The investigation began with an unbiased, medication-wide association screen. Using more than two decades of longitudinal records from Leumit Health Services, one of four nationwide healthcare providers in Israel serving roughly 730,000 members, the team compared 14,986 individuals who died before sex-specific median life expectancy thresholds (84 years for women, 77 for men) with 14,986 matched controls who survived beyond them. Each case was matched to a control on sex, birth year, socioeconomic status, year of first electronic health record, and year of a comparable ambulatory visit, and medication exposure was assessed over a fixed ten-year window before an index date. Out of 1,470 screened compounds, seven met the prespecified screening criteria of an odds ratio below 0.5 with a false discovery rate below 0.5 percent. The antimalarial class was the only therapeutic category represented by multiple top entries: mefloquine showed an odds ratio of 0.32 (95 percent confidence interval 0.15 to 0.65) and atovaquone-proguanil an odds ratio of 0.43 (0.27 to 0.69) for appearing in the long-lived group.

Because this first analysis used survival status to define the groups, the researchers treated it strictly as an exploratory, hypothesis-generating screen rather than a causal test. They then moved to a more conventional longitudinal design. They constructed a historical cohort of 3,876 individuals aged 40 to 85 who had received at least one prescription for atovaquone-proguanil or mefloquine, matching each exposed person to 20 unexposed controls on sex, age, socioeconomic status, ethnic sector, and year of first health record. The index date was the first recorded purchase of the drug, and that exact calendar date was assigned to matched controls, who were confirmed to be alive and active members at that date, an approach designed to align time zero and prevent immortal-time bias. Baseline characteristics were broadly comparable, with small differences in body mass index, blood pressure, and smoking status between groups.

In Cox proportional hazards models adjusted for age, smoking status, number of ambulatory visits, and pregnancy history, treated individuals had a hazard ratio of 0.40 (95 percent confidence interval 0.34 to 0.47) for all-cause mortality. Significant reductions were also observed for dementia (hazard ratio 0.52), diabetes mellitus (0.83), heart failure (0.76), myocardial infarction (0.78), hepatic failure (0.64), chronic obstructive pulmonary disease (0.73), gastrointestinal malignancy (0.70), and lung cancer (0.52). For each outcome, individuals with a documented diagnosis before the index date were excluded from the corresponding analysis, and the proportional hazards assumption was checked using Schoenfeld residuals.

Not every signal pointed in the same direction. Treated individuals also showed significantly increased risks of hearing loss (hazard ratio 1.25), Sjogren’s syndrome or dry eye (1.25), and lichen planus (1.54). The authors argue that these consistent increases, rather than undermining the protective associations, may offer a mechanistic clue. These conditions affect tissues that are sensitive to localized immune-mediated damage, and the researchers speculate that pharmacological clearance of intracellular pathogens could trigger localized inflammation at sites of prior colonization while simultaneously reducing systemic inflammation. They emphasize that this explanation remains speculative.

A key concern with antimalarial drugs is that they are prescribed almost exclusively to travelers, raising the possibility of a healthy-traveler effect: people who take international trips may simply be healthier and more health-engaged than those who do not. To address this, the team ran a secondary exploratory screen using a shorter five-year exposure window to find medications with potential antiprotozoal activity that are not linked to travel. This identified nirmatrelvir-ritonavir, the protease inhibitor combination marketed as Paxlovid, which was introduced in 2021 and given free of charge to eligible patients under centrally applied criteria in the Israeli health system. Ritonavir and related HIV protease inhibitors have demonstrated in vitro activity against Toxoplasma gondii, an intracellular protozoan parasite, providing a degree of biological plausibility, although nirmatrelvir-ritonavir has not been formally studied as a treatment for this parasite.

The final stage was external validation in the TriNetX network, a federated research platform drawing on electronic health records from approximately 70 healthcare organizations and more than 120 million individuals, primarily in the United States. The researchers built propensity-score-matched cohorts of people aged 70 to 89, balancing exposed and unexposed groups on age, sex, race, ethnicity, smoking status, diabetes diagnosis, body mass index, and HbA1c. The matched cohorts comprised 24,498 exposed individuals and 24,498 controls for atovaquone-proguanil, 3,400 per group for mefloquine, and 124,681 per group for nirmatrelvir-ritonavir. Follow-up was censored at the last recorded medical encounter, up to roughly 11 years after the index date, and people with a prior diagnosis of each outcome were excluded.

