Table 7.3 sums up the findings of all three of these studies; the first column refers not to the original analysis of the age fifty-five data, but to my later reappraisal, which coded individually for vascular and genetic risks. N in each column refers to living men; one of the reasons that predictive associations change is that certain risk factors, like badly adapted parasites, kill off their hosts, leaving themselves with nothing left to live on.
Study IV, 2012: How to live forever. In 2012, as I write this, sixty-eight of the active Grant Study men are still alive; fifty are still cognitively intact. I wanted to learn what I could about what makes for good health and good mentation over so many years. Ideally this study will be done again ten years from now, by which time we will know exactly when each of the men died. But to bring it all together in a provisional way, I made the best estimate I could of how much longer the men would live, based on past experience. I added seven years to a man’s current age if he still had only minor health problems, five years if he was chronically ill, and three years if he was disabled. Obviously there’s speculation here, but after the age of ninety there’s a limit to how wrong the guess can be. Then I plotted all the men’s lifespans against many factors that had been significant in other contexts to see what I could learn about their relationship to longevity. Table 7.4 is the result.
Men with no vascular risk factors lived to an average age of eighty-six. Men with three or more lived to an average age of only sixty-eight. This complex of factors subtracted eighteen years from a man’s expected life. The difference in longevity between men with very long- and very short-lived ancestors was seven years—eighty-four years as opposed to seventy-seven—a significant difference, but nothing like that predicted by the vascular risk factors. Surprisingly, parental social class made less than a year’s difference. The men with the very best and worst adjustments at forty-seven lived to eighty-five and seventy-seven respectively—an eight-year difference in longevity that was very significant, but again far less than that produced by the vascular risk factors.
A curiosity: the sixteen men with the most combat exposure in World War II died significantly younger than the rest of the sample. They were six times as likely to die from unnatural causes (e.g., murder, suicide) and they were twice as likely to die before eighty. This is another of those unanticipated findings that in the coming years may turn out to be a clue to something unexpected and important.
An interesting independent predictor of longevity, which I recognized only after Friedman and Martin published The Longevity Project in 2011, was the presence or absence in college of personality characteristics of perseverance and self-motivation (in the early Study terminology, Well Integrated and Self-Driving).23 These traits were significantly associated with longevity; the twenty-three men who manifested both in college were likely to live ten years longer than the men with neither trait (see Table 7.4).
Table 7.4 Factors Associated with Longevity (Mean Age at Death 81.5; N = 237)
Very Significant = p<.001; Significant = p<. 01; NS = Not Significant.
HOW NOT TO LIVE FOREVER
Most of the deaths before fifty-five were associated with factors over which the men had no control. We cannot pick our parents, or control the genes that lead to major depression. The vascular factors that were so strongly implicated in deaths after fifty-five, however, are a different story. They are all now widely accepted as causes of death before eighty, and we were able to confirm this. Using multiple regression analysis (unlike the Boston Museum of Science computer!) to investigate each of the five factors in the context of the other four, we found that each one made an independent contribution to health decline and death.24 As Table 7.3 illustrates, there was a very significant association between these factors and irreversible decline in health in the Grant Study men. The risk factors were of equal importance in the Glueck Inner City cohort and the Terman cohort of gifted women.25 But unlike the previously cited mental health factors, they are causes of early death over which we have considerable control. At the end of the day, good self-care before age fifty—stopping smoking, joining AA, watching the weight, and controlling the blood pressure—made all the difference in how healthy the men were at eighty and ninety.
Education is another factor to be reckoned with when it comes to health; it turned out to be critically important to the health of the Inner City sample (and modestly important for the Grant Study men, whose educational baseline was higher). The average College man who never sought post-graduate education had a life expectancy of seventy-nine years, while for the men with post-graduate educations, life expectancy was eighty-three years—a very significant difference. Years of education were powerfully associated with reduction of all of the five factors affecting vascular health. This was especially true of the Inner City men. This issue will be dealt with more fully in Chapter 10.
OUT OF OUR HANDS?
However, one interesting piece of conflicting evidence suggests that we still have to reckon with our stars. A recent paper by Thomas Perls and his colleagues on a cohort of 800-plus centenarians suggests that genes are just as important to survival past one hundred as lifestyle.26 In Aging Well I could assert that ancestral longevity was not an important predictor of survival from fifty-five to eighty. But to live past eighty, it appears, ancestral longevity again assumes significance. Another reminder that follow-up must continue as long as life does.
Cancer. Cancer is an increasingly important cause of death after seventy, and this was another factor that often seemed to lie in the men’s stars—or, if you prefer, in their genes. With the exception of lung cancer, which is closely tied to smoking, cancer in the Grant Study seemed surprisingly independent of both mental health and the vascular risk factors. In larger samples, however, and in studies focused on specific cancers, environment, dietary and sexual habits, and alcohol abuse can also be demonstrated to be important.
