Incidence of gynecological cancers following prolactin‐increasing antipsychotic use: a population cohort study

Long‐term use of prolactin‐increasing antipsychotics has been found to be associated with breast cancer ¹ . A recent meta‐analysis of observational studies has also reported an increased risk of gynecological cancers by nearly 70% in relation to antipsychotic use ² . Nevertheless, that meta‐analysis was limited by small sample size, heterogeneity across studies, narrow inclusion of specific cancer types, as well as by the case‐control design adopted by most included studies. There was also no direct comparison between users of prolactin‐increasing and prolactin‐sparing antipsychotics on a comparable timeframe ³ , ⁴ , ⁵ , ⁶ .

We took advantage of the longitudinal anonymized territory‐wide electronic health records maintained by the Hospital Authority of Hong Kong to build a large retrospective cohort and estimate the weighted incidence rate ratio (IRR) and rate difference (RD) of gynecological cancers between users of prolactin‐increasing and prolactin‐sparing antipsychotics. This study was approved by the Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster and the Hospital Authority Central Institutional Review Board (no. UW 20‐113, CIRB‐2022‐015‐5).

Women aged 18 to 85 years who initiated antipsychotic treatment between January 2006 and December 2018 in Hong Kong were identified. Patients with previous antipsychotic use, hysterectomy or mastectomy, or cancers (except non‐melanoma skin cancer) before the index date were excluded. Patients were followed from antipsychotic initiation until the diagnosis of gynecological cancer, or 365 days after change of antipsychotic (a switch to antipsychotic other than the one initiated on the index date), or 365 days after the discontinuation of antipsychotics, or death, or December 31, 2023, whichever occurred the earliest. To avoid reversed causality, patients developing cancer within 30 days of the index date were censored upon cancer diagnosis.

Study exposure was defined as use of prolactin‐increasing antipsychotics (i.e., those associated with moderately to highly elevated prolactin levels), with use of prolactin‐sparing antipsychotics (i.e., those associated with low or non‐elevated prolactin) ⁵ as the comparator (see supplementary information for the classification of antipsychotics). The main outcome was gynecological cancer (ICD‐9: 179‐180.9, 182.0‐184.9) and its subtypes, including cervical (180), endometrial (179, 182), ovarian (183) and vaginal cancers (184) ⁵ , ⁷ .

Covariates included age, reproductive history (with or without childbearing experience), year of cohort enrollment, baseline comorbidities, and number of days on other medications over the past two years. Comorbidities included diagnoses of schizophrenia, bipolar disorder, paranoid disorder, other non‐organic psychoses, depression, anxiety disorder, dementia, substance use disorders, diabetes, hypertension, obesity (body mass index ≥30), disorder of lipid metabolism, autoimmune diseases, chronic kidney disease, ovarian dysfunction, human papillomavirus infection, prolactinoma, and human immunodeficiency virus (HIV) infection, recorded since the inception of the database in 1993. Prior use of other medications included aspirin, beta‐blockers, calcium channel blockers, contraceptives, diuretics, female hormones, hypnotics and anxiolytics, non‐steroidal anti‐inflammatory drugs, opioids, paracetamol, renin‐angiotensin system inhibitors, statins and digoxin.

Poisson regression with inverse probability of treatment weight‐ing and robust variance estimation was used to estimate the IRR with 95% confidence intervals (CIs) of gynecological cancers between prolactin‐increasing and prolactin‐sparing antipsychotic users. Weighed RD was also estimated. Age‐stratified analysis was conducted by the threshold of ≥51. In addition, patients with psychotic disorders, bipolar disorder, paranoid disorder, other non‐organic psychoses, and none of these conditions were separately analyzed. Head‐to‐head comparisons were made between users of specific agents.

