Does the slope of UPDRS vs. time show neuroprotection?

I think I understand where you’re coming from about the slopes but it’s hard to be sure by email. [Note: Dr. Schwarzschild has approved this posting of his email response to Dr. Grove, who originally sent his question to Dr. Schwarzschild via email.]

 So I’m pasting below the relevant figures from ADAGIO and ELLDOPA to illustrate my thoughts and make sure we’re on the same page, [and quote from your post below.]

AG: I read your posting on PDOnlineResearch on Adagio, etc. with interest.  I wonder if we are overlooking a fairly basic indicator of disease modification (nay, neuroprotection);  the slope of the UPDRS vs. time lines.  Eyeballing the Rasagiline lines, in comparison to the slope prior to administration of Rasagiline, shows in all cases a lower slope.

Although I can see how it might look that way, I don’t think it holds up for a couple reasons.

1) Confounding symptomatic effect after the switch to active drug

To confirm we’re looking at the same lines, I  see two cases in which the slopes before and after rasagaline can be compared (Fig 1)Fig 1.ADAGIO -- Changes in Scores on the Unified Parkinson's Disease Rating Scale (UPDRS) in the Four Study Groups.The mean (±SE) change from baseline in the UPDRS score in the efficacy cohort for the second and third primary end points for patients receiving rasagiline at a dose of 1 mg per day (Panel A) and those receiving 2 mg per day (Panel B) are shown. The dashed lines indicate placebo, and the solid lines indicate rasagiline. (modified from Olanow et al., 2009):  

• First, in the “Delayed Start (placebo-rasagiline)” group at 1 mg/day (upper line in A); change occurs at week 36.
• Second, in the “Delayed Start (placebo-rasagiline)” group at 2 mg/day (upper line in B); change occurs at week 36.

Are there other cases you were referring to?

For each of the two cases I drew a straight line for the placebo phase (red in A; purple in B).

• Note while one might be tempted to draw the placebo slope just between weeks 12 and 36, the truer placebo rate is from 0 to 36. This is because though there’s a well known placebo improvement effect apparent between weeks 0 and 12 (when subjects feel/hope their getting a better on an experimental new drug). This effect dissipates and and the rate of change then seems to accelerate for the next 12 weeks before finally ‘catching up’ to the overall placebo phase slope between weeks 24 and 36. So I include the entire placebo phase from baseline time point to 36 week in drawing the lines.

For each of the two cases I copied the exact same (red or purple) line/slope, and superimposed it over the slope of the next phase (on rasagiline) AFTER the wash-in period when the symptomatic effect is manifesting as a reduction in the UPDRS.

• Most of this symptomatic improvement is seen as a rapid drop in UPDRS, but the wash-in of the symtomatic effect can take longer weeks to months.
• Note the same line/slope fits the data during the placebo phase and at the end of the treatment phase (after a symptomatic effect is presumably stable).  

2) Drop out of subjects between the placebo and active drug phases.

This is a subtler reason why one can be fooled by simply comparing before and after the starting rasagiline.  As you can imagine after the placebo improvement phenomenon dissipates and the disease continues to progress some subjects progress so much that they cannot stay off real treatment and they had to terminate the study. For the 1mg/day group there were 30 subjects (10% of 300 original who started the placebo treatment) who never made it to rasagiline and so couldn’t be included in the data after 36 weeks. These missing data will tend to be from people who have more rapidly progressive disease (i.e., such that they can’t hold out until 36 weeks when they finally get symptomatic relief by being put on rasagiline.  That means those who progress more rapidly may not be included in the line on the right, artificially improving (flattening) its slope.

For both of these reasons one should not conclude just from the before and after slopes of the “Delayed-start” arms of the study that there was an improvement in rates of decline.  Accordingly, the investigators did not include such comparisons among their 3 criteria for ‘disease modification’ (Fig. 1 in ADAGIO).  The comparison to the control group is absolutely critical for all 3 of these more meaningful tests.

My point in my posting was that when one looks at these more valid criteria there is still no conclusive evidence of ‘neuroprotection’ by rasagiline in ADAGIO that is any more compelling than the evidence for neuroprotection by levodopa in ELLDOPA study. On the contrary...

• At 2 mg/day rasagiline there was clearly no disease modification benefit.
• At the 1 mg/day dose there is some evidence but it’s not compelling as above and it’s relatively small compared to the apparent protective effect of levodopa using the matching criteria applied to ELLDOPA.  This is particularly the case for the slope comparisons in the first 36-40 weeks in both studies.  See below in the composite figure I made from the relevant figures in ADAGIO and ELLDOPA [Fig. 2].
  • Note the splay between the expected slope for ‘no disease modifying effect’ of rasagiline (i.e., the parallel red slope of the placebo line after 12 weeks superimposed on the rasagline treatment line below in Panel A) and the actual (blue line) slope during this period. It is tiny compared to the corresponding splay between red and blue slopes in the ELLDOPA graph at bottom.

But the big caveat for this disease modification criterion (#1 in ADAGIO Fig. 1) in both ADAGIO and ELLDOPA is that is that it is based on the weak/false assumption that symptomatic effect of rasagaline or levodopa is fully established within the first 12 weeks. In ADAGIO’s Methods section the authors actually noted that criterion #1 is based on this assumption, but they did not acknowledge that it appears false. So bottom line we really don’t know if either levodopa or rasagiline slows down the disease.  My concern is that although few people believe that levodopa is neuroprotective, many people will wrongly cite ADAGIO as showing rasagiline is.

Fig 2. Comparing ADAGIO to ELLDOPAFig 2. Comparing ADAGIO to ELLDOPATop figure taken from the ADAGIO study (Olanow et al., 2009: see legend in Fig 1 above); Bottom figure taken from ELLDOPA study (Fahn, et al., 2004: original figure legend follows: "Changes in Total Scores on the Unified Parkinson's Disease Rating Scale (UPDRS) from Baseline through Evaluation at Week 42.The changes in subjects treated with levodopa at different doses or with placebo were determined on the basis of the total scores on the UPDRS. The scores were obtained by blinded treating investigators who performed the evaluation before the morning dose of the daily dose of the study drug. The points on the curves indicate mean changes from baseline in the total scores at each visit. Improvement in parkinsonism is represented by lower scores, and worsening by higher scores. Negative scores on the curves indicate improvement from baseline. The bars indicate the standard error.")

AG: By contrast, the ELLDOPA lines within the acuity of my eyes, all have the same slope, suggesting no modification of progression.

As above, if anything I see the slope for 600 mg/day of levodopa is flattened/improved (certainly between 9 and 40 weeks [blue line above], and even between 24 and 40 weeks) more than for the ‘best’ (1 mg) dose of rasagiline.

 
AG: Why would that not be a sufficient indication that Rasagiline lowers the rate of progression and ipso facto, modifies the disease progression?

Sorry to have gone on but I hope I’ve made my points well enough to convince you that the slope comparisons you suggest would not be sufficient, or at least that it’s not so simple to rely on a change in slope in one group for an interpretation of neuroprotection.  But please let me know if I’m missing your point.