Our results of LTL and mortality corroborate the associations found between shorter LTL and risk of death in several other studies (17
). We found that persons in the shortest quartile of LTL had a 60% increased risk of death compared with those with LTL in the longest quartile. In terms of cause-specific mortality, shortened LTL predicted an almost threefold increased risk of death by infectious causes, which was consistent with results reported by Cawthon et al. (17
). Although we had hypothesized that, in addition to infectious disease, we would find associations with cardiovascular disease-related deaths based on previous studies and our own preliminary results with cardiovascular risk factors (9
), results here were weaker than expected (borderline significance when adjusted for other cardiovascular risk factors) However, aging-related cardiovascular compromise might increase susceptibility and diminish ability to fight infections without necessarily causing death from myocardial infarction, arrhythmia, CHF, or stroke. In addition, as LTL reflects telomere dynamics in the hematopoietic system, shortened LTL might denote aging-related loss of immune function (40
The 60% increased risk of death from any cause between persons in the shortest compared with the longest quartile of LTL found here is similar to other studies in the literature including results from 143 individuals aged 60–97 years (risk ratio 1.7, 95% CI: 0.82–3.53) (13
), and in 780 patients with stable coronary artery disease (age-adjusted HR: 1.8, 95% CI: 1.2–2.9) (23
). Our results were somewhat weaker than those reported in a study of Swedish twins comparing shortest with longest telomeres (RR: 2.8, 95% CI: 1.1–7.3) (21
), in a study of 195 nondemented post-stroke patients reporting an inverse linear relationship with longer LTL (HR: 0.52, 95% CI: 0.28–0.98) (24
) and in 204 Danish twin pairs (HR: 0.56, 95% CI: 49–63) (20
). Our results differed from several other studies, which did not find associations between LTL and mortality (16
). Nor did we duplicate the more than twofold increased risk of cancer mortality recently reported (13
), although our number of cases was larger. Differences in methods that measure LTL and demographics of the studies are likely involved. In addition, the null studies generally included small sample sizes.
The lack of an association between LTL and total or cause-specific mortality in the Health ABC Study, which analyzed data from 2,721 participant over 8.2 years of follow-up (16
), deserves greater scrutiny. Similar to CHS, the Health ABC Study is a multisite population-based study recruited from Medicare eligibility lists, the outcome ascertainment used similar protocols, and length of follow-up is comparable. However, Health ABC included a much larger percent of African Americans (47%) compared with 15% here. In spite of the greater racial diversity, Health ABC participants appear to be more homogeneous that CHS with a more restrictive age range (70–79 years) and better health based on exclusions involving both physical and cognitive function. This lack of heterogeneity may have reflected less variation in health that may be needed to discern associations. In addition, standard deviations of LTL were much larger in Health ABC corresponding to the larger coefficient of variation (5.8%) in Q-PCR compared with our results using the Southern blot method (1.7%). Average LTL was actually marginally shorter in African Americans than in whites in Health ABC (4.87 for whites compared with 4.77 for blacks), although in our study LTL was significantly longer in African Americas, which has been reported in other cohorts with LTL measured by the Southern blot method (42
). All of these factors may have contributed to the differences in results found between the two studies.
In a preliminary report, we did not find a significant association with total mortality in 419 participants of the CHS cohort (HR: 1.22, 95% CI: 0.91–1.63) (9
), only slightly lower than the HR found in the current study. Examination of the two groups revealed similar age and gender distributions; however, there was a slightly larger proportion of African Americans included in the original sample (18.1%) compared with here (14.8%). Additionally, although the sample of CHS participants in the preliminary results were selected at random, those in the follow-up article were required to have survived from 1992 through 1997 in preparation for a longitudinal study. Although this may have affected results, the larger sample size is most likely the reason for significant findings here.
Our finding of an association between LTL and infectious disease-related deaths is reasonable. Data strongly suggest that erosion of telomeres is the result of an accruing burden of oxidative stress and inflammation (27
), which is known to be enhanced by exposure to infectious and inflammatory diseases (30
). The significant relationship between LTL and interleukin-6, a biomarker of inflammation, found here and in our earlier study (9
), supports this finding. The weaker relationship found here with cardiac deaths is more perplexing but may be affected by the different mechanistic factors included in this broad category. Although the number of arrhythmic and CHF-related deaths is small (61 and 28, respectively), the HRs suggest relationships with LTL to be in opposite directions, that is shorter LTL in arrhythmic deaths and longer LTL in those associated with CHF. In some elderly persons, this could potentially be explained by increased left ventricular mass, which is often observed in CHF due to hypertension. In fact, associations between longer LTL and increased left ventricular mass have recently been reported (47
). However, in this study, we did not have information that would help to precisely discern the causes of CHF. Nevertheless, our findings and the recent reports underscore the complexity involved in the relationships between LTL and cardiovascular indices, and the need for larger and more comprehensive studies to explore them.
Although the focus of this article was on mortality, we examined associations with other morbidities for descriptive purposes. Of greatest importance, associations between LTL and age, sex, and race were found (p
< .001) in the directions consistent with other reports providing confidence in the data analyzed here. Associations were also found with hypertension, smoking status, glucose, carotid intima-medial thickness, and interleukin-6, but they were not present for body mass index, diabetes, insulin, C-reactive protein, and history of prevalent myocardial infarction, CHF, and stroke. Confirming inconsistencies in the literature, these results may reflect the cross-sectional nature of these data, imprecise measurement, or human variability. It is also possible that LTL reflects the combination of risk and burden of disease rather than associations with individual factors that may reflect “somatic fitness” (49
). This has been reported using years of healthy life in Health ABC (16
) and physical function in some (15
) but not in all (50
The large sample size available for analysis here is a major strength of this study, which allowed us to evaluate individual causes of death in addition to total mortality. Other strengths include the high-quality data reflecting both clinical and subclinical disease that has been collected in the CHS, the detailed standardized evaluation of death events using clinical and other data in addition to death certificates, and the use of Southern blot methodology considered the gold standard for telomere length measurement. However, a limitation in our study is the relatively small number of cases within specific causes of death, which reduced power for analyses and prevented the ability to address other etiologies of death. Although multiple testing was inherent in this investigation and should be considered when evaluating p values, these analyses were predicated on previous work and hypotheses developed a priori. Although prospectively collected data were analyzed here, it is unclear whether LTL may influence adverse events in older adults or merely reflects poorer health. Regardless, results showing variability by cause of death confirm that environmental and genetic effects in humans interact differentially with telomere attrition.
This study, one of the largest to date to evaluate associations between LTL and mortality, provides additional evidence that LTL may be a proxy for underlying mechanisms that bring about pathophysiological changes in persons surviving to old age. As oxidative stress alters a number of cellular metabolites during the aging process, LTL may be one of many biomarkers, which as a group can be assessed to determine oxidative status of a person. Study of these mechanisms may help increase knowledge about cell processes involved in disease progression and chronic conditions involving both genetic and environmental influences. Additional research involving telomere attrition and repair at both the cellular and population level are needed to further elucidate LTL response or influence on the aging process.