The era of centenarians: mortality of the oldest old in Sweden

Dear Sir, Demographic evidence has shown that life-expectancy has been improving for more than 160 years [1, 2]. This remarkable trend was first fuelled by reductions in infant and child mortality; however, since the 1950s most of the reduction has been because of a decrease in mortality in the elderly. The annual decline in mortality at older ages is continuing in many countries, including Sweden [3]. According to the Human Mortality Database, Swedish life-expectancy at the age of 80 years increased by 1.84 years for men and 2.78 years for women between 1969 and 2009 [4]. The number of Swedish centenarians has risen almost exponentially in recent years, mainly because of the decrease in mortality below the age of 100 [5]. The increase in mortality has been accompanied by a steady and almost linear increase in the maximum age at death [6]. These observations raise important questions about the pattern of mortality for centenarians in Sweden. Here, we explore whether the decline in mortality that was observed below 100 years of age also continued above this age. Swedish mortality data are known to be of exceptional quality with no signs of inaccurate data after 1861, even in very old age [7]. We obtained information on the exact dates of birth and death for all Swedes who reached the age of 100 between 1 January 1969 and 31 December 2009. The data were collected from Swedish population registers and registers of yearly deaths from Statistics Sweden. The data were organized in cohorts by year of birth. Our final data set included a total of 15 231 centenarians. During follow-up, 13 471 died, 112 emigrated, and status was unknown for 12 individuals. The remaining 1636 individuals were censored at 1 January 2010. We calculated life-expectancy by cohort, which can also be interpreted as the mean age of death. One of the problems associated with measurement of the mean is that single cases of extreme age may influence the estimates of the average age at death for small populations. In addition, cohort life-expectancy can only be calculated after all members of the cohort have died. Recently, it was suggested that, although rarely used as a measure to study longevity, the median age at death is important [8]. This measure describes the age at which half of a cohort of centenarians has died. Thus, it can be calculated earlier than the mean. It is also more robust against statistical outliers. Linear trends based on weighted least-squares regression were fitted for each measure of the average age at death according to sex. The weights are inversely proportional to the variance of an observation and account for the increasing number of centenarians over time. Figure 1 shows the number of people celebrating their 100th birthday between 1969 and 2009. In 2009, 163 men reached the age of 100, which is 9-fold greater than the number 40 years earlier. The increase in the number of female centenarians was even more pronounced, with 712 women reaching 100 years of age in 2009, which represents an 18-fold increase during the 40-year period. Number of men and women reaching 100 years of age (1969–2009). Next we calculated the two measures of average age at death: the mean and the median. The results for each cohort of female and male centenarians are given in 2, 3, respectively. Average age at death for cohorts of female centenarians (1969–2007). Average age at death for cohorts of male centenarians (1969–2008). The figures for both women and men show a remarkably stable pattern over the entire 40-year period for both measures of the average age at death, even for the most recent cohorts. Because of the lower number of individuals reaching the age of 100 in the past, the mortality pattern was slightly more erratic at the beginning of the observation period. The mean age at death fluctuated around 717 days for women and around 611 days for men. The respective values for the median age at death are 522 and 454 days. The trend lines fitted to the data follow a weighted linear regression equation. The slope of this linear equation is very close to zero for both sexes; however, the lines indicate a small increase in the median age at death and a small decrease in the mean age at death over time. The aim of this study was to use high-quality data to estimate the mortality of Swedish centenarians over time. Our results are based on individual-level information for all cohorts of Swedish centenarians from 1969 until almost the present day (2009) and show that the mortality of both sexes remained remarkably stable throughout the 40-year period, even in the most recent cohorts. There are several possible reasons for this. First, we know that centenarians are a selected subgroup of the general population. With ongoing improvements in the medical system in Sweden, it is likely that increasingly more frail individuals will reach 100 years of age. These changes in the composition of the centenarian population could mask the ongoing mortality improvements also at such an old age. Second, it is conceivable that the healthcare system in Sweden does not provide the same level of care to these very elderly individuals. It may prioritize patients who are younger, or elderly patients who switch from curative treatments to palliative care. The lack of change of mortality at these extreme ages would then reflect a condition of the particular healthcare system that cannot be generalized. We also found that the difference in the level of mortality between the sexes persisted over the whole observation period. Even at these extreme ages, men consistently have a poorer outcome in terms of mortality than women of the same age. Despite the stable mortality pattern amongst Swedish centenarians, we saw a continued increase in the maximum age at death [6]. Our observations demonstrate that the survival pattern after age 100 has not contributed to the rise of the maximum age at death and that the postponement of mortality in Sweden has not yet reached the very highest ages. According to pooled data from 13 countries included in the Kannisto–Thatcher database, mortality amongst centenarians improved by about 0.5% per year from the 1960s to the 1980s [9]. In Japan, a strong decline in mortality was observed [10]. With regard to the Swedish population, Kannisto [11] reported in 1996 that life-expectancy of centenarians has tended to increase at least in the first half of the 20th century. Thus, it is important to assess whether our results for Sweden in the second half of the 20th century are an exception or whether they can be extended to the other Nordic countries. None of the authors has any conflicts of interest to declare. This work was supported by the Swedish Research Council through the SIMSAM project on Register-based Research in Nordic Demography.