We studied a population of 41,063 individuals deceased in Gipuzkoa (region of the Basque Country) from the beginning of 2014 to December 31, 2019. The mean age of the population was 82 years, with the youngest individual deceased at 50 years and the oldest at 109 years. The proportion of women and men was almost the same (49.54% women vs 50.46% men). At lower lifespans, the proportion of men was higher (65% vs 35% at 60 s), with this trend reverting until the centenarians, where women far surpassed men (85% vs 15%). The demographic characteristics of the population were summarized in Table 1.

First, we performed survival analysis in order to study the relationship between the age of first diagnosis and lifespan. For that, we built Kaplan–Meier curves for each ICD-10 disease category and sex. We observed differences between diseases with neoplasms and circulatory, digestive endocrine, and musculoskeletal system having the higher incidence. Noteworthy, we obtained a continuous spectrum, independently of the disease, in which the higher the lifespan, the higher the age of onset (Fig. 1). Some diseases, such as neoplasms and diseases of respiratory and musculoskeletal systems, displayed different patterns between women and men. We also saw that some diseases, such as those related to the circulatory system, had a higher incidence when increasing lifespan, while neoplasms, particularly in the case of women, showed the opposite behavior (Fig. 1).

Kaplan–Meier curves of the age of onset of disease for different disease groups in function of lifespan and sex

Next, we analyzed the HR provided by the Cox regression to test the effect of lifespan and sex in the onset of each disease categories. For all the diseases, lifespan had a protective effect: the higher the lifespan, the lower the HR (Table 2). Regarding sex, it also had a significant HR for every category, with women showing a protective effect for every disease except for those belonging to the nervous and musculoskeletal systems. When assessing interactions between lifespan and sex, we found that for most disease groups, women at shorter lifespans had higher HR for developing diseases (Table 2).

We repeated the same approach to evaluate the age of onset of individuals with multiple systems affected. We assessed the age of onset of the nth disease category affected, from the second to the eighth. As in the previous analysis, the Kaplan–Meier curves showed the same delay on the onset as the lifespan increased, indicating that this effect was not specific for single diseases. Furthermore, as the number of affected systems increased, the incidence decreased (Supplementary Fig. 1). In Cox regression, we saw that once again lifespan had a protective effect, exhibiting lower HR as the number of systems involved increased. On the other hand, women showed protection against having multiple systems affected, a protection which increased along with the number of systems affected (Supplementary Table 1). These results showed that the most long-lived individuals displayed longer health spans being women particularly protected.

Next, we evaluated the capacity to avoid diseases for each lifespan. We defined “escapers” as those individuals who died without a specific disease; for instance, escapers for neoplasms were individuals that died without having a diagnosis of neoplasm, even if they had diseases from other categories. We represented the trajectory of the prevalence of escapers at different lifespans for each disease group and gender. All diseases reached a minimum of escapers at 70–90 years. Furthermore, most categories showed similar prevalences at the lowest and highest lifespans. We observed some extreme patterns, such as that less than 10% of the population with a lifespan of 80–90 years was able to avoid circulatory diseases. On the other hand, diseases of the nervous and genitourinary systems had almost no incidence at lifespans of 50–60 years (Fig. 2A).

Trajectories of escapers for each disease group in function of lifespan and sex. A Overall trajectories of the prevalence of escapers for each disease group. B Analysis of cluster for escaper trajectories for different disease groups in women (first row) and men (second row)

There were some differences associated with sex, with women having a higher prevalence of escapers in the majority of diseases. The cluster analysis revealed that, in the case of women, all disease categories followed a similar trend of escapers except for neoplasms. Coherently with the survival analysis, neoplasms showed a lower incidence of escapers at younger lifespans, and the prevalence increased along with the lifespan, which indicated that women who reached higher ages were able to avoid this disease. On the other hand, all the other groups of diseases showed a concave-shaped pattern, reaching the minimum number of escapers at a lifespan of around 85 years. In the case of men, we observed a similar pattern, but in this case, metabolic diseases were included in the same cluster as neoplasms. In men, both diseases reached a minimum prevalence of escapers before a lifespan of 80 years, and at higher lifespans, the prevalence of escapers was higher when compared to the other diseases (Fig. 2B).

