• Log in with Facebook Log in with Twitter Log In with Google      Sign In    
  • Create Account
              Advocacy & Research for Unlimited Lifespans

- - - - -

Smoking does not accelerate leucocyte telomere attrition: a meta-analysis of 18 longitudinal cohorts


  • Please log in to reply
1 reply to this topic

#1 Zaul

  • Guest
  • 104 posts
  • 11
  • Location:unknown

Posted 24 June 2019 - 10:25 AM



Using meta-analytic methods to compare LTL dynamics over a mean of 8.6 years of follow-up in 4678 current smokers and 7901 non-smokers, we found no evidence for significantly faster telomere attrition in smokers. Our analyses confirmed that LTL shortened with increasing age and that smokers had shorter LTL than non-smokers at all measured time-points. However, there was no evidence that this cross-sectional difference increased with age, as would be expected if smoking causes LTL attrition. There was also no significant difference in the rate of LTL attrition measured within smokers and non-smokers. Moreover, the negligibly greater rate of attrition observed in smokers was totally insufficient to produce the cross-sectional difference between smokers and non-smokers within a human lifetime. Taken together, these findings provide no support for the hypothesis that smoking causes a sustained increase in the rate of LTL attrition.

The above conclusion is based on assuming a linear effect of smoking on telomere attrition over time (figure 1). There are strong mechanistic reasons for assuming a linear model. Current smoking causes a chronic elevation in the levels of oxidative stress and inflammation, both of which are implicated in accelerated telomere attrition in a dose-dependent manner. However, some authors have attempted to explain the lack of an effect of smoking on telomere attrition in previous studies by assuming that smoking has an initial accelerating effect on attrition that rapidly reverses with continued smoking [2022]. The evidence claimed for this nonlinear effect is the strong positive correlation observed in all longitudinal studies between baseline telomere length and telomere attrition: individuals starting with longer telomeres have faster attrition than those starting with shorter telomeres. On the basis of this observation, Farzaneh-Far et al. [22] have argued that any rapid initial shortening of telomeres caused by smoking is subsequently offset by a length-related decrease in attrition caused by this ‘homeostatic’ process. However, it has now been recognized that a more parsimonious explanation for the observed correlation between baseline LTL and attrition rate is regression to the mean resulting from measurement error [29]. Thus, the apparent ‘homeostatic’ mechanism proposed to account for nonlinear attrition rates in smokers is largely a statistical artefact. Importantly, our analysis of the current dataset provided no evidence that the association between smoking and LTL attrition changed over time, as would be expected if an accelerating effect of smoking was present early on and subsequently diminished or reversed. It is a limitation of our dataset that our youngest cohorts were already in their 20s, meaning that the majority of smokers are likely to have been smoking for at least a decade. It is therefore possible that we could have missed an effect of smoking on LTL attrition that occurred prior to the baseline telomere measurements. To decisively rule out a highly nonlinear effect of smoking that is completely restricted to the decade immediately after starting smoking, we need to know whether telomere length differences in childhood precede the initiation of smoking [17]. Cohorts are currently becoming available in which it will be possible to test this prediction.

Is the lack of an effect of smoking on telomere attrition a limitation of low power? Some authors have argued that longitudinal studies have low power for detecting effects of smoking on attrition due to their small sample sizes [20,21,26]. While it is often the case that sample sizes are larger in cross-sectional than longitudinal studies, this is not true of the 18 cohorts included in the current analysis, where the same individuals were studied both cross-sectionally and longitudinally. Furthermore, longitudinal studies eliminate the substantial between-individual variation in LTL by measuring within-individual changes in LTL and are therefore much more powerful than cross-sectional studies of the same sample size for detecting effects of smoking on telomere dynamics [44]. Importantly, our meta-analysis of 18 longitudinal datasets shows no significant heterogeneity among cohorts in the effect of smoking on telomere attrition rates. This suggests first, that it is valid to compute a summary estimate for the difference in attrition between smokers and non-smokers, and second, that the precision of the meta-analysis exceeds that of its constituent cohorts [39]. The resulting negligible difference in attrition between smokers and non-smokers of −0.51 bp yr−1 is therefore the most powerful estimate yet of the association between smoking and telomere attrition.

