Next, the investigators sought to compare patients by standardizing their follow-up time. To do this, they calculated what they called the ARMSS-rate by dividing each patient’s ARMSS integral by the number of years of follow-up. The ARMSS-rate offers a “snapshot of disease severity and progression,” said Dr. Ramanujam. When the researchers compared ARMSS-rates at 2 years and 10 years for each patient, they found that the measure was “extremely stable over time and strongly correlated with future disability.” The correlation improved slightly when the researchers compared ARMSS-rates at 4 years and 10 years.
The investigators then categorized patients based on their ARMSS-rate at 2 years (e.g., 0 to 1, 1 to 2, 2 to 3). When they compared the values in these categories with the median ARMSS-rates for the same individuals over the subsequent 8 years, they found strong group-level correlations.
To analyze correlations on an individual level, Dr. Ramanujam and colleagues examined the ability of different metrics at the time closest to 2 years of follow-up to predict those measured at 10 years. They assigned the value 1 to the most severe quartile of outcomes and the value 0 to all other quartiles. For predictors and outcomes, the investigators examined ARMSS-rate and the integral of progression index, which they calculated using the integral of EDSS. They also included EDSS at 10 years as an outcome for progression index.
For predicting the subsequent 8 years of ARMSS-rates, ARMSS-rate at 2 years had an area under the curve (AUC) of 0.921. When the investigators performed the same analysis using a cohort of patients with MS from British Columbia, Canada, they obtained an AUC of 0.887. Progression index at 2 years had an AUC of 0.61 for predicting the most severe quartile of the next 8 years. Compared with this result, ARMSS integral up to 2 years was slightly better at predicting EDSS at 10 years, said Dr. Ramanujam. The progression index poorly predicted the most severe quartile of EDSS at 10 years.
The main limitation of the ARMSS integral and ARMSS-rate is that they are based on EDSS, he added. The EDSS gives great weight to mobility and largely does not measure cognitive disability. “Future metrics could therefore include additional data such as MRI, Symbol Digit Modalities Test, or neurofilament light levels,” said Dr. Ramanujam. “Also, self-assessment could be one area to improve in the future.”
Dr. Ramanujam had no conflicts of interest to disclose. He receives funding from the MultipleMS Project, which is part of the EU Horizon 2020 Framework.