The study, published as an open-access research article in BMC Endocrine Disorders, drew on the Korea National Health and Nutrition Examination Survey, known as KNHANES, specifically the fourth wave conducted between 2007 and 2009. The investigators, led by Ziran Qiu of the Department of Breast and Thyroid Surgery at Loudi Central Hospital in Hunan, China, together with colleagues including corresponding author Na Jin, focused on postmenopausal women, a group in which both menopausal hormone therapy and thyroid disease converge with unusual frequency. Thyroid disorders, which range from underactive and overactive gland function to nodules and autoimmune inflammation, are markedly more common in women than in men, and their prevalence rises with age, making the postmenopausal population a natural setting for asking whether exogenous hormones tip the balance.

The analytical sample comprised 3,974 postmenopausal women, of whom 582 reported a history of using female hormone medication. The exposure was deliberately simple: a questionnaire item recording whether the participant had ever taken female hormone drugs. Importantly, the survey did not capture whether use was current, which formulation was taken, at what dose, or for how long, a limitation that shapes how the results can be interpreted. The outcomes were equally grounded in self-report: thyroid disorders diagnosed by a doctor served as the primary endpoint, while current thyroid disorder, lifetime thyroid disorder, and currently treated thyroid disorder served as secondary endpoints. This reliance on questionnaire-defined disease means the study measures the burden of recognized and recorded thyroid illness rather than laboratory-confirmed thyroid dysfunction in the moment.

The headline numbers are striking. Weighted prevalence of doctor-diagnosed thyroid disorder was 10.87 percent among women with a hormone therapy history versus 6.87 percent among women without one. Current thyroid disorder showed a similar gap, 6.79 percent versus 3.79 percent, as did lifetime thyroid disorder at 10.95 percent versus 6.96 percent and treated thyroid disorder at 4.88 percent versus 2.95 percent. In every case, women reporting prior hormone medication use carried roughly a one-and-a-half to nearly two-fold heavier burden of thyroid disease on the raw, survey-weighted scale.

Because raw differences in observational data can easily reflect age, body weight, income, reproductive history, or the simple fact of seeing doctors more often, the team turned to survey-weighted logistic regression, the standard tool for estimating associations in complex national surveys. Their models adjusted for demographic, socioeconomic, metabolic, reproductive, and survey-year factors. To guard against the statistical trap of cherry-picking significant results across multiple tests, they applied Benjamini-Hochberg correction, computing q values that control the expected proportion of false discoveries across the four outcomes. After adjustment, a history of female hormone therapy remained associated with doctor-diagnosed thyroid disorder with an adjusted odds ratio of 1.52 and a 95 percent confidence interval of 1.06 to 2.16, with a q value of 0.031. The estimate for current thyroid disorder was even stronger, at an odds ratio of 1.75 with a confidence interval of 1.09 to 2.83, and lifetime thyroid disorder yielded an odds ratio of 1.50 with a confidence interval of 1.06 to 2.14, both also surviving the false discovery correction at q equal to 0.031. The estimate for currently treated thyroid disorder was directionally similar but less statistically secure, at an odds ratio of 1.60 with a confidence interval spanning 0.96 to 2.68 and a q value of 0.071.

Recognizing that any single modeling choice can swing results, the researchers ran an unusually thorough battery of sensitivity analyses. When they expanded the covariate set further, the primary estimate attenuated to an odds ratio of 1.38 with a confidence interval of 0.96 to 2.00, no longer excluding the possibility of no association. When they adjusted for healthcare contact, an attempt to account for the possibility that hormone therapy users simply interact with the medical system more and therefore get diagnosed more often, the association held at an odds ratio of 1.46 with a confidence interval of 1.01 to 2.10. Restricting the analysis to narrower age bands, however, produced estimates that crossed the null, suggesting the signal may be sensitive to the age composition of the sample. Adding gravidity, the number of times a woman had been pregnant, to the adjustment left the estimate essentially unchanged at an odds ratio of 1.52 with a confidence interval of 1.05 to 2.20. Exploratory analyses of treatment duration returned heterogeneous results, offering no consistent picture of whether longer exposure carried greater risk.

What might connect ovarian hormone preparations and the thyroid gland biologically? The mechanistic story is plausible but unresolved. Estrogen influences thyroid physiology in several documented ways: it can stimulate thyroglobulin, the protein scaffold on which thyroid hormones are synthesized, and it modulates immune activity in a gland where autoimmune disease, particularly Hashimoto’s thyroiditis, is disproportionately common in women. Hormone therapy could also plausibly promote thyroid growth or nodule formation in a gland already rich in estrogen receptors. Conversely, the association could run in the opposite causal direction: women with thyroid symptoms such as fatigue, weight change, or mood disturbance might seek menopausal symptom relief through hormone therapy, or their thyroid conditions might simply be diagnosed around the same life stage. Residual confounding by health-seeking behavior, physician vigilance, or unmeasured metabolic traits remains a live possibility in any cross-sectional design.

