Showing disease modifying effects requires an objective method for analyzing UPDRS versus time curves

The different views of Drs. Grove and Schwarzschild on whether or not slopes of the UPDRS versus time curves in the ADAGIO and ELLDOPA trial show disease modifying effects demonstrates that the analysis of these kind of  trials requires a more objective method than visual interpretation of treatment arm average curves. This method should allow separation of the time course of the disease from the time course of a symptomatic effect (fast onset of action, which will washout after stopping treatment) versus that of a disease modifying effect (slow onset, which will persist after stopping treatment). Recently, a quantitative model for Parkinson’s disease has been applied to analyze the DATATOP cohort (PSG, 1989), which confirmed the symptomatic action of the dopaminergic agents levodopa, bromocriptine and pergolide and provided evidence that these drugs also have disease modifying effects (Holford et al., 1996). More recently in a MJFF funded study, we have confirmed the symptomatic and disease modifying effect of levodopa by analyzing the ELLDOPA data with the same quantitative model used for the DATATOP analysis. In this analysis we show that the slopes of the UPDRS versus time curves attributable to disease progression in the active treatment arms are significantly lower compared to the placebo arm.

We think that analyzing the ADAGIO data with a quantitative model for Parkinson’s disease will provide objective insights in the discussion of whether or not rasagiline is disease modifying. This method takes many of the issues addressed by Dr. Schwarzschild into consideration e.g. the confounding symptomatic effect after switching to the active drug is explicitly taken into consideration in the analysis. In addition, the method uses all available data up to a patient’s drop-out and does not rely on imputation methods, which makes this method less sensitive to bias from informative drop-out, such as patients with higher disease progress who are more likely to drop-out.

We hope that the continuing discussion on the possible disease modifying effects of anti-Parkinsonian treatments will result in the willingness to use longitudinal quantitative methods both in the design and analysis of trials. We have recently shown in a simulation study that a delayed start design is less powerful to show disease modifying effects compared to a washout design (Ploeger and Holford, 2009). We believe it is time for clinical investigators and statisticians involved in Parkinson’s disease research to become familiar with newer methods that do not ignore the crucial role of time in understanding the response to treatment.

Parkinson Study Group, DATATOP: a multicenter controlled clinical trial in early Parkinson's disease. Parkinson Study Group. Arch Neurol, 1989. 46(10): p. 1052-60.

Holford, N.H., et al., Disease progression and pharmacodynamics in Parkinson disease - evidence for functional protection with levodopa and other treatments. J Pharmacokinet Pharmacodyn., 2006. 33(3): p. 281-311.

Ploeger, B.A. and N.H. Holford, Washout and delayed start designs for identifying disease modifying effects in slowly progressive diseases using disease progression analysis. Pharm Stat, 2009; 8: 225-38