Monday, December 19, 2022

Curing Alzheimer's? Lies, damn lies and statistics

If you flip a coin three times and all three times it shows heads, you would not be shocked. If you flipped it 50 times and it came up heads every time you would ask for a new coin! Statisticians have ways of describing the odds that something happens by chance, and this type of calculation is broadly used in medical trials.

When treatment A gives a different outcome than treatment B, this is generally accompanied by a “P value,” which expresses the likelihood that the difference was purely by chance. The commonly accepted metric that there is a real difference between two arms of a trial is a P value less than 0.05. If P<0.05, this says there is less than a 1 in 20 chance the results were not really different. The lower the P value, the greater the chance the results were not coincidental.

While many researchers worship at the altar of P<0.05, I urge caution. If you include very large numbers of patients, even small differences in outcome can be “statistically significant,” when the difference is meaningless to patient outcomes.

Let’s say you treat 10 patients with an aggressive cancer with A and 10 with B. After 3 years, all the patients given A are dead and 9 of those given B are alive. You don’t need a statistician to choose your treatment.

Instead, let us say you treat 1000 patients with A or B, and the average lifespan for those given A is 2 years and two months, while those given B live an average of 2 years and a month. Because of the large number of patients in the trial, it is reported that A is statistically better than B. Yes, but… Those given A were twice as likely to have to stop treatment for a time due to serious side effects, and spent several weeks more in hospital. Choice depends on individual values.

There is often a serious difference between statistical significance and clinical significance.

This brings me to lecanemab, the newest “wonder drug” for Alzheimer’s disease. Like several others released or in testing, this monoclonal antibody targets amyloid-beta, a protein that accumulates in the brain of patients with Alzheimer’s. I must note that experts are still not agreed on whether amyloid causes Alzheimer’s or is simply a marker of the disease.

The trial of lecanemab involved 1795 patients, half getting an infusion of the drug every two weeks and half getting a placebo. The results, touted loudly by the drug’s manufacturer showed a highly statistically significant difference (P<0.001) in favor of the drug.

When you dig deeper, the results are less impressive. Patients in both arms showed steady deterioration in mental acuity, though the decline was less for those on the drug. The absolute difference after 18 months was 0.45 on an 18-point scale, a difference, but a modest one.

A quarter of those given the drug had serious reactions to the infusion, and one in eight had brain swelling. There were two deaths in the active treatment group that remain unexplained. Potentially useful? Maybe. A game-changer? No.

Existing drugs like the cholinesterase-inhibitors (Aricept et al) also slow the decline but do not cure the disease. There is also recent evidence that this class of drugs prolongs life.

In the meantime, do things proven to help that are low cost and safe: do not drink to excess; every drink over 1/day progressively shrinks your gray matter. Eat less processed and more unprocessed food. Be sure your Vitamin D levels are in the normal range; if you do not live in the tropics, taking a daily supplement is advised. Be sure your blood pressure is below 140/80.

My bias is that chasing amyloid is chasing down a blind alley. Time will tell, but be cautious about “statistically proven” treatments of marginal clinical benefit.

Prescription for Bankruptcy. Buy the book on Amazon

No comments:

Post a Comment