The authors are explicit about the limits of their evidence. Because the exposure and outcomes were both self-reported and captured at a single point in time, the study cannot establish temporality, let alone causality. It cannot distinguish whether hormone therapy preceded the thyroid disorder, followed it, or merely co-occurred with it. The questionnaire also could not identify current users or treatment details, blurring any dose or duration effects. Yet the consistency of the association across three of four endpoints, its survival of false discovery correction, and its persistence under healthcare-contact adjustment give the finding enough weight that it should not be dismissed as statistical noise. At the same time, the attenuation under expanded adjustment and the null results in age-restricted analyses are honest reminders that the true effect, if any, is likely modest and context-dependent.

For clinicians and the millions of postmenopausal women weighing hormone therapy, the practical takeaway is one of measured vigilance rather than alarm. The study’s authors state plainly that their findings do not support universal thyroid screening for women with a hormone therapy history, nor do they justify changing menopausal hormone therapy decisions solely on the basis of that history. Standard practice, in which thyroid function is evaluated when symptoms or clinical findings warrant it, remains the appropriate course. What the research does provide is a well-powered, nationally representative East Asian data point in a literature long dominated by Western cohorts, and a clear template for the kind of study needed next: prospective, longitudinal designs with pharmacy-verified hormone exposure, laboratory-measured thyroid function, and careful sequencing of treatment and diagnosis. Until such evidence arrives, the association between hormone therapy history and thyroid disorder burden in postmenopausal Korean women stands as an intriguing signal, a question mark that modern endocrinology will now be pressed to resolve.

Subject of Research: The association between self-reported menopausal hormone therapy history and thyroid disorder prevalence among postmenopausal Korean women in the KNHANES IV survey.

Article Title: Self-reported female hormone therapy history and thyroid disorder burden among postmenopausal Korean women: a cross-sectional KNHANES IV study

Article References: Qiu, Z., Long, H., Li, R., Yang, J., Cao, W., Chen, S., & Jin, N. (2026). Self-reported female hormone therapy history and thyroid disorder burden among postmenopausal Korean women: a cross-sectional KNHANES IV study. BMC Endocrine Disorders. https://doi.org/10.1186/s12902-026-02540-3

Image Credits: AI Generated

DOI: 10.1186/s12902-026-02540-3

Keywords: menopausal hormone therapy, thyroid disorder, postmenopausal women, KNHANES, cross-sectional study, South Korea, endocrinology, odds ratio, thyroid screening, estrogen, women's health, observational study

A new systematic review and meta-analysis published in BMC Cancer has now brought the weight of all available evidence to bear on that question, and its conclusions are both reassuring and subtly intriguing. Led by Mohammad Amouzadeh-Lichahi, Maryam Jafari, Nazila Sattari, Sara Zamani, Negin Letafatkar and Shahriar Ghodous of Guilan University of Medical Sciences in Rasht, Iran, the research team systematically searched PubMed, Scopus and Embase for observational studies published up to July 2026. Their exhaustive sweep of the literature identified twenty-seven reports that met their rigorous inclusion criteria, covering the use of oral contraceptive pills, menopausal hormone therapy, estrogen therapy alone, and combined estrogen–progestin therapy in relation to pancreatic cancer risk. By pooling adjusted effect estimates across studies with random-effects models, the team aimed to produce the most definitive statistical picture to date of how these hormonal exposures relate to one of oncology’s deadliest cancers.

The methodological machinery behind the analysis was considerable. The researchers followed PRISMA reporting guidelines and evaluated study quality using the Newcastle–Ottawa Scale, while the certainty of the resulting evidence was graded with the GRADE framework. Heterogeneity among studies was quantified with the I-squared statistic, and publication bias was probed with both Egger’s and Begg’s tests. Crucially, the team went beyond simple pooled estimates: they computed ninety-five percent prediction intervals, which indicate the range of effects that might plausibly be expected in a future study conducted in a new population, and they performed both linear and non-linear dose–response meta-analyses to test whether longer durations of hormone use translated into progressively altered risk. This layered approach matters because a pooled relative risk alone can mask enormous variability in what individual studies actually find.

The headline result for oral contraceptives was, in statistical terms, a null finding. Across twenty studies, women who had used the pill showed a pooled relative risk of pancreatic cancer of 0.99, with a ninety-five percent confidence interval spanning 0.86 to 1.14 — a range that comfortably includes no effect. The prediction interval was strikingly wide, stretching from 0.62 to 1.59, and heterogeneity between studies was substantial at 61.4 percent. The GRADE assessment rated this evidence as very low certainty. In plain language, the analysis found no overall association between oral contraceptive use and pancreatic cancer, but the underlying studies disagreed with one another enough that the true effect in any given population could conceivably fall anywhere between a meaningful reduction and a meaningful increase in risk.

