Time elapsed from symptom onset was recorded for over 99% of the included participants, thus defining reported disease duration. The δALSFRS (points decline per month) could therefore be calculated as either drop in ALSFRS from 40 divided by disease duration, or by subtraction of time-separated ALSFRS assessments divided by the inter-visit time-interval. Calculated δALSFRS was then used to extrapolate back to the date at which ALSFRS = 40, i.e. no disability.
This approach was then validated using FVC data and ALSFRS-R scores. 4168 records in PRO-ACT include at least two FVC measurements separated by at least one month. The smaller portion of PRO-ACT records with ALSFRS-R was supplemented by 217 individual longitudinal data records from the Lithium Carbonate in ALS (LiCALS) study, resulting in 1709 individual records suitable for analysis, that demonstrated incremental disability over a median time-interval of 11.0 months.
Within the PRO-ACT database 1863 individual records included mortality data, and from this subset only 464 remained alive at the last census. Hazard curves were constructed from both the entire population and the mortality subset to represent time-dependent risk of death or significant disability (defined as ALSFRS < = 21, the median final assessment ALSFRS across participants).
Trial drop-out mitigation methods appraised included the re-assignment of missing values with either (1) imputed values of ALSFRS based on linear extrapolation using δALSFRS calculated from the first assessment to the last available assessment, or (2) ALSFRS values carried forward unchanged from the last available assessment for that participant, with or without (3) assignment of ALSFRS = 0 if the participant died prior to the planned assessment.
Data were analysed using Matlab and SPSS 21. Paired t-tests compared alternative measures for individual patients and Spearman’s rho was used for correlations to minimize the effect of non-normative data. Mean value was followed by standard deviation in parenthesis.