Association of Thyroid Hormone Therapy With Quality of Life and Thyroid-Related Symptoms in Patients With Subclinical Hypothyroidism: A Systematic Review and Meta-analysis

Question  Among patients with subclinical hypothyroidism, is the use of thyroid hormone therapy associated with improvements in general quality of life or thyroid-related symptoms?

Findings  In this meta-analysis of 21 randomized clinical trials including 2192 participants with subclinical hypothyroidism, thyroid hormone therapy was not significantly associated with improvements in general quality of life (standardized mean difference, −0.11) or thyroid-related symptoms (standardized mean difference, 0.01).

Meaning  These findings do not support the routine use of thyroid hormone therapy in adults with subclinical hypothyroidism.

Importance  The benefit of thyroid hormone therapy for subclinical hypothyroidism is uncertain. New evidence from recent large randomized clinical trials warrants an update of previous meta-analyses.

Objective  To conduct a meta-analysis of the association of thyroid hormone therapy with quality of life and thyroid-related symptoms in adults with subclinical hypothyroidism.

Data Sources  PubMed, EMBASE, ClinicalTrials.gov, Web of Science, Cochrane Library, CENTRAL, Emcare, and Academic Search Premier from inception until July 4, 2018.

Study Selection  Randomized clinical trials that compared thyroid hormone therapy with placebo or no therapy in nonpregnant adults with subclinical hypothyroidism were eligible. Two reviewers independently evaluated eligibility based on titles and abstracts of all retrieved studies. Studies not excluded in this first step were independently assessed for inclusion after full-text evaluation by 2 reviewers.

Data Extraction and Synthesis  Two independent reviewers extracted data, assessed risk of bias (Cochrane risk-of-bias tool), and evaluated the quality of evidence (GRADE tool). For synthesis, differences in clinical scores were transformed (eg, quality of life) into standardized mean differences (SMDs; positive values indicate benefit of thyroid hormone therapy; 0.2, 0.5, and 0.8 correspond to small, moderate, and large effects, respectively). Random-effects models for meta-analyses were applied.

Main Outcomes and Measures  General quality of life and thyroid-related symptoms after a minimum follow-up of 3 months.

Results  Overall, 21 of 3088 initially identified publications met the inclusion criteria, with 2192 adults randomized. After treatment (range, 3-18 months), thyroid hormone therapy was associated with lowering the mean thyrotropin value into the normal reference range compared with placebo (range, 0.5-3.7 mIU/L vs 4.6 to 14.7 mIU/L) but was not associated with benefit regarding general quality of life (n = 796; SMD, −0.11; 95% CI, −0.25 to 0.03; I²=66.7%) or thyroid-related symptoms (n = 858; SMD, 0.01; 95% CI, −0.12 to 0.14; I²=0.0%). Overall, risk of bias was low and the quality of evidence assessed with the GRADE tool was judged moderate to high.

Conclusions and Relevance  Among nonpregnant adults with subclinical hypothyroidism, the use of thyroid hormone therapy was not associated with improvements in general quality of life or thyroid-related symptoms. These findings do not support the routine use of thyroid hormone therapy in adults with subclinical hypothyroidism.

Subclinical hypothyroidism, defined as elevated thyrotropin in combination with normal-range free thyroxine,¹ is common.^2,3 According to the NHANES III report,⁴ an estimated 13 million people have subclinical hypothyroidism in the United States. The prevalence is higher in women and in older people.^2,3 Subclinical hypothyroidism is often treated with thyroid hormones (levothyroxine),⁵ particularly when it co-occurs with symptoms potentially attributable to hypothyroidism, such as tiredness, constipation, and unexplained weight gain.⁵

Relatively limited evidence exists from randomized clinical trials (RCTs) to guide therapy of subclinical hypothyroidism. Systematic reviews have been inconclusive and clinical practice guidelines have varied regarding recommendations for managing subclinical hypothyroidism.^6-10 Two large randomized trials of levothyroxine therapy in patients with subclinical hypothyroidism were recently completed.^11,12 This meta-analysis and systematic review incorporated recent trials and evaluated whether thyroid hormone therapy was associated with improved symptoms and other benefits in nonpregnant adults with subclinical hypothyroidism.

We followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement¹³ and published the protocol of this systematic review in the PROSPERO database (identifier: CRD42017055536).

We considered randomized trials that included nonpregnant adults with subclinical hypothyroidism. Subclinical hypothyroidism was defined as a thyrotropin level above the reference range in combination with a free thyroxine level within the reference range (according to center-specific reference ranges). The intervention had to consist of thyroid hormone therapy (triiodothyronine, thyroxine, or a combination of both) for at least 1 month, with a minimum follow-up of 3 months. The control group had to receive either placebo or no therapy. To be included, studies had to report quantitative data for at least 1 of the study’s primary or secondary outcomes: general quality of life, thyroid-related quality of life/hypothyroid symptoms, depressive symptoms, fatigue/tiredness, cognitive function, pain, muscle strength, blood pressure, body mass index, cardiovascular events (myocardial infarction, stroke, revascularization), mortality, or adverse effects (hyperthyroidism due to overdosing). Data had to be reported with effect estimates and measures of precision (standard deviations or standard errors). The primary outcomes were general quality of life and thyroid-related quality of life/hypothyroid symptoms, whereas depressive symptoms, fatigue/tiredness, cognitive function, pain, muscle strength, blood pressure, body mass index, cardiovascular events, mortality, and adverse effects were secondary outcomes. Studies that included only patients with subclinical hypothyroidism in combination with another specific condition (eg, patients with diabetic nephropathy and subclinical hypothyroidism) were excluded because this type of study population is not representative of most patients with subclinical hypothyroidism. We excluded studies that exclusively enrolled pregnant women or women who wanted to become pregnant. Pseudorandomization (eg, pre-post comparisons) did not qualify for inclusion.

We searched MEDLINE, EMBASE, Web of Science, Cochrane Library, CENTRAL, Emcare, and Academic Search Premier from inception until July 4, 2018, in cooperation with a trained librarian. Search terms were adapted according to the syntax of each specific database, and no language restrictions were applied. We searched trial registries on ClinicalTrials.gov for upcoming and not yet published trials on this research topic and asked authors for the status of the trial if not published yet. We screened references of key articles for additional potentially relevant articles. Details of the search strategy are presented in eAppendix 1 in Supplement 1.

Two researchers (M.S. and M.D.M.) evaluated eligibility independently based on titles and abstracts of all studies retrieved in the electronic search. Studies not excluded in this first step were independently assessed for inclusion after full-text evaluation by 2 reviewers (M.F. and M.S.). We manually screened bibliographies of the included studies as well as guidelines and review articles for additional studies. Discrepancies were resolved by consensus among the study team.

A standard data extraction form was used, adapted from a template suggested by Cochrane (eAppendix 2 in Supplement 1).¹⁴ Two researchers (M.S. and E.M.) independently extracted bibliographic details, funding source, eligibility criteria, information about the study population and setting, study design, risk of bias, intervention/control intervention, and results and independently evaluated the quality of evidence using the GRADE tool.¹⁵ If a study reported more than 1 outcome measure for a specific outcome domain (eg, more than 1 cognition test to assess cognitive function), we chose measures that were most relevant to the largest number of patients, including international usage. This was determined by consensus among the study team. As an example, Parle et al¹⁶ reported 5 different cognition tests. We analyzed results from the Mini-Mental State Examination because it is used worldwide and because it is a more general assessment of cognitive function than more specific alternative tests, such as the Trail Making Test. When a study mentioned an outcome of interest without providing estimates (eg, the study reported no difference in body mass index between the intervention and the control group without providing data on mean differences and standard deviations), we contacted the author for the data. If studies reported results for an outcome at multiple time points during the intervention (eg, body mass index at 6 and 12 months), only the most recent measurement was used in statistical analyses. Data were extracted in duplicate by 2 independent reviewers (M.S. and E.M.) and differences were resolved by consensus.