Menopausal hormone therapy told a similar story. Drawing on eighteen studies, the pooled adjusted relative risk was 0.92, with a confidence interval of 0.79 to 1.06 that again crossed the threshold of no effect. Heterogeneity was even higher here, at 75.9 percent, and the prediction interval ranged from 0.54 to 1.56, once more signaling that results varied dramatically across populations and study designs. Estrogen therapy used alone, examined across six studies, yielded a pooled relative risk of 0.82 with a confidence interval of 0.63 to 1.07 — a point estimate hinting at a possible protective trend, but one that failed to reach statistical significance and was likewise graded as very-low-certainty evidence. For all three of these exposures, the adjusted analyses simply did not support a consistent link with pancreatic cancer.

The one exception emerged with estrogen–progestin therapy, the combined form of menopausal hormone therapy in which estrogen is paired with a progestin to protect the uterine lining. Across five studies, this regimen was associated with a modest but statistically robust inverse association: a pooled relative risk of 0.85, with a tight confidence interval of 0.78 to 0.93 and, remarkably, zero heterogeneity among the contributing studies. The prediction interval of 0.77 to 0.94 mirrored that consistency. This was the only finding in the adjusted analyses to achieve statistical significance, and it was graded as low-certainty evidence — a notch above the very-low ratings given to the other exposures, but still short of the certainty clinicians would want before drawing firm conclusions.

The biology behind such a signal is not far-fetched. Pancreatic ductal adenocarcinoma, the dominant form of the disease, expresses estrogen receptors and the G protein-coupled estrogen receptor, and experimental work has implicated estrogen signaling in pathways central to cancer cell behavior, including the PI3K/AKT and MAPK cascades, insulin-like growth factor signaling, and inflammatory and oxidative stress processes. Progestins, meanwhile, can modulate cell proliferation and differentiation. A modest protective effect of combined estrogen–progestin therapy is therefore biologically plausible, even if the observational nature of the evidence and the small number of contributing studies demand caution. Confounding by indication, differences in body mass index, smoking status, diabetes prevalence and other lifestyle factors across study populations could all shape the observed associations in ways that statistical adjustment cannot fully remove.

Indeed, the secondary crude analyses illustrated exactly how fragile such associations can be. When the researchers pooled unadjusted effect estimates instead of adjusted ones, they found substantially lower odds of pancreatic cancer among oral contraceptive users (an odds ratio of 0.70) and estrogen–progestin users (an odds ratio of 0.55). But both of these crude analyses were plagued by substantial heterogeneity and wide prediction intervals — the estrogen–progestin prediction interval stretched from 0.16 to 1.97 — suggesting that the apparent protection in crude data may partly reflect confounding by factors such as socioeconomic status, healthcare access, or healthier baseline profiles among hormone users rather than a genuine biological effect. The divergence between crude and adjusted results is a textbook demonstration of why observational hormone epidemiology is so treacherous.

The dose–response analyses added a further layer of nuance. Neither oral contraceptive use nor menopausal hormone therapy showed a significant linear relationship with pancreatic cancer risk per five-year increase in duration of use: the relative risks were 1.07 (p = 0.158) for the pill and 0.84 (p = 0.128) for menopausal hormone therapy. Non-linear models likewise failed to identify any significant threshold or U-shaped pattern. In other words, there was no evidence that taking hormones for longer periods progressively shifted risk in either direction, which weakens the case for a strong causal relationship and suggests that any real effects, if they exist, are likely to be modest and context-dependent.

The authors’ conclusions are measured and appropriately cautious. The adjusted evidence did not show a consistent association between oral contraceptive, menopausal hormone therapy, or estrogen therapy use and pancreatic cancer risk, and although estrogen–progestin therapy was linked to a modest reduction in risk, that finding rests on low-certainty observational evidence from only five studies. The wide prediction intervals for oral contraceptives and menopausal hormone therapy indicate that associations may vary across populations and settings, and the team explicitly calls for further large prospective studies to clarify the long-term hormonal effects on pancreatic carcinogenesis. For the millions of women weighing the risks and benefits of hormonal medications, the practical takeaway is largely reassuring: nothing in this comprehensive synthesis suggests that these widely used therapies meaningfully raise the risk of pancreatic cancer. And for researchers, the modest inverse signal from combined therapy offers a tantalizing, testable hypothesis — one that could ultimately illuminate how estrogen and progestin signaling shapes one of medicine’s most lethal tumors.

Subject of Research: The association between oral contraceptive and menopausal hormone therapy use and pancreatic cancer risk

Article Title: Oral contraceptive and menopausal hormone therapy use and risk of pancreatic cancer: a systematic review, meta-analysis, and dose–response analysis

Article References: Oral contraceptive and menopausal hormone therapy use and risk of pancreatic cancer: a systematic review, meta-analysis, and dose–response analysis. (n.d.). https://doi.org/10.1186/s12885-026-16902-6

Image Credits: AI Generated

DOI: 10.1186/s12885-026-16902-6

Keywords: pancreatic cancer, oral contraceptives, menopausal hormone therapy, estrogen therapy, estrogen–progestin therapy, meta-analysis, systematic review, dose–response analysis, observational studies, exogenous hormones, BMC Cancer, cancer epidemiology