International application of an optimized harmonization approach for longitudinal cognitive data in people with HIV

We previously developed a refined longitudinal data harmonization method to address the challenge of nonoverlapping cognitive tests across cohorts, successfully harmonizing data from 5 large-scale US HIV studies. Building on this harmonized data set, we now aim to apply this method to an additional US HIV study and cognitive data from HIV studies in China, India, and Uganda. This effort will result in a more comprehensive data set with a larger, internationally diverse sample that includes both people with HIV and people without HIV.

The new cohorts to be harmonized included cognitive tests that did not fully overlap across studies, a challenge for traditional harmonization methods. We applied our refined approach, designed for scenarios without direct test linkage. In the Uganda cohort, where a key method assumption was violated, we implemented targeted adjustments.

The harmonized cognitive domain scores were consistent across cohorts and strongly correlated with raw or log-transformed cognitive test data (eg, timed outcomes). These scores preserved key patterns of variation observed in the raw data for key demographics-such as age, education, and race-and maintained age-related longitudinal trajectories of cognitive performance derived from all participants' visits.

The resulting harmonized data set includes 18,270 participants across multiple countries, significantly enhancing its diversity and utility. It lays the groundwork for developing normative data and conducting more robust analyses to address critical neuro-HIV research questions. This study also demonstrates the adaptability of the refined harmonization method in integrating new data and accommodating methodological challenges.

People with HIV (PWH) often face a variety of cognitive challenges, but these issues can look different for each person. As different studies use different tests to measure cognitive abilities, it is difficult to combine the results from multiple studies and draw clear conclusions. In our previous work, we developed a refined method to harmonize data from 5 large US-based HIV neuro studies. Such method could handle the scenarios where nonoverlapping cognitive tests are used in certain domains across different studies. We now aim to include additional cohorts from the United States, China, India, and Uganda. Because these new cohorts also use nonoverlapping cognitive tests in certain domains, we applied our developed approach to harmonize the new data into our previously harmonized data. Our refined method created "harmonized scores" for cognitive abilities that closely matched the original test results. These scores captured differences related to age, education, and other factors while preserving how each person's cognitive abilities changed over time. By using this method to combine new data with existing data, we were able to create a more comprehensive and diverse data set. This will aid researchers to better understand the wide range of cognitive changes in PWH, leading to stronger, more inclusive studies on the impact of HIV on cognition.