 per 100,000 in 2017 (1). Over the same period, the disability-adjusted life-years (DALYs) that measure the years of healthy life lost due to fatal and nonfatal burden attributed to pancreatic cancer had doubled from 4.4 million to 9.1 million (2). Owing to the lack of clinical symptoms and difficulties in diagnosis, pancreatic cancer is usually detected at advanced stages, leading to poor 5-year survival rates between 2% to 9% (2). In 2019 alone, pancreatic cancer caused more than 531,000 deaths, making it the fifth leading cause of cancer death worldwide (1).

Menopausal hormone therapy (MHT), including various estrogen-only and estrogen-progestin combined regimens, is effective to alleviate menopausal symptoms (3). It has long been postulated that estrogen has inhibitory effects on pancreatic cancer development (4–6) and therefore, MHT could potentially be associated with a reduced risk of pancreatic cancer. A previous meta-analysis, incorporating publications before 2013, reported a null association between MHT use and the risk of pancreatic cancer (7). Since then, more studies have been conducted with conflicting results (8–13). More importantly, whether the association of MHT with pancreatic cancer varies by MHT types, recency, or treatment duration has not been quantified in aggregation. Given the clinical importance of MHT and discordant results on pancreatic cancer risk associated with MHT, a systematic review and meta-analysis summarizing the most updated evidence is appropriate.

This systematic review and meta-analysis aimed to assess the evidence from cohort studies of the association between MHT and pancreatic cancer risk. In addition, we explored the associations by different MHT formulations, recency, and duration of treatment.

This study reported in accordance with the Preferred Reported Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guidelines (Supplementary Table S1). Two of the reviewers (C.Y. Leung and H.L. Huang) systematically searched PubMed, Embase, and Cochrane Library for articles published in peer-reviewed journals, from inception to April 20, 2022, using keywords relating to MHT, pancreatic cancer, and risk. We applied no language restrictions. Only studies provided relative risks (RR) with 95% confidence intervals (CI) or provided sufficient data that allow risk estimates to be computed were eligible for inclusion. Preestablished review protocol was registered in PROSPERO (CRD42022327363). Details of the search strategies are described in Supplementary Tables S2–S4.

Titles and abstracts of articles retrieved from the search were screened independently by the two reviewers. Full texts of potentially relevant studies were obtained and evaluated. The search was also supplemented by screening reference lists of included articles and relevant reviews. We included clinical trials and cohort studies that reported associations between the use of MHT and risk estimates for pancreatic cancer in women. We excluded case–control studies, case reports, reviews, letters, and conference abstracts, along with studies only assessing mortality risk. Disagreement on eligibility was resolved by consensus.

The two reviewers independently extracted relevant data from individual studies using a predefined data extraction form. Extracted data included the name of the first author, publication year, country, study design, study period, age, sample size, information on MHT (i.e., former or current user, treatment type, duration of treatment), source of pancreatic cancer diagnoses, adjustment, and risk estimates. We extracted risk estimates that were the most adjusted. If reported, estimates stratified by recency (former or current users), treatment types (estrogen plus progestin or estrogen only), and duration of treatment (years) were also extracted. The risk of bias in each study was assessed using the Newcastle-Ottawa Quality Assessment Scale (14). Disagreements between reviewers were resolved through discussions with a third reviewer (Y.C. Jang).

Pooled RRs and 95% CIs were estimated using random-effects meta-analysis. When risk estimates required were unavailable, we harmonized reported subgroup risks to a combined risk estimate using fixed-effect meta-analysis. For example, the risk estimate for the overall use of MHT from Skinner and colleagues was obtained from the reported estimates of former and current users (15). Details of combined results are available in the Supplementary Tables S5–S7. The 95% prediction intervals (PI) describing the heterogeneity in random-effects meta-analysis were computed to inform potential future treatment effects in 95% of all populations. In case of heterogeneity, the 95% PIs enable evaluation in addition to the risk estimate and its corresponding 95% CIs (16). We also performed prespecified stratified meta-analyses defined by treatment formulations (estrogen plus progestin or estrogen only), recency (former or current users), and duration of treatment (years). The percentage of variation attributable to between-study heterogeneity was quantified by I² statistics, with I² 0% to 25% indicating low heterogeneity, 26% to 50% indicating moderate heterogeneity, 51% to 75% indicating substantial heterogeneity, and > 75% indicating high heterogeneity (17). We performed random-effect meta-regression to explore potential whether the sources of heterogeneity varied according to participant number (≤133,690 or >133,690), MHT assessment (prescription records or self-reported), publication year (<2013 or ≥2013), and year of study enrollment (<2002 or ≥2002, publication year of the Women's Health Initiative results). Potential publication bias and small-study effects were assessed visually using funnel plots and statistically using Egger test (18). If evidence of publication bias existed, trim and fill method was performed (19). In the current analysis, exact P values are provided unless P < 0.0001. All statistical analyses were using STATA version 16.1 (College Station, TX).

