Should doctors prescribe lecanemab (Leqembi) to women? The answer, given available evidence, is probably No

Data from the CLARITY tri­al ear­li­er this year was sup­posed to be the crown­ing glo­ry of the amy­loid hypoth­e­sis, vin­di­ca­tion for pro­po­nents of this long-held but much-maligned the­o­ry of Alzheimer’s disease.

Yet the results left many feel­ing under­whelmed, and even the study authors noncommittal.

The CLARITY tri­al has many admirable fea­tures. It recruit­ed close to 1800 peo­ple from around the world, pret­ty bal­anced between women and men. While the major­i­ty were white, 17% of the cohort was Asian and 12% Latino.

Choice of pri­ma­ry and sec­ondary out­comes were impec­ca­ble. The pri­ma­ry out­come was the Clin­i­cal Demen­tia Rat­ing sum of box­es (CDR-SB), a score­card of sorts rat­ed by a clin­i­cian across the domains of mem­o­ry, ori­en­ta­tion, prob­lem solv­ing, com­mu­ni­ty affairs, home duties and per­son­al care. Sec­ondary out­comes includ­ed stan­dard mea­sures of glob­al cog­ni­tion, dai­ly func­tion, as well as bio­mark­ers in the brain, from the CSF and blood.

So what hap­pened? In short, the group who received fort­night­ly infu­sion of amy­loid anti­bod­ies dete­ri­o­rat­ed by 1.21 points on the CDR-SB whilst the place­bo group dete­ri­o­rat­ed by 1.66 points, the aver­age dif­fer­ence being 0.45 points. That’s right, less than half a point on the CDR-SB! Less than the small­est change prac­ti­cal­ly pos­si­ble at an indi­vid­ual lev­el and only cal­cu­la­ble by com­par­ing group averages.

At the same time, par­tic­i­pants’ PET scans showed mas­sive reduc­tions of amy­loid. Almost 75% of base­line cere­bral fib­ril­lar amy­loid was removed; the major­i­ty of indi­vid­u­als mov­ing from amy­loid “pos­i­tive” to amy­loid “neg­a­tive”. Whilst amy­loid removal has been seen in pri­or tri­als using dif­fer­ent anti-amy­loids, this is the most con­vinc­ing bio­log­i­cal result yet and indeed dif­fi­cult to imag­ine a bet­ter brain imag­ing result.

Hav­ing said this, the most trou­bling aspect was no sta­tis­ti­cal­ly sig­nif­i­cant result–let alone clin­i­cal­ly mean­ing­ful outcome–in women on the pri­ma­ry out­come or any clin­i­cal sec­ondary out­come (Sup­ple­men­tary Fig­ures 1–4). In oth­er words, the head­line read­out was dri­ven by pos­i­tive out­comes in men, and it is beyond remark­able this was not dis­cussed in the paper. This can­not be explained by a lack of sta­tis­ti­cal pow­er, and sug­gests a more fun­da­men­tal bio­log­i­cal inter­ac­tion at the lev­el of ther­a­peu­tic poten­cy or mech­a­nism of action. This will need a lot of care­ful analy­sis and fur­ther inves­ti­ga­tion to sort out.

Recall these ‘meh’ results are in the face of not incon­sid­er­able or incon­se­quen­tial risk of side effects. 26% of those receiv­ing their fort­night­ly IV dose expe­ri­enced an infu­sion reac­tion, and a 21.5% inci­dence of ARIA, a form of brain imag­ing abnor­mal­i­ty relat­ed to bleed­ing and oede­ma unique to those treat­ed by anti-amy­loids. Whilst most turn out to be harm­less, about a quar­ter are clin­i­cal­ly sig­nif­i­cant, includ­ing stroke, and in any case ARIA abnor­mal­i­ties need expert diag­no­sis and management.

Alto­geth­er, lecanemab is the most potent anti-amy­loid to date but like all its ther­a­peu­tic cousins has not been able to show a clin­i­cal­ly impor­tant dif­fer­ence com­pared to place­bo, espe­cial­ly 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 CDR-SB with my anno­ta­tion in pink. Click to expand the image:

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 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 high inci­dence of ARIA and even death. In biotech inno­va­tion such risk may 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.

– Pro­fes­sor Michael Valen­zuela is a Vis­it­ing Pro­fes­sor at the Cen­tre for Healthy Brain Age­ing at the Uni­ver­si­ty of New South Wales, Aus­tralia, mem­ber of the Clin­i­cal Con­sor­tium on Healthy Age­ing of the World Health Orga­ni­za­tion and Co-Founder and CEO of Skin2Neuron Pty Ltd. This arti­cle is an edit­ed com­bi­na­tion of two pre­vi­ous blog posts by him at www.skin2neuron.org.

News in Context:

• CMS: anti-amy­loid drug Leqem­bi (lecanemab) doesn’t meet the “rea­son­able and nec­es­sary” stan­dard required for wider Medicare coverage