Look at 148 Week Data for Alzheimer Drug Anavex 2-73.

The data presented for the 148 weeks period gave readings for 26, 5, 96 and 148 weeks adjusted for the SIGMAR1 and COMT gene variations from the pool of patients previously reported as High Concentration.  The data points are separately given for two measures; ADCS-ADL and MMSE. One has to remember that ADCS-ADL is on 70 points scale and MMSE is on 30 point scale so the drops at 148 weeks are commensurate.  Here is a graph of the deltas from the base:

Our premise is to identify patients by matching averages by trial or inspection to individual patient scores as given in previous data releases.  This might be fool’s errand but it might be also instructive yet not conclusive to see either discrepancies or even convergence of averages.

The presentation of the 148-week data listed the ADCS-ADL first as for the patients selected through the filter of the full list of co-variants in MMRM-LME model (previous ones, plus SIGMAR1 and COMT).  CTAD November 2017 lists patients data for 57 weeks on the slide 24:

Out of the 9 patients, 8 are listed, patient 2002 has to be discarded as his drop is over 20 points making any average out of agreement with 148-week data.  It is equally possible that the pool can be just two patients or all 9.  Upon the knowledge that the 80% of the general population is carrying genes assuring therapeutic response to Anavex 2-73 the percentage of such population within the pool of 9 patients in the cohort with High Concentration can indicate the degree to which therapeutic effect in Phase 2a can be replicated in the general population.  The average for all other patients is:

(7.8+3.7+2.0-.9+.9-1.9-4.0)/7=1.09  compare this to the reading of 1.0 at 57 weeks on 148-weeks data and we come near to the mark.

Inspection of the performance of patient 1013 reveals drop in ADCS-ADL score at 96 weeks of -14 points. This can be found on the slide 29 of CTAD November 2017 presentation.  Such a big drop would require the other patients score to go up in order to keep the average at -1.0 reading.  If we remove patient 1013 then:

(7.8+3.7+.9-.9-1.9-4.0)/6=0.9333 which is even closer to the reading of 1.0 (148-weeks data).

Let’s see if the MMSE data confirm one or the other pool of patients.

For seven patients this is:

(3.7+2.8+2.2+3.8+.1-2-6.2)/7=0.63  vs. the reading of 0 for adjusted to MMRM-LME model.

The same for six patients:

(2.2+2.8+3.8+0.1-2.0-6.2)/6=.12  This gives a closer match to 0 than the previous seven patients average.

Out of the pool of 9 patients, 3 are excluded.  Only patients 1013 performs well and then declines which is statistically within probabilities of the natural course of the disease.  Patients 1002 and 2002 are deteriorating quickly and to such extent that the given averages cannot include them.  The inclusion of the SIGMAR1 and COMT gene variation whittles the pool to six patients, that is 60% of the initial population which is congruent with the 80% of the general population to be with the “right” genes.  The data points are for averages so they include those who declined and those who improved or kept their scores unchanged. In conclusion, the p-numbers given with the 148-weeks data leave little doubt that the data is not a fluke.

The slide 29 does not include all patients in the pool of six so we can not try to have a second data point to assess the validity of the presented conclusions.

Due to some problems with my health, this could be my last post.  I thank everyone who read these rather simplistic posts.  Thank you, again.