By David N. Leff Science Editor

"There are three kinds of lies," someone once said; "plain lies, damn lies and statistics." Whoever coined that truism was lying.

Statistical probability is the fourth dimension of molecular genetic research, as witnessed by a paper in the latest Lancet, dated July 13, 1996. Its title: "Specificity, sensitivity and predictive value of apolipoprotein-E genotyping for sporadic Alzheimer's disease."

The article's authors, at Duke University in Durham, N.C., describe how they used simple statistical tests to sharpen the diagnoses of senile-dementia patients thought to be suffering from Alzheimer's disease (AD).

It's a grim irony that to date autopsying the brain after death is the only way to verify that an elderly individual with failing memory and cognition was truly a victim of AD. So far, a neurologist's finding of Alzheimer's is a judgment call, an educated guess, not a biologically founded finding.

That occurs only when microscopic examination of brain tissues reveals the telltale neurofibrillary plaques and tangles specific to AD.

The trouble is that AD is only one of many forms of senile dementia, some of which _ unlike AD _ are treatable. Thus, a clear-cut differential diagnosis would spare true AD patients futile drug therapies, and speed up prescribing of such beneficial treatments to non-AD individuals.

Molecular geneticist and cell biologist Ann Saunders, lead author of the Lancet paper, told BioWorld Today that the most common of such AD look-alikes "is multi-drug prescription-taking. As you become elderly," she explained, "you start taking blood-pressure medicine and other drugs, and frequently the combination interactions will result in a type of dementia that can be easily corrected."

Brain tumors, too, she added, produce memory and cognition deficits that can be mistaken for AD, as does "an extremely rare condition, vitamin-B12 deficiency."

Every year in the U.S. alone, some 400,000 people are diagnosed with "possible" AD. What's wanted is a genomic "mark of Cain," identifying a true AD victim, to the exclusion of others with similar progressive mental and social disabilities. Put another way, how to get those neurofibrillary plaques and tangles out of the closet and into the open.

Blowing APOE's Cover

In 1993, molecular neurologist Allen Roses and his colleagues at Duke put the finger on apolipoprotein-E (APOE), a molecule better known for its role in managing the body's cholesterol traffic. They built on earlier showings that the gene encoding APOE exists in three genetic variants, epsilon-2, epsilon-3 and epsilon-4 (e2, e3, e4).

Of these three alleles, the third, e4, conferred on its bearer a higher risk of developing AD than the other two. Now in the current Lancet paper, researchers have quantified these odds for accurate estimates of the aberrant gene's diagnostic prowess.

For starters, Roses and his co-authors reviewed the records of 67 patients at Duke's Memory Disorders Clinic who, over a 10-year period, had received a diagnosis of "probable" AD at the time of their death.

Brain autopsies revealed that 10 of these 67 patients (15 percent) did not have AD, nor did they have an e4 gene allele. But one-fourth of the 57 remaining patients, with verified AD, didn't carry the e4. Three-fourths did.

The epsilion-4 allele, their study showed, had AD specificity and predictive value of 100 percent. That is, all of the patients with e4 had the disease, whereas none of the non-AD individuals carried the e4 version.

"In a clinical setting," said Saunders, "APOE genotyping will identify three out of four patients who have the disease. It is no help with patients who do not have an e4."

She added: "This is the first actual measurement of genotyping in a series that could evaluate misdiagnoses. In short, it can tell a physician which patients are almost certain to have Alzheimer's."

These results, Roses said, "confirm the rough estimate long made by neurologists that 15 percent of patients diagnosed with AD will not actually have the disease. But it shifts the non-AD individuals into a patient group that can be defined during life." He added: "This has tremendous implications for evaluating drug trials."

Roses also emphasized that this first-ever genetic test for the AD- APOE4 connection needs to be confirmed by much larger patient cohorts. On this score, Saunders observed that Roses will report next week to the International Alzheimer's Disease Conference in Osaka, Japan, on a series of 200 patients collected at other university centers with identical work-ups to Duke's 67.

Athena Offers Duke's Test For AD Suspects Only

Roses made the point that testing the general population for the apolipoprotein E gene is "currently regarded as premature." In fact, when Duke licensed its pending patent on the AD genetic findings to Athena Neurosciences (see BioWorld Today, Mar. 27, 1996, p. 1), it stipulated that the e4 genetic test not be used to predict the disease in asymptomatic people.

"Since rolling the tests out at the Academy of Neurology meeting in March of this year," Athena President and CEO John Groom told BioWorld Today, "we've been talking to neurologists about it. We didn't really start detailing them to neurologists until April."

The response, he added, has been, "So far, so good. We're encouraged by the growth in the number of tests that have been used from month to month, though so far it isn't going to double the stock price of Elan [which recently acquired Athena]. But I think we have an important product that will become more valuable as time goes on.

"Some neurologists," Groom continued, "are more enthused with our story than others. Their sort of standard response has been: `Well, I can diagnose AD.' And there's no doubt that they can, after a period of time.

"What these tests offer neurologists is greater clinical information in the majority of patients, and a guided way of changing the course of their former practice.

"Allen Roses' latest paper," Groom went on, "does add quite significantly to that, because what Duke did was take a sequential series of patients, in which they accepted all comers over a 10-year period that had come through their center."

He concluded: "It's probably one of the best centers in the world for the diagnosis of AD. Yet even they had this inherent 15 percent error rate in terms of misdiagnosis, which is certainly higher in terms of accuracy than will be achieved in the average neurologist's office." n

(c) 1997 American Health Consultants. All rights reserved.