The major finding of this study was that participants with higher LS benefited the most with respect to absolute mortality risk reduction (higher LS = 67; mid = 98; low = 140 per 10,000). Furthermore, higher LS was independently associated with survival in men (HR 0.55, 0.37-0.81) but not in women. The present analysis demonstrates that the relationship between LS in men and mortality persists even after adjustment for baseline risk factors and severe sustained co-morbidities, which likely provides a conservative estimate of the overall effect of LS on survival. These findings suggest that for men regardless of their somatic and psychological health, being satisfied with one's life is protective against mortality. This is in agreement with previous reports, which clearly show the association between a global subjective perception of one's own health and mortality. These studies also found a significant, independent association that persists even after adjustment for health status indicators and other relevant covariates 
The finding that LS was not associated with mortality in women is interesting. Although there were similar frequencies of high LS in both sexes in our study sample, there was a clear association of LS and mortality for men but not for women. Only few publications have addressed sex differences and found a similar sex-specific effect [27, 28]. It has been suggested that in men, morbidity-related factors are the most important predictors of mortality while in women the predictors were spread over more domains . Additionally, Koivumma-Honkanen et al. have speculated that females may be more capable of coping with psychological distress than males, thus avoiding fatal consequences . Furthermore, the reasons for these differences may also include different lifestyles and different biological vulnerability . Still, in considering mortality, statistical power may be compromised by the small number of deaths in women (62 deaths out of 1,252 participants). Nevertheless, this difference deserves further investigation.
Inclusion of individuals with pre-existing illness is potentially problematic because their perspective on life may likely be negatively affected by their disease experience, and thereby could drive down LS in the entire population. Additionally pre-existing illness is likely to be associated with both LS and mortality. Therefore, a sensitivity analysis was performed with exclusions made for patients suffering from cardiovascular disease at baseline (n = 89) and, against expectations, returned essentially identical results. Healthy participants in the high LS tertile showed a 38% mortality risk reduction compared with those in the lower tertile. These values are within the range reported previously in a meta-analysis: mean HR of 0.82 (CI = 0.76-0.89) of 21 studies with healthy populations . Additionally, it has been shown that LS significantly predicted a lowered risk of all-cause and natural cause mortality, and this association is especially salient in the healthy subsample . Again, as observed in the entire cohort, in sex-stratified analyses this association remained true for men, but not for women.
Our data suggest, as previously reported,  that LS has a favourable effect on survival in healthy and disease populations, which was lost after adjusting for other psychological determinants. The fact that LS lost significance in the psychological model, could have been caused in part by the well-established association between self-rated health and mortality , that may have weakened the relation between LS and mortality. Indeed, only when either self-rated health or social network index were excluded from the psychological model, was the LS association with mortality restored (data not shown). The exclusion of none of the other variables from the "psychological model" (including depressed mood) did not modified the association between LS and mortality.
Determinants of LS
The present study provides a broad range of somatic and psycho-social determinants to elucidate both the determinants of LS as well as possible underlying factors that may explain the substantial survival benefit of LS. Consistent with previous reports , a significant decline in LS was observed across the life span of women in our population. However, the lack of decline in LS among men reported in the literature  was not seen in our population. How LS changes with age is an intriguing question, especially in light of prior findings that it improves from middle age onward, even in the face of physical health decline; little is known about the determinants of this pattern [33, 34]. The decline in LS across life span for men and women could be partially explained by the fact that older people are more often ill and health-related factors play an important role in LS. Indeed, when the analyses are repeated only for "healthy participants" (sensitivity analysis) we can see the previously reported U-shape pattern with lowest LS levels in middle 50s for both men and women  (data not shown).
It was previously shown that socio-demographic variables explain roughly about 8-15% of the variance in LS  and psychological and social characteristics explained 62% of the variance in LS . The powerful impact of psychological and social characteristics as independent determinants of high LS is well illustrated in the present logistic regression analysis, where 73% of the variance in LS could be explained. An individual's positively-perceived health (lack of somatic complaints, good self-rated health, and good health status), a healthy psychological status (no depressed mood, anger disposition, or suppression of angry feelings) and good socio-demographic conditions (higher income, high social support) were associated with higher LS. Interestingly, although some co-morbidities (angina, insomnia, acute illness last week) were associated with differences in LS, none of these variables were relevant determinants of LS according to the logistic regression analysis.
The lack of association between co-morbidities and lifestyle factors (physical activity, diet) with LS in our comprehensive, holistic model may seem surprising. However, the relationship between well-being and medically-based health measures is still unclear. There is conflicting evidence with some studies showing that healthier people are more satisfied with life , and others which indicate that the relation between medically based health and well-being is weak . Nonetheless, this lack of association between LS and ill health in our analyses may have been due to the small number of subjects with co-morbidities in our sample, which is a reflection of the population-based character of our sample.
Strengths, limitations and guidelines for future studies
The present study has several important strengths. Foremost, it is a population-based sample, in which healthy and ill participants were included, well defined health outcomes, and inclusion of an exhaustive list of relevant covariates. The prospective study design allowed for a reasonable follow-up time to assess health outcomes. Some limitations, however, need to be addressed. In the present study, sub-syndromal depressive mood was assessed by the DEEX scale, which is a less rigorous instrument to assess depressed mood, although a recent re-examination of its validity and reliability is promising . The assessment of LS with a one-item question is disputed, however, previous studies have used similar questions [13–15] and the factorial load in the total Satisfaction With Life Scale is very high (.82 to .89) . The baseline measurement of life satisfaction assessed on average 12 years before follow-up provided strong risk estimates similar to other studies in which validated LS measurements in different time intervals were employed . The inclusion rate for this study was 55% of all participants in the survey, potentially limiting the generalizability of our findings . A cautionary note must be taken when interpreting prior findings on "positive" factors and health because it is still unclear from the literature whether "positive" traits are associated with better health or "negative" psychological traits are associated with worse health. The design of the current study, like any observational study that did not extend across the life course, cannot determine a causal relationship. However the results demonstrate an inverse association between LS and mortality.