Methods

Mixed effects model

In this study, data analysis was complicated by the high patient withdrawal rate (n = 30). In a multi-group clinical trial, patients lost to follow-up may not be representative of those randomized to each group. Thus, analyzing only the collected clinical data can introduce serious biases. Censorship of observation at time of dropout is only valid if loss to follow-up is independent of outcome, but the precise reason for a patient withdrawing from a study is often unavailable to investigators, as was the case in this research. To overcome these potential biases, a mixed effects model was employed in analysis of the topiramate placebo cohort dataset. This model is a well established technique that weights each patient according to their number of collected data points.1Data derived from a patient with measurements at six time points is considered more robust than data obtained from a patient with only two measurements. It is interesting to note that statistical analysis of the topiramate placebo cohort excluding data from patients who withdrew consent (completers only) was similar to the findings reported in this study (data not shown).

  1. LairdNM, Ware JH. Random-effects models for longitudinal data. Biometrics 1982; 38(4):963-974.

Results

Table (E)T-1. Annual rate of decline and variability of functional outcome measures of the topiramate placebo group over the 12-month study period.

Outcome measure / Mean rate of decline per year, Mean (SE) / Standard deviation within subjects,
Mean (SE)* / Standard deviation between subjects, Mean (SE)†
FVC (%) / -29.3 (2.8) / 5.9 (0.3) / 22.3 (2.3)
ALSFRS total score (units) / -11.1 (1.0) / 1.6 (0.1) / 8.2 (0.8)
MVIC arm Z score (units) / -0.91 (0.09) / 0.17 (0.01) / 0.70 (0.07)
MVIC grip Z score (units) / -1.02(0.10) / 0.20 (0.01) / 0.79 (0.08)

*Standard deviation within subjects reflects the extent to which a patient’s functional outcome measure varies from the rate of decline over the 12-month period.

†Standard deviation between subjects reflects the extent to which the total cohort varies from the overall rate of decline over the 12-month period.

Table(E)T-2. Number of patients assessed at each time point during the study according to outcome measure.

FVC% / ALSFRS / MVIC arm / MVIC grip
Baseline visit / 97 / 97 / 97 / 95
Month 1 / 89 / 93 / 92 / 91
Month 3 / 80 / 83 / 79 / 80
Month 6 / 67 / 74 / 64 / 66
Month 9 / 56 / 62 / 56 / 58
Month 12 / 48 / 51 / 48 / 48

Figure (E)F-1. Survival of ALS patients enrolled in the placebo group of the topiramate study according to their rate of decline of ALSFRS in the first three months of study*

*log-rank test p < 0.0001

Figure (E)F-2. Survival of ALS patients enrolled in the placebo group of the topiramate study according to their rate of decline of MVIC grip Z score in the first three months of study*

* log-rank p = 0.49

Appendix (E)A-1

The authors thank:

The North East ALS (NEALS) Consortium:

Stanley Appel (BaylorCollege of Medicine)

George Baquis (BayStateMedicalCenter)

Jerry Belsh (RobertWoodJohnsonMedicalSchool)

Walter Bradley (University of Miami)

William David (HennepinCountyMedicalCenter)

Peter Donofrio (WakeForestUniversity)

Kevin Felice (University of Connecticut)

Terry Heiman-Patterson (DrexelUniversity)

Steven Novella (YaleUniversity)

Robert Pascuzzi (IndianaUniversity)

Erik Pioro (Cleveland Clinic Foundation)

Jeffrey Rosenfeld (CarolinasMedicalCenter)

Jeffrey Rothstein (JohnsHopkinsUniversity)

James Russell (Lahey Clinic)

George Sachs (Rhode IslandHospital)

Jeremy Shefner (SUNYUpstateMedicalUniversity)

Rup Tandan (University of Vermont)

And theGeneral Clinical Research Centers Program, Massachusetts GeneralHospital (Grant Number M01-RR-01066)

Corresponding email address: