# Derivation and external validation of community-acquired pneumonia subphenotypes in Southeast Asia: a secondary analysis of prospective cohort studies

## TL;DR

This study identifies two subtypes of pneumonia in Southeast Asia, which differ in severity and outcomes, and validates a model to classify these subtypes.

## Contribution

The study derives and externally validates pneumonia subphenotypes specific to Southeast Asia, and explores their relevance in a U.S. cohort with COVID-19.

## Key findings

- Two CAP subphenotypes (CAP1 and CAP2) were identified in Thailand, with CAP1 associated with higher mortality and inflammation.
- A four-variable model accurately classified subphenotypes and showed consistent results in external validation.
- Subphenotypes in a U.S. COVID-19 cohort revealed treatment effect heterogeneity with corticosteroids.

## Abstract

Identifying pneumonia subphenotypes in understudied populations can advance equitable personalized medicine for pneumonia care. We aimed to derive and validate subphenotypes of patients presenting with community-acquired pneumonia (CAP) in Southeast Asia.

This secondary analysis included three prospective cohorts conducted between March 2013 and April 2020. First, we performed latent class analysis to identify subphenotypes using clinical and laboratory variables in a prospective cohort of adults hospitalized with CAP in northeastern Thailand. Next, we compared clinical and biological features between the subphenotypes and then developed a parsimonious classifier model (PCM) for accurate subphenotype assignment. We then validated the accuracy of PCM subphenotype assignment in an external, multinational prospective cohort of patients hospitalized with CAP in Southeast Asia. Finally, in an exploratory analysis, we used the PCM to assign pneumonia subphenotypes in a prospective cohort of patients hospitalized with COVID-19 in the United States of America and evaluated a heterogeneity of treatment effect with corticosteroids.

Among 953 CAP patients in the Thai derivation cohort, we identified two subphenotypes: CAP1 (141, 15%) and CAP2 (812, 85%). We observed greater respiratory failure, 28-day mortality, inflammatory cytokines and metabolic derangements among CAP1 patients. A four-variable PCM discriminated subphenotype assignment in bootstrap internal validation (optimism-corrected C-statistic 0.97). In the Southeast Asian external validation cohort, CAP1 and CAP2 subphenotypes assigned by the PCM shared similar differential clinical features and outcomes with the Thai derivation cohort. In the cohort of patients with COVID-19, CAP1 and CAP2 subphenotypes assigned by the PCM differed by key clinical characteristics and revealed an interaction between corticosteroid treatment and mortality (P = 0.002).

In Southeast Asia, CAP subphenotypes are associated with distinct outcomes, inflammatory profiles, and metabolomic signatures. These subphenotypes may represent unique targets for future CAP interventional trials.

Supported by US NIH awards T32HL007287, F32HL168809, K08HL157562, U01AI115520, R01AI137111, R01GM114029, R21AI173435, R01HL113382, the 10.13039/100010269Wellcome Trust grants 090219/Z/09/Z, 101103/Z/13/Z, 106680/B/14/Z, and 106698/B/14/Z, the US National Cancer Institute, 10.13039/100000002National Institutes of Health HHSN261200800001E, and 10.13039/100005815Firland Foundation award 20220012. For the purpose of Open Access, the author has applied a CC BY public copyright license to any Author Accepted Manuscript version arising from this submission.