Vaginal estrogen prescribing following a diagnosis of recurrent urinary tract infection was associated with lower unadjusted rates of sepsis, hospital admission, and all-cause mortality across every age group, according to an analysis of nearly 1.9 million female patients in the Epic Cosmos database, published in Urology.

The findings add exploratory outcomes data to a therapy that guidelines support for reducing recurrent urinary tract infection in postmenopausal women, and that the study authors described as underutilized.

The main caveat: this was an observational, unadjusted database analysis using aggregate-level data. The researchers identified healthy prescriber/detection bias as the study's most important limitation, noting that patients who received vaginal estrogen may also have received earlier infection treatment, specialist referral, or more proactive comorbidity management. They wrote that the prescription itself may be a marker of higher-quality care rather than the causal agent driving outcomes.

What the Study Did

Researchers used Epic Cosmos, a HIPAA-compliant, deidentified, aggregate-level database drawing on more than 304 million patient records from more than 2,070 hospitals and 47,000 clinics as of May 2026.

They identified patients aged 20 to 99 years with documented legal sex of female and recurrent urinary tract infection between October 2017 and October 2025. Birth sex and gender identity were not available in the data set.

Recurrent urinary tract infection was defined as 2 separate coded urinary tract infections 1 to 6 months apart. Patients were classified as vaginal estrogen recipients if they had a documented prescription within 2 months of their second urinary tract infection, or as nonrecipients if they had no history of vaginal estrogen in the medical record. Patients prescribed vaginal estrogen outside that 2-month window were excluded.

Outcomes included sepsis, any hospital admission, and all-cause mortality within 8 years. Because Cosmos provides aggregate-level data, the researchers could not perform multivariable adjustment for individual confounders such as comorbidities; all reported odds ratios were unadjusted.

Key Numbers

Of 1,891,956 patients with recurrent urinary tract infection, 97,109, or 5.1%, received a vaginal estrogen prescription within 2 months of diagnosis; 1,794,847, or 94.9%, did not.

Across the 4 age strata, unadjusted rates were consistently lower among recipients:

• Sepsis: 4.2% to 10.4% in recipients vs 8.5% to 24.1% in nonrecipients

• Hospital admission: 7.5% to 12.0% vs 15.7% to 27.5%

• All-cause mortality: 0.3% to 3.9% vs 0.8% to 9.6%

The authors reported that the associations were stronger for sepsis and death in the 40- to 69-year age groups and for admission in the 20- to 39-year age group. They also reported that a shorter 3-year follow-up window, performed to assess lead-time or differential follow-up bias, showed consistent and proportional results.

Baseline comorbidities differed between groups. In younger cohorts, vaginal estrogen recipients had higher rates of chronic kidney disease, immunodeficiency, neurogenic bladder, urinary retention, and urinary device use than nonrecipients. Among patients aged 55 to 69 years, diabetes and chronic kidney disease were more common among nonrecipients. The researchers described most measured imbalances as small to modest and inconsistent in direction, but also noted that residual confounding likely persisted because multivariable adjustment was not feasible.

Subgroup Considerations

The 20- to 39-year admission gap may be confounded by pregnancy. This group showed the largest relative difference in admission, which the researchers wrote was "most likely attributable to a confounding variable." They noted that patients of childbearing age may be admitted for labor and delivery and may be more likely to be observed inpatient for urinary tract infection during pregnancy.

Vaginal estrogen use in premenopausal patients is clinically atypical. The researchers wrote that vaginal estrogen is generally reserved in this age group for specific scenarios such as primary ovarian insufficiency, hormone manipulation for breast cancer treatment, breastfeeding, or use of nonbarrier hormonal contraception. They cautioned that premenopausal patients prescribed vaginal estrogen may reflect a specific clinical phenotype rather than a broadly generalizable group.

Underuse, Even Where Guidelines Are Clear

In the broader Cosmos population, not restricted to the 2-month prescribing window used for the primary analysis, 25% of patients with recurrent urinary tract infection had ever received a vaginal estrogen prescription. This included 37.5% of patients aged 55 to 69 years and 31.7% of those aged 70 to 99 years. The researchers wrote that this low rate may reflect provider hesitancy, patient concerns, systemic barriers to care, lack of capture in Cosmos data, or a combination of these factors.

Limitations

The authors emphasized that vaginal estrogen prescription may reflect healthier prescribing environments or more proactive care rather than a causal treatment effect.

They also noted several additional constraints:

• All reported odds ratios and relative risks were unadjusted, and no multivariable regression was possible because Cosmos provides aggregate-level data.

• The study could confirm that a prescription was written, but not whether it was filled or used consistently.

• Death was captured through discharge disposition codes, which the researchers wrote likely underestimated mortality.

• Neither P values nor confidence intervals were adjusted for multiplicity, and the researchers described the findings as exploratory.

• The Epic Cosmos population may skew toward larger, urban, better-resourced health systems and may not generalize to safety-net, rural, or under-resourced settings.

Bottom Line

The authors themselves frame this analysis as a strong and consistent association between timely vaginal estrogen prescribing and lower unadjusted rates of sepsis, hospital admission, and death after recurrent urinary tract infection — while explicitly cautioning that, because the study was observational, unadjusted, and limited to aggregate-level data, it cannot establish that vaginal estrogen caused the lower rates of serious adverse outcomes.

Mortality odds ratios were as low as 0.20 in some age strata — a magnitude that makes the authors' own residual-confounding caveat particularly consequential to weigh. The researchers called for prospective observational studies to better understand outcomes, including the potential impact of vaginal estrogen use on serious adverse events.

Read alongside the design limitations the authors describe, the analysis is best understood — in the authors' own framing — as hypothesis-generating, and as a prompt to examine why a guideline-supported therapy remains underused in postmenopausal patients with recurrent urinary tract infection, rather than as evidence that vaginal estrogen prevents sepsis, hospital admission, or death.

Disclosures can be found in the published study.

Source: Urology