What is the likely impact of telomere measurement error on our findings? The variation in the correlation between baseline and follow-up LTL measurements suggests substantial variation in measurement error among cohorts [27]. While this error should not cause bias in our estimates of the association between smoking and LTL attrition, it will affect the precision of these estimates [30]. We found no evidence that weighting cohorts according to the correlation between baseline and follow-up measurements caused a significant change in our estimate of the association between smoking and LTL attrition. Indeed, the weighted meta-analysis actually yielded a marginally lower rate of LTL attrition in smokers compared to non-smokers. This result strengthens the evidence against the causation hypothesis, because it implies that no amount of smoking could yield the observed cross-sectional difference in LTL.

Is it possible that some kind of bias is masking a true effect of smoking on telomere attrition? Restricting our dataset to participants that survived to follow-up undoubtedly introduces selection based on mortality (e.g. [26]). Since both smoking and short LTL have been argued to cause earlier mortality, mortality is a collider variable in our analyses. Selection based on the value of a collider is usually discussed in the context of producing spurious associations between independent variables, so-called ‘collider bias’ (e.g. [45]), but collider bias can also potentially mask true associations. For example, by selecting against individuals who die between baseline and follow-up, longitudinal studies could underestimate the effects of smoking on telomere attrition, because they retain only smokers who are resistant to the damaging effects of tobacco smoke. However, this argument fails to explain the substantial difference in baseline LTL between smokers and non-smokers that we observed, even in the studies where the age of participants at baseline was quite advanced. If selection bias is present, it should affect cross-sectional associations as well as measures of attrition, yet we found no change in the difference in LTL between smokers and non-smokers with increasing cohort age (figure 3b). Furthermore, selection based on mortality should be negligible in cohorts in their 20s or 30s, but substantial for cohorts over 60, yet we found no effect of baseline age on the size of the association between smoking and LTL attrition, despite an age range of over 54 years. Taken together, the above findings argue against selection bias masking a true association between smoking and LTL attrition.

We deliberately elected to use raw attrition measures in our analyses as opposed to effect sizes derived from multiple regression models controlling for known sources of variation in telomere attrition rates (such as baseline telomere length). Our rationale for this choice came from recent work showing that controlling for baseline telomere length in multiple regression models of telomere attrition biases the effects of any predictor variables that also correlate with telomere length at baseline (typically age, sex and smoking status; [30]). This latter finding suggests that the published effects of variables such as age, sex and smoking status on telomere attrition are likely to exaggerate the true effect sizes of these variables, raising the probability of type I errors above 5% (see also [46]). This could explain why some of our individual cohorts report significant effects of smoking on LTL attrition [18,19].

A corollary of our decision to use raw LTL and attrition measures in our analyses is that we did not control for potential confounds including age and sex. However, the majority of the nine published studies that we have been able to find reporting effects of smoking on telomere attrition use multiple regression models that control for age and sex [1826]. Considered together, the results of these nine studies support the general conclusions of the current paper: there is strong evidence for an effect of smoking on LTL (six out of eight studies that tested for a difference report that LTL is significantly shorter in smokers), but there is much less evidence for an effect of smoking on LTL attrition (only two out of nine studies report that LTL attrition is significantly faster in smokers) (see [17, table 3]). Furthermore, it is reassuring that our meta-analysis of the cross-sectional effect of smoking on LTL produces a summary effect size for smoking (−0.13) that is similar to that reported in another meta-analysis based on published effect sizes derived from cross-sectional studies that control for potential confounds such as age and sex (−0.11 in [1]). Thus, the conclusions drawn from analyses that do and do not control for age and sex appear very similar and there is no evidence to suggest that controlling for age and sex would alter the conclusions of the current paper.