Of 471 potentially relevant articles identified through the literature search, 27 were eligible for full-text review after 18 duplicates and 426 irrelevant studies were excluded. The search details and study selection process are summarized in Fig. 1. Eleven unique cohort studies met inclusion criteria and were included in the final meta-analysis (Fig. 1). Table 1 presents a summary of included studies and Supplementary Table S8 gives details on excluded studies. In total, the studies comprised 2,712,313 women, with 6,482 cases of pancreatic cancer diagnosed. The studies began between 1976 and 2005, with six published after 2013. Six studies were conducted in North America, four in Western Europe, and one in Asia. Ten studies had low risk of bias and one had a moderate risk (Supplementary Table S9).

Overall, there was no association between MHT and pancreatic cancer risk (Pooled RR, 0.92; 95% CI, 0.83–1.02; I², 64.0%; Fig. 2). The 95% PI (0.68–1.25) suggested that the association would likely be similar in future studies (Supplementary Fig. S1). Meta-analyses of four studies with data stratified by MHT formulations showed that the use of estrogen-only MHT was inversely associated with the risk of pancreatic cancer. The pooled RRs for pancreatic cancer was 0.77 (95% CI, 0.64–0.94; I², 56.6%) for women receiving estrogen-only MHT (Fig. 3A). Similarly, compared with nonusers, the pooled RR for pancreatic cancer was 0.85 (95% CI, 0.75–0.96; I², 0%) for women receiving estrogen plus progestin MHT (Fig. 3B). Prespecified subgroup analysis defined by recency did not reveal evidence of associations between MHT and pancreatic cancer risk among current users (Pooled RR, 0.94; 95% CI, 0.80–1.10; I², 61.1%) and former users (Pooled RR, 0.97; 95% CI, 0.89–1.04; I², 0%) when compared with never-users of MHT (Fig. 4). The pooled RR of < 5 years of MHT with never use was 0.99 (95% CI, 0.90–1.09; I², 0.01%). Pooled RRs comparing women had a longer duration of MHT to those who never received MHT were 0.97 (5–9 years; 95% CI, 0.84–1.11; I², 0%) and 0.90 (≥ 10 years; 95% CI, 0.79–1.02; I², 23.2%; Fig. 5).

Meta-regression analyses showed no between-group difference by participant number (P = 0.90), MHT assessment (P = 0.18), initiation year (P = 0.18) and publication year (P = 0.59). Funnel plots did not demonstrate evidence of publication bias (Supplementary Fig. S2). There was no evidence of publication bias by the Egger test (P = 0.66).

This systematic review and meta-analysis, including 2,712,313 women from 11 cohorts, found no association between overall use of MHT and risk of pancreatic cancer. This finding was consistent across MHT recency and duration. However, subgroup analysis defined by MHT formulations showed that estrogen-only therapy was associated with a 23% reduction in risk of pancreatic cancer and combined estrogen-progestin therapy reduced pancreatic cancer risk by 15%. The associations were robust to the assessments of publication bias.

In several large randomized clinical trials, MHT has been shown to reduce risk of diabetes (20–22) and the level of fasting glucose (20–23), which are important risk factors for pancreatic cancer (24, 25). Laboratory studies also suggested inhibitory effects of estrogen on pancreatic tumor growth (4–6). This hypothesis was supported in The California Teachers Study, which demonstrated that MHT users were at a 30% lower risk of pancreatic cancer compared with never users (26). Similar findings were reported by a nationwide cohort study in Sweden, where the use of MHT was associated with a 23% reduction in pancreatic cancer risk (9). Nevertheless, other cohort studies found a null overall association of MHT use with risk of pancreatic cancer (8, 10–13, 15, 27–30). The discordance might be partly due to various degrees of residual confounding across studies. In addition, from four included studies with data on MHT formulations, our analysis demonstrated that estrogen-only and combined estrogen-progestin therapy were associated with a reduced risk of pancreatic cancer. In the current analysis, seven of 11 included studies only provided data on overall use of MHT, which may also explain the null association in our study. Future studies focusing on different MHT formulations are needed to confirm our findings.

