he study came with impeccable pedigree — published in a peer-reviewed journal, using the most rigorous approach — and it seemed to prove what countless worried baby boomers want to believe: that breaking a sweat is good for the brain.
Researchers had older women with mild cognitive impairment, which often becomes Alzheimer’s disease, exercise twice a week for six months. In women who did resistance exercises — with weights, for instance — or balance training, the size of the hippocampus decreased about 2 percent, on average. But in women who did aerobics, this memory-forming structure increased 4 percent, scientists reported in 2015. The 6-point difference was hailed in dozens of news stories, with headlines calling exercise the brain’s “miracle drug.”
That and similar studies have persuaded aging boomers, physicians, and even some professional medical groups that hitting the gym has brain benefits. But they have also precipitated a clash of titans.
Late last month, the American Academy of Neurology said for the first time that neurologists “should recommend regular exercise” for patients with mild cognitive impairment (MCI), which affects memory, thinking, and judgment. (The guidelines don’t specify the intensity or duration of exercise.)
Nine days earlier, however, researchers who examined 262 studies on the possible brain benefits of exercise — from tai chi to aerobics — for a panel of the National Academies of Science, Engineering, and Medicine concluded that empirical support is woefully “insufficient” for claims that exercise can prevent MCI or dementia.
The disagreement sheds light on a deep ideological schism between equally respected experts about what constitutes proof, and how rigorous it must be for a public health recommendation. In the study that triggered the “miracle drug” headlines, for instance, one-fifth of the aerobics group dropped out, which the experts reviewing research for the National Academies considered a bright red flag. The authors of the study didn’t discuss the attrition — that fact was buried in a data table in their paper — let alone explain how it might have affected their results.
Everyone involved is being diplomatic.
“I can imagine a practitioner thinking, well, it can’t hurt” to tell patients with MCI to try exercising, said Michelle Brasure, of the University of Minnesota, who led the team that reached the “insufficient evidence” conclusion for the National Academies, which is regularly tapped by government agencies and others to review scientific questions. (This one was requested by the National Institute on Aging.) “From their point of view, why not recommend it?” (Her team also looked at drugs and brain training, reaching equally dismal conclusions.)
“I wouldn’t go to the bank with the claim” that exercise can keep MCI from progressing to dementia, said Dr. Ronald Petersen, director of the Alzheimer’s Disease Research Center at the Mayo Clinic, who led the guidelines panel telling neurologists to recommend exercise. “Absolutely not. But we thought it was a reasonable thing to say, especially since [exercise] doesn’t have a lot of risk. To be honest, we’re looking for something positive to tell people.”
For an intervention that has animal studies, logic, and some human data behind it, evidence that physical exercise can help keep mild cognitive impairment from becoming dementia (or MCI from developing in the first place) has been elusive. “It’s really not known” how exercise might benefit the brain, Petersen said, though by improving cardiovascular health (which is undisputed), it might help cerebrovascular health, the integrity of the brain’s blood supply.
Mouse studies have found that in animals genetically engineered to develop a rodent version of Alzheimer’s, their brains’ level of amyloid, the diagnostic marker for the disease, was inversely proportional to how much they had exercised. “People postulated that physical activity induced enzymes that are known to metabolize amyloid in the brain,” Petersen said. “I don’t know if that has any application for humans, but it was interesting that exercise might have that effect.”
In human studies, a lack of physical exercise, especially before or by middle age, leaped out as something that increases the risk of Alzheimer’s. It made sense to assume the flip side of that: Exercise should reduce that risk.
Despite a steady drumbeat of “magic bullet for the brain” headlines, that has been frustratingly difficult to prove, however. When the National Academies panel dug in to the hundreds of studies on exercise and MCI, “about a quarter to a third had outcomes suggesting a benefit,” said Dr. Howard Fink of the Minneapolis VA Health Care System, a co-author of the evidence review. All of them were randomized controlled studies, considered the most rigorous way to determine whether an intervention has a benefit. But when the reviewers scrutinized this supposedly gold-standard research, it toppled like a Jenga tower.
Participants dropping out was a chronic issue. In addition to the hippocampus-measuring study, one in which people with MCI were randomly assigned to practice tai chi or to do toning and stretching concluded that tai chi lowered the chance of developing dementia within the next year by a whopping 79 percent. But only 54 percent of the tai chi group stuck it out.
The reason that’s a red flag is that people who volunteer for a study on exercise and brain function (everyone is told the purpose of the study) and who don’t notice any improvement in memory might decide, partway through, that they no longer want to do aerobics or tai chi. That would leave only people who are, for whatever reason (including random fluctuations in brain health), seeing such a benefit. By including only “winners” in the analysis, the studies would have inadvertently skewed the results in favor of the “exercise helps the brain” hypothesis.
“The problem is, we don’t know what happened to the people who aren’t in the analysis” because they dropped out, Brasure said. “Maybe the reason they dropped out was that they were not seeing a cognitive benefit. The reported benefit would then be artificially inflated.” In the analysis she led for the Academies, study after study had dropout rates of 20 percent, 30 percent, or more, undercutting their credibility.
Selective reporting was also a problem. “They would report using 10 tests [of cognitive function] and report results of only four,” said the VA’s Fink. “The suspicion, cynically, is that the outcomes they didn’t report looked worse and that they were cherry-picking.”
In the analysis Brasure led for the Academies, study after study had dropout rates of 20 percent, 30 percent, or more, undercutting their credibility.
A number of studies also reported only subgroup analyses, meaning results for only a subset of the total study population. “That’s always a little suspect,” Brasure said. “You know they did the analysis on the whole group, so you have to wonder.”
Although study authors blew past these weaknesses, the National Academies group didn’t. The “committee searched in vain for convincing evidence of effective preventive interventions from the randomized controlled trials,” said Dr. Eric Larson, of Kaiser Permanente, who served on the committee. “Nothing even approached the evidence level required” for a government-backed public health recommendation.”
As for the neurology association, it cited only two studies to support its new advice that physicians recommend exercise to patients with MCI. In one, more than 40 percent of the exercise group dropped out. In the other, just under 8 percent did, but like the first it included fewer than 100 people and followed them for only six months. Small studies of short duration are considered relatively weak. “There just weren’t that many strong studies of intervention for MCI,” Petersen said.
All the studies were done by reputable scientists and published in peer-reviewed journals, the magic words that lead the press to publicize them. Yet because of the flaws identified by the National Academies, “I think their conclusion was fair,” said Dr. Steven DeKosky, a neurologist and an Alzheimer’s researcher at the University of Florida. “If there is a signal there, we can’t quite see it.”