In the absence of any evidence supporting the hypothesis that smoking causes a sustained increase in the rate of LTL attrition, it is worth considering the alternative hypothesis that selective adoption is occurring. Selective adoption predicts that a difference in LTL between future smokers and non-smokers should exist prior to the start of smoking. Two alternative causal pathways could underlie selective adoption [17]. First, it is possible that telomere shortening could directly cause changes in behaviour. There is emerging evidence that telomere shortening causes changes in regulation of more than 140 genes [47,48], making this idea theoretically possible (although none of the genes identified thus far as regulated by telomere shortening is obviously linked to behaviour, let alone smoking). A second causal pathway yielding selective adoption is that exposure to a third variable both shortens LTL and makes subsequent adoption of smoking more likely. One possibility, that still attributes a causal role to smoke exposure, is that parental smoking both causes early-life LTL attrition and increases a child's probability of starting smoking. A recent study found that telomere loss between birth and young adulthood was positively associated with distance to a major road at the residential address occupied at birth [49], suggesting air pollution as a possible cause of childhood telomere attrition. Thus, it is possible that passive smoking in early life could cause telomere attrition. However, it is not necessary to attribute any causal role to smoke exposure to explain the data in the current paper. We suggest that a plausible third variable supported by substantial existing data is exposure to early-life adversity. Developmental telomere attrition is accelerated by exposure to early-life adversity of various types including family disruption and physical and emotional abuse [5053]. Furthermore, these same sources of early-life adversity are also associated with a greater probability of starting smoking, smoking more and being less likely to quit [5456]. Thus, although childhood LTL has not thus far been examined as a predictor of adult smoking behaviour, there is strong indirect evidence to expect associations to exist. It is worth noting that the available data lead us to predict not only an association between childhood LTL and the presence of adult smoking, but also between childhood LTL and the amount smoked. An association between LTL and amount smoked is often regarded as strong evidence for the causation hypothesis [1]. However, it is now clear that such evidence is equally compatible with selective adoption.

In conclusion, we find no evidence that smoking accelerates the rate of leucocyte telomere attrition in adults. Our findings should prompt more critical appraisal of data underlying the claim that smoking is the most important, ‘broad range’ ageing accelerator [5,6]. Where these data come from cross-sectional studies, and in vivoexperimental studies are lacking, we suggest that selective adoption should be considered as an alternative explanation for associations between smoking and biomarkers of ageing such as telomere length.

Our findings have consequences for how measures of telomere length are used in human epidemiology and behavioural ecology. Under the currently prevailing view that certain types of behaviour cause accelerated telomere attrition, measures of telomere length can be used to identify those behaviours that are most harmful and those that are protective [57]. Changes in telomere dynamics could also potentially be used to monitor the somatic consequences of behaviour change (e.g. the positive effects of quitting smoking). However, if we are correct, and selective adoption turns out to be an explanation for observed associations with telomere length, then we need to reinterpret shorter telomeres as a relatively static biomarker as opposed to as a dynamic consequence of current adult behaviour.

As a final note, although we found no evidence that smoking accelerates the rate of leucocyte telomere attrition, our results do not preclude the many other well-established negative effects of smoking on human health and longevity. We chose to focus on smoking in the current paper simply because there are more data available on the associations between smoking and telomere length than for any other behaviour [17]. Our intention was to question prevalent assumptions in the telomere dynamics literature concerning the mechanisms underlying associations between behaviour and telomere length, rather than to question the damaging effects of smoking.




Edited by Nigeria Custom Officer, 24 June 2019 - 10:27 AM.

#2 QuestforLife

  • Member
  • 889 posts
  • 464
  • Location:UK

Posted 25 June 2019 - 09:32 AM

Maybe the relationship between LTL and smoking is not linear, i.e. you lose some length initially, and then your body's defensive systems counteract further attrition.


Using LTL is not very informative anyway; what really matters is how smoking affects the telomere length of lung fibrobalsts. Now that would really tell us something.

  • Good Point x 1

Also tagged with one or more of these keywords: telomere

0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users