Age and cognition. We used the Telephone Interview for Cognitive Status, or TICS, to measure cognitive skills in men without an existing diagnosis of dementia.27 In the Grant Study, the factors most associated with a high TICS score appeared to be the absence of vascular risk factors, good vision, college IQ and class rank, exercise at sixty, and, surprisingly, a warm relationship with one’s mother when young. This is another curiosity. As we’ve gone along, I’ve pointed out that a warm childhood relationship with his mother—not maternal education—was significantly related to a man’s verbal test scores, to high salary, to class rank at Harvard, and to military rank at the end of World War II. At the men’s twenty-fifth reunion, it looked, to my surprise, as though the quality of a man’s relationship with his mother had little effect on overall midlife adjustment. However, forty-five years later, to my surprise again, the data suggested that there was a significant positive correlation between the quality of one’s maternal relationship and the absence of cognitive decline. At age ninety, 33 percent of the men with poor maternal relationships, and only 13 percent of men with warm relationships, suffered from dementia.
Dementia, like arthritis, is a curse of longevity. Vital aging at ninety is closely dependent upon the preservation of cognitive faculties, and the best way to achieve this is by minimizing the vascular risk factors. Alzheimer’s disease, however, is a special case. It is a major source of health decline after eighty, but, unlike the purely vascular dementias, it seems to be surprisingly independent of the factors that I have been enumerating.
Death before eighty can be avoided to some degree by wise lifestyle choices, but so far the Study has provided no clues as to how to prevent the two greatest and most dreaded sources of total disability after eighty—most cancers and Alzheimer’s disease.
WHAT WAS NOT IMPORTANT
Ancestral longevity. Lacking lifetime studies of humans, scientists have studied aging in fruit flies. You can breed and study many generations of fruit flies in a year, and their longevity, it appears, depends heavily upon genes. Therefore, one of the first variables the Study loo
ked at was ancestral longevity.
The myth that ancestral longevity is passed on is hard to test well; this is because most people are either so old that they no longer remember the exact ages of their grandparents’ deaths, or so young that they don’t yet know when their parents will die. It takes at least a two-generation study to learn from the subjects’ parents the age that the grandparents died, and then to follow the subjects until the last of their parents have died as well. (In the Grant Study, the last parent died in 2002, sixty-five years and three generations after the Study began. The Study men themselves were very unreliable informants about the age of death of their grandparents.)
For the College men, we calculated ancestral longevity by averaging the ages of death of the oldest of the first-degree ancestors (parents and grandparents) on the maternal and paternal sides. As I’ve noted, the longevity of the forty-four men with the longest-lived first-degree ancestors was seven years longer than the longevity of the thirty-six men with the shortest-lived ancestors. That difference was significant, but just barely. Moreover, to my surprise, the average lifespans of the ancestors of the men with the best and worst mental and physical health at eighty were identical.28 In the Inner City sample as well, the longevity of the men’s parents seemed irrelevant to the quality of their own aging at seventy.29 Fruit flies are clearly not always good models for human aging. Obviously, specific genes are very important in predicting the specific illnesses that shorten life, and soon the precise genes that facilitate longevity may be discovered. In the meantime, however, the McArthur twin studies and investigators using the Swedish Twin Registry confirm our finding that variance in longevity cannot simply be attributed to genetic inheritance.30
Psychosomatic stress. When the Study began, psychosomatic medicine was in high fashion. Hans Selye had shown the world that stress could kill, and psychoanalytic theories about the role of emotions in medical illness were all the rage, attributing peptic ulcers to repressed anger or longing for love.31 It would be decades before the medical profession brought itself to accept that the most common cause of duodenal ulcers is a gram-negative bacterium of the genus Helicobacter.
Besides, many people do feel physically sick under stress. They have headaches; they can’t sleep; their stomachs ache; they get itchy; they’re running to the bathroom all the time. This observation suggests an attractive hypothesis, to which in the 1960s I earnestly subscribed: that individuals who experience stress psychosomatically in midlife would suffer poor physical health in old age.32 When I joined the Grant Study in 1966, I was thrilled to be in possession of prospective Study data that I suspected would prove this hypothesis.
A secondary hypothesis in psychosomatics, to which I also subscribed, held that individuals have characteristic “target organs” in which they experience stress.33 This proposition was based on the observation that one person’s place of experiencing somatic symptoms under stress may be quite different from another’s, and it was much emphasized in my residency training. Theoreticians anticipated that the organ affected by stress would be the same one in which signs of psychosomatic illness appear.
A third related hypothesis, one that I felt certain of proving, was the implicit assumption that the development of “psychosomatic” illnesses (for example, colitis, asthma, and hypertension) reflect more psychopathology than the development of “real” illnesses (such as diabetes, myocardial infarction, and osteoarthritis). Indeed, many of the mental health screening tests of the 1950s and ’60s, some of which, like the Minnesota Multiphasic Personality Inventory, are still in use today, use the presence of multiple physiological symptoms under stress as indicators of emotional illness.
In the 1970s I finally analyzed the data to test these three psychosomatic hypotheses so favored by me and other armchair speculators. My intention was to demonstrate 1) that psychosomatic illness leads to accelerated aging; 2) that “target organs” for stress (the stomach, the lungs, and so on) are real and remain stable over time; and 3) that psychosomatic illness is a reliable hallmark of mental illness. Longitudinal study proved them all wrong.