Several sensitivity analyses were further conducted. First, in a repeated analysis, only patients who switched to the different prolactin‐inducing category of antipsychotics were censored. Second, we excluded antipsychotics with moderate prolactin‐increasing effects from the prolactin‐increasing group. Third, we reduced the period for defining drug discontinuation from 90 to 60 days. Fourth, we followed up patients for an additional 180 days (instead of 365 days) if they discontinued antipsychotics. Fifth, we limited the maximum follow‐up time to 10 years. Sixth, we extended the initial censoring window for cancer occurrence from 30 to 60 days after the index date. Seventh, we conducted the analysis without an initial censoring window for cancers after the index date. Eighth, patients on aripiprazole were excluded, because this drug may counteract the prolactin‐increasing effects of other antipsychotics. Ninth, we used breast cancer (174) as a positive control outcome, and tuberculous meningitis (013) as a negative control outcome, for a repeated main analysis. Lastly, we repeated the analysis with weighted Cox regression. All the analyses were conducted in the R statistical programming environment (version 4.1.2). Statistical tests were two‐tailed, and a p value ≤0.05 was indicative of statistical significance.

In total, 84,061 female new antipsychotic users were included, with a follow‐up period up to 15 years (median: 1.18; 25th‐95th percentile: 1.02‐1.99). Of these, 61,771 (73.5%) were prolactin‐increasing antipsychotic users. The three main antipsychotics prescribed were quetiapine, haloperidol and risperidone. During the follow‐up period, 126 cases of gynecological cancers were identified, with a crude incidence rate of 77.86 per 100,000 person‐years. Of these, 96 cases occurred in the prolactin‐increasing group (88.12 per 100,000 person‐years), and 30 in the prolactin‐sparing group (56.72 per 100,000 person‐years).

Poisson regression showed that prolactin‐increasing antipsychotic users had a significantly higher rate of gynecological cancers compared to prolactin‐sparing antipsychotic users (IRR=1.99, 95% CI: 1.13‐3.50). The weighted RD of gynecological cancers was 42.58 (95% CI: 11.77‐69.32) per 100,000 person‐years, with the rate notably elevated among patients aged over 51 years (IRR=2.09, 95% CI: 1.02‐4.26), and those diagnosed with psychotic disorders (IRR=9.69, 95% CI: 1.32‐71.10).

In head‐to‐head drug comparisons, the increased rate remained statistically significant for the quetiapine‐haloperidol and quetiapine‐risperidone comparisons (IRR=2.27, 95% CI: 1.21‐4.26; and IRR=2.20, 95% CI: 1.10‐4.37). Regarding gynecological cancer subcategories, elevated risks of endometrial cancer (IRR=1.94, 95% CI: 1.02‐3.69) and ovarian cancer (IRR=3.68, 95% CI: 1.17‐11.56) were observed. The increased risk of cervical or vaginal cancer did not reach statistical significance (IRR=1.33, 95% CI: 0.44‐4.04 and IRR=2.14, 95% CI: 0.43‐10.73, respectively). Similar results were seen in the sensitivity analyses, including the positive control outcome analysis on breast cancer (IRR=1.63, 95% CI: 1.10‐2.43), with the negative control outcome analysis showing a non‐significant IRR close to 1 (IRR=0.96, 95% CI: 0.37‐2.55).

To the best of our knowledge, this is the first cohort study, and the largest real‐world study, to compare prolactin‐increasing with prolactin‐sparing antipsychotic use in terms of incidence of gynecological cancers. Findings suggest a two‐fold increased rate of this incidence among users of prolactin‐increasing antipsychotics compared with those using prolactin‐sparing antipsychotics, particularly evident for ovarian and endometrial cancers, but with small RDs, i.e., approximately one case per 2,300 person‐years. Preclinical studies suggesting biological links ⁸ and observational studies supporting an empirical association ⁹ are far from scarce, with preliminary evidence suggesting that prolactin may promote the growth of ovarian and endometrial surface epithelial cells, enhance ovarian cancer cell survival and migration, and inhibit apoptosis ⁸ . Key limitations of the study include the lack of randomization and potentially limited generalizability beyond Chinese populations.

In conclusion, our study showed that prolactin‐increasing antipsychotics are associated with an increased rate of gynecological cancers compared with prolactin‐sparing antipsychotics, although the RD is small, entailing little clinical significance.