A 58-year-old white woman presented to the internal medicine clinic for a 3-month follow-up visit. Her medical history was significant for dyslipidemia and hypertension. Daily medications included a multivitamin, lisinopril 10 mg, and simvastatin 80 mg (a dose that had been increased at the previous visit). Family history was significant for hypertension and dyslipidemia; the woman's mother died of an MI at age 62 years. The patient exercised moderately three to four times per week for 30 to 45 minutes each time, and she had followed the Dietary Approaches to Stop Hypertension (DASH) diet for several years. She denied illicit drug or alcohol use and had never smoked. At a height of 5 ft 6 in and weight of 173 lb, her BMI was 27.9 kg/m2. Vital signs were as follows: BP, 138/85 mm Hg; heart rate, 92 beats per minute; respiration rate, 18 breaths per minute; and temperature 97.9°F. Findings on physical examination were remarkable only for cool, dry skin. Results of a fasting laboratory panel were significant for thyroid-stimulating hormone (TSH), 12.0 mIU/L; total thyroxine (T4), 0.9 µg/dL; total cholesterol, 250 mg/dL; LDL, 175 mg/dL; HDL, 35 mg/dL; and triglycerides, 200 mg/dL. An assay for antithyroid antibodies was negative. These recent results were consistent with laboratory findings obtained 3 months earlier.
CLINICAL QUESTION
In a woman with subclinical hypothyroidism (SCH) and dyslipidemia, does treatment with levothyroxine reduce the risk of developing cardiovascular disease (CVD)?
SEARCH CRITERIA
AND RESULTS
Data for this article were retrieved from Medline and the Cochrane Database of Systematic Reviews using the key terms subclinical hypothyroid AND thyroxine AND cardiovascular. The search, which was conducted in July 2011, utilized the Clinical Queries function and was limited to articles published within the past 5 years, humans, and English language. To ensure consideration of all pertinent articles, a secondary search using the terms subclinical hypothyroidism, cardiovascular disease, and levothyroxine was also performed. In total, 37 articles were recalled through the search strategy. Three meta-analyses were available for evaluation related to the clinical question and represented the highest-level evidence for this therapy question. Of the three articles, the meta-analysis by Rodondi and colleagues1 was selected for final review because it stratified outcome measures based on TSH levels; reflected the most comprehensive, up-to-date investigation of the subject; and included the studies that were identified through the other meta-analyses.
EVALUATING THE EVIDENCE
Rodondi and colleagues formed a Thyroid Studies Collaborative to conduct a meta-analysis using individual participant data from large prospective cohort studies that investigated the association between subclinical hypothyroidism and coronary heart disease (CHD). The goal of this collaborative was to reconcile conflicting data and define the influence of age, TSH levels, and preexisting CVD on outcomes in patients with SCH. The databases of Medline and EMBASE were searched without language restrictions for full-text studies published between 1950 and 2010 that included baseline thyroid function testing and a comparison group, had adequate follow-up, and reported clinically meaningful outcome measures (eg, CHD events, mortality). Search terms were broadly defined by using the Medical Subject Headings (MeSH) term in Medline and similar terms in EMBASE. The 13 authors who participated in data acquisition used a previously defined search filter and did not explicitly partner with medical librarians to improve their search strategy and terms. Of the 12 studies that met inclusion criteria, 11 study investigators agreed to release their individual patient data for analysis (see Table 1 in Rondondi and colleagues1). A formal validity assessment of the quality of each study was conducted based on methods of adjudication and ascertainment, consideration of confounders, and thoroughness of follow-up.
In order to limit outcome heterogeneity, the Framingham risk score was used as a model to define the primary end points (CHD events, CHD mortality, and total mortality). To further maximize comparability between studies, SCH was uniformly defined as serum TSH levels of 4.5 to 20 mIU/L and T4 levels of 0.7 to 1.7 µg/dL. Euthyroidism was defined as serum TSH levels of 0.5 to 4.5 mIU/L and T4 levels of 0.7 to 1.7 µg/dL. To assess the degree of heterogeneity across studies, the researchers used the I2 statistic, which indicates variability due to heterogeneity rather than chance. An I2 greater than 50% suggests moderate heterogeneity.1 Because of the moderate degree of heterogeneity in CHD events (I2 = 59%) and total mortality (I2 = 66%), a random-effects model was used to calculate pooled estimates for each outcome. The presence of publication bias was assessed via the Egger test and funnel plots, which revealed little evidence of publication bias related to CHD events and no publication bias for other outcome measures.
All participants were included in the primary analysis, which was adjusted for age, sex, systolic BP, smoking, total cholesterol, and diabetes (see Figure 1 in Rondondi and colleagues1). In this unstratified analysis, participants with elevated TSH levels were compared with those in the euthyroid control group. Results showed no statistically significant increased risk for CHD events (hazard ratio [HR], 1.18; 95% confidence interval [CI], 0.99-1.42), CHD mortality (HR, 1.14; CI, 0.99-1.32), or total mortality (HR, 1.09; CI, 0.96-1.40). The study assessed the impact of common risk factors (except for family history of CHD) on the outcome measures. Thus, the study conclusions are generalizable to the population of patients with SCH unless there is a history of premature CHD in their family. Fixed-effect, subgroup, and sensitivity analyses stratified by age, sex, race, TSH concentration, and preexisting CVD were also performed. These secondary analyses included only complete data sets and excluded participants who used thyroid hormone replacement at any point during the study. Loss to follow-up across all studies was 5% or less.