19. Conclusions
Mortality in old age has in recent decades declined in a manner so unprecedented that it can rightly be called a new stage in epidemiological transition (Olshansky and Ault 1986; Rogers and Hackenberg 1987) causing a shift from fertility-dominated to mortality-dominated population aging (Yu and Horiuchi 1987; Horiuchi 1991). This unexpected but by now well-known decline encompasses all elderly age groups as high up as can today be reliably measured, that is to say around ages 106-108. The best available data on the oldest-old in the Odense database show that the characteristic pattern of this decline is one of larger relative gains around age 80 which then gradually diminish towards ages 100 and over. This is in line with the contention of BOURGEOIS-PICHAT (1952, 1978) that the causes of death prevalent in older ages are more resistant to reduction. Our own data do not reach below age 80 but the death rates published by the United Nations show a widespread decline among the elderly in general and, in many cases, larger reductions among people in their late sixties or seventies than above it. (KANNISTO 1994, p. 48). Exceptions to this pattern of diminishing progress with advancing age which have been observed among men in some countries are clearly associated with unfavourable development in the mortality of middle-aged men and thus alien to the new transition.
The decline so far observed has varied between countries as to time of onset, speed, depth and acceleration or slow-down but has reached virtually all low-mortality countries and is going on unabated. A sustained decline in old age mortality began earlier for women than men (the average year of onset in 25 countries was for women 1968, for men 1976; KANNISTO 1994, p. 42) and has been, in percentage terms, deeper for women than men in virtually every country.
We are not able to measure the mortality of the oldest-old in the U.S. or Canada with the extinct-cohort method from available data because of age overstatement. At ages below 75, however, the official data are undoubtedly reliable and show that the mortality of middle-aged and elderly U.S. and Canadian men began declining rapidly in late 1960s, and among women somewhat earlier. (U.N. Demographic Yearbook). There is hardly any doubt that this decline extended to the oldest-old. North America therefore entered the new stage of epidemiological transition earlier than most of Western Europe. Only in France, did a sustained decline begin for both sexes - and in some others for women - already in the 1950s.
All data of good quality indicate a monotonic, though less than exponential, rise in mortality with advancing age and no evidence of a plateau or peak is seen before age 110 at least.
The often assumed convergence of male and female death rates in old age is revealed as only apparent. Although the sex ratio of mortality falls, the difference per population-at-risk remains roughly constant at least some years past age 100.
Data which are based on accurate age information have not produced any significant mortality crossover by age either over time or between populations. Trajectories calculated from unreliable data, however, have a very strong tendency to flatten out and intersect more reliable curves.
We suggest that change in old age mortality may be better understood if it is observed through an array of indicators selected for their relevance at high ages. Thus, the relative reduction in the death rate is usefully complemented by the absolute (per population) reduction which has a different age pattern and possibly a different gender relationship. At highest ages the ebbing force of life may be more appropriately measured by the probability of survival than that of dying. The relative increase in the former describes the expanding frontier of survival. Since the desired end result in combatting death is increase in lifetime, the gain in years lived per person, broken down by age in which it is lived, is perhaps the most significant parameter of change. Finally, the age shift is another meaningful and easily conceptualized indicator which can be taken to reflect delay in individual aging.
It has been, we believe, conclusively demonstrated that the recent decline in old age mortality has been caused by period factors with simultaneous and apparently immediate effect at all old age groups (KANNISTO 1994, pp. 55-59). It is manifestly not the result of supposedly healthier cohorts growing older and gives no support to BARKER's fetal origins hypothesis of adult disease (Barker, 1992). As deaths are often the result of long term morbid processes, the past has certainly an effect on today's mortality and cohorts may differ regarding their frailty. Yet, factors "here and now" seem to have affected these processes sufficiently to make an immediate difference and to submerge possible cohort effects.
This is not necessarily always the case. Smoking by women, for example, is reported to have increased in many countries. If this tendency persists, its adverse effects will be felt with a considerable time lag. If the incresed smoking is cohort-related, the habit being acquired when young, the eventual mortality effect will also be cohort-dependent.