When assessing the multisystem involvement, we saw a similar pattern: around 85 years, the number of systems free of disease reached their minimum, while the lowest and highest lifespans exhibited the maximum. More specifically, the minimum number of systems free of disease was reached at a lifespan of 83 years for women and 81 years for men (Fig. 3A–B). We also observed differences in the number of systems free of disease for both sexes across different lifespans, with women maintaining an overall lower number of comorbidities. However, men reached higher lifespans with fewer systems affected (Fig. 3C). The prevalences of the number of systems free of disease for each age decade were summarized in Table 3. These results showed that the long-lived individuals presented more rates of disease escapers with women presenting particularly lower number of comorbidities. Indeed, 10% of centenarians were free of diseases in contrast to only 2–3% individuals between 70 and 89 years-old groups. The comparison with Catalonian data revealed that the mean number of systems free of disease for individuals deceased between 50 and 80 years was slightly lower for both sexes in our cohort.

Trajectories of the number of systems free of disease. A Cumulative prevalence of individuals with each number of systems free of disease for women and men in function of lifespan. B Overall trajectories of the prevalence of individuals with each number of systems free of disease in function of lifespan and sex. C Trajectories of the number of systems free of disease for women and men in function of lifespan

The next step was to assess the health span of individuals, or the period of lifespan that individuals had without diseases. For that, we focus on individuals that had at least one disease. We evaluated the percentage of life free of disease, defined as the years from birth until the time of onset of the first disease divided by the lifespan, for each category, lifespan, and sex.

First, we observed different patterns both among systems and among genders. There was a notable overlap among the different diseases at lower lifespans; however, as lifespan increased, two main trajectories could be distinguished: those that maintained a higher prevalence of life free of disease, with more than 90%, and those with a lower prevalence, which reached around 85%. Diseases from both trajectories seemed to converge at higher lifespans. Notably, at lower lifespans, men exhibited a higher percentage of life free of disease for neoplasms when compared to women, while metabolic diseases displayed the opposite behavior (Fig. 4A).

Trajectories of the life free of disease. A Overall trajectories of the prevalence of life free of disease for each disease group in function of lifespan and sex. B Analysis of cluster of the trajectories of life free of disease for the different disease groups in men (first row) and women (second row)

The cluster analysis revealed two main clusters for women and two clusters for men. In the case of women, in the first cluster, neoplasms, diseases of the circulatory, digestive, and metabolic systems were grouped. Except for neoplasms, diseases of this cluster reached their minimum prevalence of health span at a lifespan of 70 years, while neoplasms showed a decrease in health span as lifespan increased. In the second cluster, there were diseases of the musculoskeletal, respiratory, nervous, and genitourinary systems, which had a tendency of increasing their health span period along with lifespan. In the case of men, diseases of the nervous system were in the first cluster, and diseases of the digestive system were in the second one, while the rest remained the same (Fig. 4B). These results described differences on health span between genders.

Finally, we evaluated the capacity of the previously assessed variables to explain lifespan. For that, we performed an MFA. We included each pathology, sex, the number of systems free of disease, and the percentage of life free of disease for each category as variables. We represented a random sample of the 10% of the population in terms of lifespan using the first two dimensions of the MFA. First, we saw that our variables were able to discern the differences in lifespan, since the distribution of ages of death followed a pattern, with the lifespan decreasing along the vertical axis. We also observed two differentiated clusters of individuals, with the first one being all women and the second one being men (Fig. 5A). When assessing the contribution of the variables to the algorithm, the different pathologies were the group of variables that most contributed to explaining lifespan, followed by multimorbidity, sex, and the percentage of life free of disease (Fig. 5B).

Multiple factor analysis of the assessed variables. A Representation of a random sample of 10% of the population in terms of lifespan and sex according to the first and second dimensions of the MFA. B Contribution of the groups of variables to the MFA. The red line indicates the expected value considering uniform distribution