The cohort included 314 male and 272 female patients. Their mean age was 43.0 ± 11.5 years, and analyzable electrocardiographic data were available for 585 patients because one record contained a physiologically implausible value. At admission, which the investigators designated as the pre-treatment measurement, women had a mean Bazett-corrected QT interval of 430.3 ± 23.3 milliseconds, compared with 416.6 ± 23.9 milliseconds in men.

The mean difference was 13.7 milliseconds, with a 95% confidence interval of 9.9 to 17.6 milliseconds. The baseline sex difference remained significant with the Fridericia, Framingham, Hodges and Rautaharju corrections, all yielding P values below 0.001. Men, however, had wider QRS complexes and longer PR intervals: 91.6 ± 10.2 versus 86.0 ± 9.2 milliseconds for QRS duration and 150.1 ± 20.5 versus 144.8 ± 18.5 milliseconds for PR interval.

Heart rate did not differ significantly between the sexes at admission. The investigators used sex-specific QTc thresholds to characterize prolongation. A “prolongation tendency” was defined as above 440 milliseconds in men and above 460 milliseconds in women, while “marked prolongation” meant above 450 milliseconds in men and above 470 milliseconds in women.

At admission, 15.0% of men met the threshold for prolongation tendency, with a 95% confidence interval of 11.5% to 19.4%, compared with 8.8% of women, with a 95% confidence interval of 6.0% to 12.8%. Marked prolongation was also more frequent among men, affecting 5.4% versus 1.5% of women, with a Fisher exact P value of 0.013. The unadjusted odds ratio for marked prolongation in men was 3.85, with a 95% confidence interval of 1.28 to 11.58.

After adjustment for age, hospital stay and antipsychotic drug group, female sex was associated with lower odds of prolongation tendency, with an odds ratio of 0.49 and a 95% confidence interval of 0.29 to 0.85. Age, length of stay and drug group were not significantly associated with that outcome in the multivariable model. Over the course of hospitalization, Bazett-corrected QTc fell by 4.6 ± 22.9 milliseconds in men, a statistically significant within-group change with P = 0.0004.

The corresponding change in women was a smaller, nonsignificant reduction of 2.2 ± 25.4 milliseconds, with P = 0.165. The unadjusted difference between these changes was not statistically significant under Bazett or any of the four alternative formulas. After accounting for baseline QTc, age and length of stay, however, women showed less shortening than men: the baseline-adjusted female coefficients across the five correction methods ranged from +6.48 to +9.75 milliseconds, with every P value below 0.001.

In the fully adjusted model, which also included antipsychotic group, the female coefficient was +9.91 milliseconds, with a 95% confidence interval of +6.53 to +13.30 milliseconds. No antipsychotic group differed significantly from risperidone in adjusted QTc change. Threshold-defined outcomes told a more specific story.

Marked QTc prolongation among men declined from 5.4% at admission to 0.6% before discharge, a within-sex change supported by an exact McNemar P value of 0.0007. Among women, the proportion moved from 1.5% to 1.1%, with no significant within-sex change and an exact McNemar P value of 1.000. The discharge difference between the sexes was therefore not significant, with a Fisher exact P value of 0.668.

The proportion meeting the broader prolongation-tendency threshold declined from 15.0% to 10.8% in men and from 8.8% to 6.6% in women, although the sex difference at discharge did not meet the study’s significance threshold, with P = 0.081. Not every apparent treatment-period pattern survived scrutiny. Male heart rate declined by 4.8 ± 15.9 beats per minute, compared with 2.7 ± 16.2 beats per minute in women, and heart rate can influence rate-corrected QT values.

Crude QTc shortening in men was significant under the Bazett, Hodges and Rautaharju formulas, but not under Fridericia or Framingham. Females showed no statistically significant shortening under any of the five methods. The magnitude and statistical significance of the crude change therefore varied with the correction approach.

Applying the same sex-specific thresholds to Fridericia-corrected values reduced the apparent prevalence of marked prolongation at admission to 0.6% in men and 0% in women, illustrating how absolute prevalence depends on the correction formula. By contrast, the baseline-adjusted sex difference in QTc change remained significant under every formula. The sex-by-time interaction was not significant for QTc in the mixed-effects model, with P = 0.215, and none of the exploratory between-sex comparisons of change survived false-discovery-rate correction.

