Subclinical Hypothyroidism And The Risk Of All-cause Mortality: A Meta-analysis Of Cohort Studies

BackgroundSubclinical hypothyroidism, defined as a serum TSH concentration higherthan the normal range with normal levels of thyroid hormones, is a mild form ofhypothyroidism. The etiology of subclinical hypothyroidism is similar to that ofovert hypothyroidism and is most often caused by autoimmune chroniclymphocytic thyroiditis. Other causes of elevated TSH levels must be excluded;mild elevations in TSH are commonly seen in recovery from nonthyroidalillness. Female sex, advanced age, and the presence of elevated thyroidperoxidase antibodies are all risk factors for subclinical hypothyroidism.subclinical hypothyroidism has adverse effects on cardiovascular risk factorsand cardiovascular function, including elevated total cholesterol and LDL-Clevels, impaired endothelial function and altered ventricular structure etal.Whether subclinical hypothyroidism leads to adverse clinical outcomes andshould be treated remains controversial.Some studies showed positive resultsbut others did not.A meta-analyses assessing the relationship between CVD, cardiovascularmortality and all-cause mortality has been performed previously, showing thatsubclinical hypothyroidism was not associated with all-cause mortality. Sincethe latest meta-analysis had been performed, several new large prospectivestudies on this issue have been published. Now data is sufficient to update themeta-analysis.Objective1. To investigate the relationship between subclinical hyothyroidism andall-cause mortality.2. To assess the stability of the results by performing meta-regression andsensitivity analysis. Methods:1. Search strategy and Study selection. PubMed and CNKI database weresearchedfor relevant studies assessing the association between subclinicalhypothyroidism and all-cause mortality. Studies were considered eligible if theymet the following criteria:1) the study design was a cohort one;2) thyroidfunction was measured at baseline;3) the outcome of interest was all-causemortality; and4) relative risk (RR) or hazard ratio (HR) and the corresponding95%confidence interval (CI) were reported.2. Data extraction. The following data were abstracted: the first author’sname; thepublication year; the country of origin; the number, mean age, and sex ofthe participants; the definition of subclinical hypothyroidism and cardiovascularevents, based on the information as provided in the primary studies; theprevalence of subclinical hypothyroidism; the study design details, includingpopulation source, starting year of study, and study duration; whether thereported RR or HR was adjusted for potential confounders.3. Statistical analyses. The study-specific maximally adjusted RR or HRwas used to compute a summary RR and its95%CI. Heterogeneity acrossstudies was tested by using the Q and I2statistic. The combined risk estimateswere computed using either fixed-effects models or random-effects models withthe presence of heterogeneity. Meta-regression was used to explore possiblecontributors to heterogeneity. We also investigated the influence of a singlestudy on the overall risk estimate by omitting1study in each turn. Potentialpublication bias was assessed by Egger’s test and Begg’s test to evaluatepublication bias. All analyses were performed using STATA version11.0. A Pvalue <0.05was considered statistically significant.Results:1. Study characteristics and Quality assessment. The characteristics of20cohort studies were included. These studies were published between2001and2013. The mean length of follow-up ranged from2.5to20years. In most studiesparticipants were recruited from communities. All studies used second or thirdeneration method for TSH assay. Most studies had a TSH cutoff value of4.0to5.0mU/L. Most studies adjusted for a group of conventional risk factors for all-cause mortality.2. Subclinical hypothyroidism and risk of All-cause mortality. Twentystudies were pooled to calculate the overall RR for all-cause mortality. Asignificant association was observed when combining all studies (RR:1.134，95%CI:1.020-1.260, P=0.020) with presence of heterogeneity(P=0.000, I2=60.1%).3. Exploration of the sources of heterogeneity. Exploratory univariatemeta-regression was performed with age, publication year, length of follow-up,and sample size as the potential sources of between-study heterogeneity. Noneof variables abovementioned was identified as a potentially important source ofbetween-study heterogeneity in combining RRs for total mortality.4. Sensitivity analysis. No evidence of any individual study havingexcessive influence on the pooled effect in combining risk estimates ofsubclinical hyperothyroidism for all-cause mortality was observed.5. Publication bias. The Begg’s funnel plot and Egger’s regression test didnot indicate evidence of publication bias.Conclusion1. Subclinical hypothyroidism was associated with an increased risk ofall-cause mortality.2. A certain degree of heterogeneity existed and the result should beinterpreted with caution.