Regarding the causes of the mortality decline no unanimity exists and could hardly be expected but among researchers who have worked extensively on the subject, and many of whom have carried out representative, large-scale follow-up studies, a wide body of opinion emphasizes the role of such lifestyle factors as improved diet, less smoking, and physical excercise while many include as additional factors new or more widely available medical practices such as control of hypertension (Al-roomi et al. 1989; Bah and Rajulton 1991; Epstein 1989; Valkonen et al. 1993). Utilization of some medical improvements is cited (La Vecchia et al. 1991) while some others see no impact by new hospital facilities (Mackenbach et al. 1988, 1989; Wing et al. 1986) or by intervention projects (Aase 1989). A follow-up study in the Finnish provinces of highest mortality from ischaemic heart disease showed that the rapid decline in the mortality was mostly due to change in three major risk factors: serum cholesterol, hypertension and smoking (Vartiainen et al. 1994).
Our own data do not provide direct evidence on the causes of the decline but the results are compatible with the existence of a major lifestyle factor. Advances in medical science and availability of new preventive or curative practices have certainly played a role but do not explain the 15 year time lag of Western Germany behind France and Belgium nor the lack of progress for Dutch and Norwegian men. The adverse development regarding men in Eastern Europe is usually ascribed to environmental pollution and deterioration of medical services but even here, lifestyle factors may have played a role and contributed to the divergent trends for men and women.
The existence of spatial differences needs to be carefully considered as they may be due to more permanent factors such as genetic heritage (Koskinen 1992) or the mineral composition of the soil (Gavrilov and Gavrilova 1991).
The new stage in mortality transition has spread to most of the advanced countries and possibly to elites in others. It has gathered considerable momemtum and further acceleration is noticeable at least in some of the latest data. As, furthermore, mortality differences between countries, as shown in the relevant chapters, have not narrowed and as such differences are found to persist also between social groups in a given country despite continuous decline in all groups (Valkonen et al. 1993; Martelin 1994), ample scope for further decline in old age mortality clearly exists.
If this potential can and will be taken advantage of, and the progress of the last 20 years is repeated three more times, death rates of 80-year-old men will be reduced by nearly a half and women by two-thirds. Even the mortality of centenarians would decline by one third. The present death rate of 80-year-old men would shift to age 86 and that of women to 87. The probability of surviving from 80 to 100 would increase for both sexes by a factor of 7 or 8 and approach 5 percent for men and exceed 10 percent for women. As a consequence, the numbers of very old people would continue to grow very rapidly - and the faster, the higher the age.
The bulk of the absolute increase, however, would be below age 100. If mortality declines only above age 80 and not below it, the greatest increase would be at ages 85-94. Considering also the probable decline below 80, very large gains can be expected at all ages from 80 to 94 and quite substantial ones at 95-99.
This development would change the modal (most common) age at death of men from the present 80 years to 85 and that of women from 85 to 92. Life expectancy at age 80 would increase for men by 2.8 years and for women by 4.3 years. The contribution to life expectancy at birth would be somewhat less than this. Under the rather extreme assumption that the present mortality decline would continue at the same speed for one hundred years and all mortality below age 80 is eliminated, female life expectancy at birth would still not reach more than 94.8 years. The extension of the length of life will ultimately meet strong resistance near age 100 unless mortality at these very highest ages will in the future decline faster than has been the case so far.
Barring unexpected adverse developments on one hand, and unforeseen scientific breakthroughs on the other, the new stage of mortality transition is likely to continue for some time to come, possibly for an extended period. This could be construed as better preservation of an individual's life potential and, therefore, the scope for improvement would be finite. The forward scenarios based on recent change describe how the continuation of the trend would increase and modify the oldest sector of modern populations but would not cause a revolutionary increase in the length of life of man.
Updated by V. Castanova, 1 November 1999