Subgroup analyses examined whether particular antipsychotic regimens altered the overall pattern. Baseline QTc remained longer in women than men among patients receiving olanzapine, risperidone or combination therapy, but the sex difference was not significant in the smaller clozapine subgroup. No significant sex difference in QTc change appeared within any drug subgroup.

Of the 253 patients receiving combination therapy, 192, or 75.9%, had a regimen containing clozapine or olanzapine; across the full cohort, 344 of 586 patients, or 58.7%, received a higher-QTc-risk regimen. When the models were repeated using this binary classification in place of detailed drug groups, the baseline-adjusted sex difference in QTc change was essentially unchanged. The higher-risk classification itself showed only a weak association with QTc change under Bazett, with a coefficient of +3.43 milliseconds and P = 0.048, but no association under Fridericia, with a coefficient of −1.48 milliseconds and P = 0.35.

The study also recorded lipid and glucose measures. At admission, women had higher total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol and fasting blood glucose than men, while triglycerides did not differ significantly. Both sexes showed significant declines in high-density lipoprotein and fasting blood glucose during hospitalization.

Triglycerides increased significantly in women by 0.19 ± 1.27 mmol/L, with P = 0.009, and showed a nonsignificant upward trend in men. After adjustment, the study reported sex differences in changes in QTc, heart rate, QRS duration, high-density lipoprotein and total cholesterol. In mixed-effects models, however, the only significant sex-by-time interaction among the metabolic measures was for high-density lipoprotein, and that result did not survive false-discovery-rate correction, with q = 0.108.

The metabolic assessment was limited to fasting glucose and a standard lipid panel; glycated hemoglobin, insulin and homeostatic indices of insulin resistance were not measured, and acute illness or stress could have affected the values. Several features constrain interpretation. The study was retrospective and observational, and most participants were already receiving antipsychotics at admission; the first electrocardiogram therefore did not represent a medication-naive baseline, and subsequent changes could reflect drug switching, dose titration or adjunctive treatment rather than treatment initiation.

Important potential determinants—including body mass index, serum electrolytes, antipsychotic blood concentrations, smoking, alcohol use, physical activity, menopausal status, benzodiazepine use and other concomitant medications—were missing or incomplete. Baseline differences in age and length of stay were included as covariates but could not eliminate all confounding. The two hospitals were not represented by an explicit study-site variable, although the direction of the baseline sex difference was similar in an unvalidated proxy derived from medical-record numbers.

Heart rate and other electrocardiographic measurements came from automated algorithms with clinician over-reading rather than blinded manual measurement. Raw digital waveforms were unavailable for direct recalculation from the original QT interval or assessment of dynamic QT–RR relationships, and no clinical arrhythmic outcomes were available. The results are best described as exploratory observations rather than evidence of a biological treatment effect.

Women began with longer QTc intervals across all correction formulas, while men were more likely to exceed the marked-prolongation threshold at admission. The baseline-adjusted evidence of greater male shortening was consistent across formulas, but the crude result was formula-dependent, the unadjusted between-sex change was not significant, and the QTc sex-by-time interaction was not significant. The analysis also cannot establish that the electrocardiographic differences translated into different risks of torsades de pointes, arrhythmia, sudden cardiac death or other clinical outcomes.

Prospective studies incorporating complete site information, detailed drug exposure, validated QT measurement, dynamic QT–RR analysis and clinical outcomes are needed to determine whether these sex-stratified patterns persist outside this hospitalized Chinese cohort.

Subject of Research: Sex differences in QTc and electrocardiographic measures before and during antipsychotic hospitalization for schizophrenia

Article Title: Sex differences in QTc prolongation and electrocardiographic changes before and after antipsychotic treatment of schizophrenia

Article References: Song, X., Fan, F., & Liang, J. (2026). Sex differences in QTc prolongation and electrocardiographic changes before and after antipsychotic treatment of schizophrenia. Frontiers in Psychiatry, 17, Article 1988458. https://doi.org/10.3389/fpsyt.2026.1988458

Image Credits: AI Generated

DOI: 10.3389/fpsyt.2026.1988458

Keywords: schizophrenia, QTc interval, antipsychotic treatment, sex differences, electrocardiography, QT prolongation, heart-rate correction, observational study, cardiac monitoring, China, differences, prolongation