(Note that the evidence for this belief system had been derived over the years from retrospective data. Furthermore, the source of the data was patients who presented themselves repeatedly for medical attention—a group likely to have mental health issues.34 The Grant Study, on the other hand, was a prospective study of men who explicitly were not patients.)
Over the years, the Study men were systematically asked where in their bodies they experienced emotional stress. Decades of follow-up revealed that this site varied considerably over time.35 The idea of stable target organs (hypothesis II) was not supported.
Nor did the number of physical symptoms under stress before age fifty predict physical health at age seventy-five or at ninety. If you wait a few decades, people often recover from psychosomatic illnesses. And at age eighty, the physical health of the men with multiple apparently psychosomatic illnesses was actually, if insignificantly, better than that of men with none. Hypothesis I was not confirmed either.
Hypothesis III, the supposition that there’s a connection between psychosomatic illness and mental health, didn’t fare any better than the other two. The Study has carefully followed the men’s objective physical health from the age of forty, when it was still very good, up through the present or until their deaths. By fifty, over half of the Study men had required medical treatment for conditions thought by some physicians to be psychosomatic: hypertension, respiratory allergies, ulcer, colitis, and chronic musculoskeletal complaints. Admittedly, these five disorders do not define psychosomatic illness conclusively, but they make a good start. At age forty-seven, after almost thirty years of observation, the men were ranked for emotional health. There was no correlation with the number of psychosomatic complaints. The number of psychosomatic illnesses that the men suffered didn’t predict mental health at age sixty or eighty, either. As a faithful believer in psychosomatic medicine, I was profoundly disappointed.
We also found that the childhoods of the men who experienced relatively few physical symptoms under stress and no psychosomatic illnesses were no warmer in blind ratings than those of men cursed with a great deal of psychosomatic symptomatology.36 As I’ve noted, in our first study of aging, premorbid psychopathology did correlate with early onset of chronic (real) physical illness, but only up to the age of fifty-five. We had some middle-aged Study men who frequently visited medical doctors and who took more than five or more sick days a year. They displayed our distinguishing markers for mental health issues—drinking, psychiatric intervention, use of tranquilizers and antidepressants, and so on—and in this they resembled the self-selected populations of the old retrospective studies. In other words, people with mental health issues brought more illnesses of all kinds—real and imagined—to the attention of the medical community.
When in 1970 I first presented these findings at a national psychosomatic meeting, I was met with outrage and disbelief as a traitor to dynamic psychiatry. Today, my findings would be regarded as unremarkable. Over the last forty years, a medical model of psychosomatics has displaced the earlier one, thanks in part to the Helicobacter story and the fading influence of psychoanalysis in academic psychiatry. My research life, and the Study’s, spanned both eras.
Cholesterol. Magazines would lose valuable advertising revenues, and probably readers, if they talked too much about really significant risks to health—like Virginia Slims and wine coolers. But it’s OK to fuss about cholesterol, because the butter and egg lobbies don’t advertise in Cosmopolitan. Moreover, as the TV ads tell us, your cholesterol can be magically lowered by statins, even if you don’t eat less or exercise more. So the war on cholesterol would appear to be a win-win-win situation—for patient, for doctor, and for the pharmaceutical industry.
Countless studies show that the ratio between high- and low-density lipoproteins (HDLs and LDLs) is important and that reducing LDL levels is good for the heart. But in neither the Coll
ege nor the Inner City sample did average cholesterol levels at age fifty distinguish the men who lived to ninety from those who died before eighty. The estimated age of death of the fifty-eight Grant Study men with cholesterol levels below 206 mg/dl was eighty-three. The estimated age of death of the fifty-seven men with cholesterol levels over 254 mg/dl was eighty-one, which was not a statistically significant difference. This finding has been confirmed by much larger and more representative studies.37
It’s situations like these that persuaded me to include my experiment with the Boston Museum of Science computer—not as entertainment, but as an alert. To understand longevity we need longitudinal images, not snapshots. The more sacred our cows, the more they need longitudinal testing. This is one of the ones we were able to test.
Bleak childhood. Alas, we cannot choose our families. Without asking permission they endow us with their genes. They also bathe us in their warmth and riches—or parch us with emotional and financial poverty. Many aspects of childhood are important to aging, as I’ve described in Chapter 4. But most of them do not predict length of life. Parental social class, stability of parental marriage, parental death in childhood, and IQ (at least in our samples with their two different but limited ranges) were not important to longevity. The men from the warmest childhoods lived only a year and a half longer than the men from the bleakest childhoods, a difference that was not statistically significant.
Vital affect and general ease in social relationships. These were the two personality traits most highly valued by the original Study investigators, and for the Harvard cohort, they correlated highly with good psychosocial adjustment in college and in early adulthood.38 But they, too, failed to predict healthy aging.
Triumphs of Experience: The Men of the Harvard Grant Study Page 25