The TriNetX results were consistent with the Israeli findings. All-cause mortality hazard ratios were 0.26 (95 percent confidence interval 0.24 to 0.28) for atovaquone-proguanil, 0.33 (0.27 to 0.40) for mefloquine, and 0.50 (0.48 to 0.52) for nirmatrelvir-ritonavir. Dementia incidence was reduced (hazard ratios 0.48 to 0.64 across medications), as were diabetes diagnoses (0.56 to 0.75), new HbA1c values of 6.5 percent or higher (0.57 to 0.68), kidney disease defined as eGFR below 45 milliliters per minute per 1.73 square meters (0.53 to 0.71), heart failure (0.55 to 0.75), myocardial infarction (0.55 to 0.89), hepatic failure (0.28 to 0.48), COPD (0.45 to 0.69), lung cancer (0.47 to 0.69), gastrointestinal malignancies (0.56 to 0.70), and brain malignancies (0.63 for atovaquone-proguanil and nirmatrelvir-ritonavir). The same adverse signals reappeared: hearing loss (hazard ratios 1.53 to 1.65), Sjogren’s syndrome or dry eye (1.50 to 2.11), retinal tear, and lichen planus.

The authors propose Toxoplasma gondii as a candidate, though unproven, mechanism. The parasite is ubiquitous, with seroprevalence estimated at roughly 20 to 80 percent depending on geography and age, persists lifelong inside human cells, and has been associated in epidemiologic studies with dementia, cardiovascular disease, schizophrenia, diabetes, and adverse population-level health outcomes. Atovaquone is a well-established anti-Toxoplasma agent used in treatment and prophylaxis of toxoplasmosis, and mefloquine has shown activity against the parasite in experimental models. The researchers also report that in a separate matched case-control analysis within their health system, prior mefloquine exposure was less frequent among individuals with positive T. gondii serology (odds ratio 0.45, 95 percent confidence interval 0.27 to 0.74), which they describe as indirect support for an effect on parasite persistence. They caution that these observations are consistent with, but not direct evidence for, a protozoal contribution to aging-related disease.

The limitations are substantial and the authors enumerate them candidly. As an observational study, residual confounding cannot be rigorously excluded, and in the exploratory screen individuals exposed to antimalarials did show slightly healthier baseline characteristics, which may explain part, though not all, of the observed risk reduction. Not all downstream outcome analyses were adjusted for multiplicity, so some associations may represent false positives. The mortality effect sizes are larger than typically seen after short-duration pharmacologic exposures, and alternative explanations include microbiome-mediated effects, immunomodulatory properties of the drugs, differences in healthcare engagement, and unmeasured confounding. Generalizability to low- and middle-income countries, where protozoal exposure prevalence and healthcare access differ substantially, remains uncertain. The researchers frame their conclusion as a hypothesis to be tested: chronic protozoal colonization, possibly involving T. gondii, may be an underrecognized contributor to age-related morbidity, and mechanistic studies, serologic investigations, microbiome analyses, and prospective randomized trials will be needed to determine whether the observed associations reflect causal effects.

Subject of Research: Pharmacoepidemiologic associations between antiprotozoal drug exposure and reduced long-term mortality and age-related morbidity

Article Title: Antiprotozoal drug exposure and reduced long-term mortality and morbidity: a multi-cohort pharmacoepidemiologic study

Article References: Israel, A., Weizman, A., Israel, S., Ashkenazi, S., Ruppin, E., Vinker, S., Magen, E., & Merzon, E. (2026). Antiprotozoal drug exposure and reduced long-term mortality and morbidity: a multi-cohort pharmacoepidemiologic study. Frontiers in Pharmacology, 17, 1861376. https://doi.org/10.3389/fphar.2026.1861376

Image Credits: AI Generated

DOI: 10.3389/fphar.2026.1861376

Keywords: antiprotozoal drugs, atovaquone-proguanil, mefloquine, nirmatrelvir-ritonavir, Toxoplasma gondii, longevity, pharmacoepidemiology, electronic health records, drug repurposing, all-cause mortality, aging-related disease, observational study