Have to take with a grain of salt but I put all the data from ADAS COG and CDR SB into chat GPT with the 'n' from week 0, 12, 24, 36 and 48 and the score at each point (as far as I could judge with the naked eye based on the chart) and explained the dropouts due to dizzines asked if it could see any clear evidence of surivorship bias with the dropouts helping the drug look effective. Here was reply:
Key Considerations:
Dropout Rates and Timing:
The dropouts occurred more heavily in the 50mg group early on, which aligns with reported side effects like dizziness.
The completion rates for 30mg (70%) and placebo (74%) were fairly close, reducing concerns about severe dropout-related bias in that dose.
Efficacy Trends Over Time:
Both ADAS-Cog and CDR-SB show gradual improvements in the drug groups over time, especially past Week 24. This trend is steady rather than marked by sudden efficacy gains linked to periods of high dropout, which weakens the likelihood of survivor bias.
Notably, the efficacy curves for both 30mg and 50mg gradually diverge from placebo, especially by Week 48, suggesting a consistent effect that isn’t clearly tied to dropout phases.
Cross-Referencing Dropout Phases and Efficacy Changes:
There is no strong correlation between quarters with higher dropout rates and subsequent improvements in efficacy scores. For instance, in the 50mg group, the period from Week 12 to Week 24 saw a higher dropout rate, yet the efficacy improvement remained gradual rather than sharply increasing.
Conclusion:
After this thorough analysis, I could not identify a clear pattern indicating that the dropouts artificially inflated the efficacy results. The improvements in the drug groups appear consistent over time and are not strongly linked to dropout timing or rates.
Market Data 
Markets 
AI