Despite evidence demonstrating that MHT was linked to a decreased risk of diabetes and fractures in menopausal women, the prevalence of MHT use dropped significantly after the publication of studies suggesting the association between MHT use and adverse health outcomes in 2002 (31, 32). The potential impact of overstressing the risks associated with MHT use has received considerable attention (33), but little evidence was available to provide detailed and comprehensive evaluation. Indeed, the benefits and harms of using MHT are still a question that remains unsolved. Our findings show, in accordance with previous meta-analysis summarizing evidence before 2013, that the use of MHT was not associated with risk of pancreatic cancer. Tang and colleagues focused only on overall risk estimates due to the scarcity of data on different MHT formulations, recency, and duration of treatment (7). The updated evidence allowed us to perform further investigation and the results showed that estrogen-only and combined estrogen-progestin therapy were associated with a reduction in pancreatic cancer risk. Of note, our findings do not ipso facto imply that MHT should be indicated for the primary prevention of pancreatic cancer. Our results need to be considered alongside other studies and guidelines, which evaluate comprehensive benefits and harms of MHT use.

The major strengths of the current study were the large sample representative of the general population to increase the generalizability of the findings, the inclusion of most recently published studies, and the comprehensive subgroup meta-analyses that highlighted the inverse associations of the use of different MHT formulations with risk of pancreatic cancer. This study had some limitations. First, given the observational nature of the included studies, it is possible that residual confounding remains. However, known major risk factors were adjusted in the majority of included studies. Nine of the 11 studies adjusted for cigarette smoking (8, 9, 11, 15, 26–30), nine adjusted for body mass index or obesity (8, 9, 11, 15, 26–30), and six adjusted for diabetes (8, 9, 15, 28–30). In addition, we pooled fully adjusted risk estimates, which reduced the potential for residual confounding. Second, because information on MHT in eight studies was self-reported, potential measurement error is inevitable (8, 12, 15, 26, 27–30). Although meta-regression showed that MHT assessment methods (prescription data or questionnaire) was not a significant effect moderator, the method of collecting questionnaires varied among included studies. In the Nurses’ Health Study, MHT use was repeatedly assessed to ensure accurate detection of any association (15). In contrast, some studies collected data of MHT use by self-administered questionnaires at baseline only (8, 11, 12, 27–30). Therefore, some misclassification is inevitable. Nevertheless, all data on MHT were collected prospectively, misreporting could be random and biased the associations towards null. Third, albeit we applied a rigorous predefined systematic searching, only one of 11 included studies was from Asia (13). This requires further investigations in large Asian cohorts. Similarly, comprehensive evaluation stratified by ethnicity is unavailable, which limits the generalizability of the results. Nonetheless, we assessed PIs to demonstrate the possible findings of future studies. Fourthly, while the age at first MHT use is of clinical importance, current data do not allow us to perform subgroup analysis by the age at the initiation of MHT. This requires further investigation in relation to pancreatic cancer risk. Fifthly, previous research suggested that, compared with women without hysterectomy, women with prior hysterectomy were at a higher risk of pancreatic cancer (29). Estrogen-only MHT was only available for women who had prior hysterectomy; therefore, the inverse association between estrogen-only MHT and risk of pancreatic cancer could be attenuated. Finally, we cannot completely rule out the possibility of healthy user bias because MHT users may have a healthier lifestyle than nonusers.

In conclusion, the current meta-analysis of 2,712,313 women from 11 cohorts found no association between the overall use of MHT and risk of pancreatic cancer, regardless of recency or duration of treatment. The current evidence with data on MHT formulations, however, showed a reduced risk of pancreatic cancer in women taking estrogen-only therapy or combined estrogen-progestin therapy, compared with never users. Further high-quality cohort studies are needed to investigate the association of MHT with risk of pancreatic cancer by various MHT formulations. This study would be helpful to physicians, patients, and policy makers for decision making regarding the use of MHT.

No disclosures were reported.

Y.-C. Jang: Writing–original draft, writing–review and editing. C.Y. Leung: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. H.-L. Huang: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.