Longitudinal Viral Load Clustering for People With HIV Using Functional Principal Component Analysis

Area of Science:

• Immunology
• Virology
• Data Science

Background:

• Longitudinal viral load (VL) monitoring is crucial for managing HIV status.
• Limited research exists on clustering historical/longitudinal VL measures.
• Analyzing VL patterns offers deeper insights than aggregated measures.

Purpose of the Study:

• To classify longitudinal VL patterns in people with HIV (PWH) using functional data clustering.
• To characterize identified clusters by demographics, comorbidities, social behaviors, and CD4 count.
• To inform tailored treatment and intervention strategies for PWH.

Main Methods:

• Adult PWH diagnosed 2005-2015 in South Carolina with >=5-year follow-up were analyzed.
• Functional Principal Component Analysis (FPCA) was used for classification, outperforming other methods.
• ANOVA compared VL characteristics, demographics, comorbidities, substance use, and CD4 count across clusters.

Main Results:

• FPCA identified four distinct VL patterns: long-term viral suppression (17.3%), short-term viral suppression (29.8%), suboptimal viral suppression (28.3%), and viral failure (24.9%).
• The viral failure group showed higher mean VL and lower mean CD4 counts.
• Demographic and clinical characteristics varied significantly across the four identified clusters.

Conclusions:

• Functional data clustering effectively distinguishes distinct viral profiles in PWH.
• Understanding these distinct patterns is vital for developing targeted interventions.
• Personalized treatment approaches are essential for optimizing outcomes for all PWH.