Lecanemab doesn’t appear to work in women - Part 2

A previous post outlined my reservations about the apparent lack of efficacy of lecanemab in women.

In a Letter to the Editor of the New England Journal of Medicine published a few days ago, myself and Professor Alvaro Pascual-Leone of Harvard communicate these concerns to the academic community.

Our fundamental point is simple. There appears to be a repeating pattern in the trial data, whereby clinical outcomes are statistically significant in men but not so in women.

Below is the original paper’s data (Figure S1) for the primary end point of the trial, the Clinical Dementia Rating scale sum of boxes (CDR-SB) with my annotation in pink.

Adapted from Figure S1 of original NEJM paper.

For those unused to these forest plots, if the confidence interval of a particular comparison crosses zero then the difference between groups is considered statistically non-significant. Where the confidence interval does not cross zero then the comparison is significant.

Clearly, for CDR-SB the result was statistically significant in men (estimated difference between treatment vs placebo was 0.73 points). Equally clear is that the result was non-significant in women (estimated mean difference 0.20 points). 

On the basis of this trial, it is reasonable to conclude that the overall clinical impact of lecanemab as frequently cited (0.45 CDR-SOB point difference) arises from a real albeit weak therapeutic effect in men but negligible effect in women.  

If lecanemab doesn’t work in women that is a big deal. Unfortunately, the response by the authors published alongside our Letter fails to shed any light on the matter.

Firstly, they state the trial was not powered to analyze individual subgroups. But subgroup analysis was pre-planned as per their published protocol, and hence the subgroup results merit consideration.

One explanation for the null result in women is a sampling error, but the size of the subgroups was healthy (n >400 for men and women) and the authors accept at face value that point estimates in women were lower than in men. If you’re going to interpret point estimates why not also consider their degree of precision, the confidence interval?

The alternative is that lecanemab is actually not effective in women. This rises in likelihood given null findings in women were also seen in all other clinical secondary endpoints in Figures S2-S4 (ADAS-COG14, ADCOMS, ACDS-MCI-ADL).

Secondly, they assert, “The subgroup analyses indicate that lecanemab performed better than placebo with respect to all clinical, biomarker, and quality-of-life outcomes among women, findings that were consistent with the overall efficacy.”  They seem to be arguing that any numerical superiority in the treatment arm in women is sufficient, irrespective of the statistics. That is untenable and potentially dangerous.

To put it bluntly if lecanemab doesn’t work in women it would be unethical to supply it to women.

Recall this costly immunotherapy comes with substantive risks, including death. In biotech innovation such risk can be accepted in the context of a reasonable chance of clinical benefit, and should be avoided when those benefits are uncertain or unknown.

From a scientific perspective there are additional concerns. Why doesn’t it work in women? What is the biological basis for the decisive impact of sex on the mechanism of action? And is this specific to lecanemab or a problem for the whole class of anti-amyloids?

Engaging with these issues in an open and meaningful way is now critical for the credibility of this new drug, and I dare say, for the field at large.