Study results were presented separately for each outcome with estimates as reported in the original publication and transformed into standardized mean differences (SMDs) when different scales were used for the same outcome domain. We coded SMDs such that positive values indicated benefit of thyroid hormone therapy, with 0.2, 0.5, and 0.8 corresponding to small, moderate and large effects, respectively.¹⁷ In contrast, for body mass index and blood pressure, negative values indicated a benefit of thyroid hormone therapy. For estimations of treatment effects, we used mean values and their standard deviations at the end of treatment in both groups, assuming balanced baseline values due to the randomized designs.

For outcomes on which studies reported treatment effects at different time points, we included only the estimate at the most recent follow-up time point, thereby avoiding counting a study twice in a formal meta-analysis. Overall results were calculated using random-effects models unless fewer than 5 studies were included for a meta-analysis because in that case, the between-study variance could not be estimated reliably, so a fixed-effects analysis was performed. For better clinical interpretation, overall SMDs were also back-transformed to one original scale according to a method proposed by the Cochrane Collaboration¹⁷ for general quality of life, thyroid-related quality of life/hypothyroid symptoms, depressive symptoms, cognitive function, and muscle strength. Heterogeneity was assessed visually with forest plots and quantified with I² (low: 0%-40%; moderate: 40%-75%; and high: >75%). If substantial heterogeneity existed and a sufficient number of publications was available (n = 10), we aimed to explore potential sources of heterogeneity in protocol prespecified subgroup analyses (eg, restricting the analysis to high-quality studies). In addition, a post hoc sensitivity analysis was performed with the aim to evaluate heterogeneity after excluding studies showing a statistically significant benefit of placebo treatment. If a sufficient number of publications was available (n = 10), publication bias was assessed via funnel plots (visually) and more formally with the Egger test.¹⁸ Statistical significance testing was 2-sided and P<.05 was considered statistically significant. All analyses were conducted with Stata, release 14 (StataCorp).

The systematic literature search retrieved 3086 studies, and 2 additional studies were retrieved after searching references of key articles. After removing 1438 duplicates, 2 reviewers (M.S. and M.D.M.) independently screened 1650 unique articles for potential eligibility based on title and abstract. Forty-nine potentially eligible studies were evaluated in full text independently by 2 reviewers (M.F. and M.S.). Among these, 25 studies did not meet the inclusion criteria. Three additional studies were excluded because of data presentation problems^19-21; for 2 of these articles, the authors indicated that data were no longer available^19,20; for one, the author could not be reached²¹ (eTable 1 in Supplement 1).

Twenty-one studies met the inclusion criteria^{11,12,16,22-39} (eFigure in Supplement 1). Among the 21 studies, a total of 2192 adults were randomized (Table). Study sizes ranged from 20 to 737 participants; mean ages ranged from 32 to 74 years; percentages of women ranged from 46% to 100%; and baseline mean thyrotropin values ranged from 4.4 to 12.8 mIU/L. Two studies (99 participants) had a mean baseline thyrotropin level greater than 10 mIU/L.^22,23 Seven studies provided information about hypothyroid symptoms at baseline, and in these studies, the burden of symptoms was mild to moderate (Table).^12,22-27 In the thyroid hormone therapy groups, mean thyrotropin levels at the end of follow-up ranged between 0.5 and 3.7 mIU/L (eTable 2 in Supplement 1), indicating that treatment was associated with normalization of thyrotropin levels. In contrast, mean thyrotropin in the placebo/no intervention groups remained elevated at the end of follow-up, ranging from 4.6 to 14.7 mIU/L (eTable 2). The duration of the intervention (thyroid hormone therapy or placebo/no therapy) ranged from 3 months to 18 months. Three studies compared thyroid hormone therapy with no intervention and the other studies compared thyroid hormone therapy with placebo.^11,28,29 Two studies were supported by industry (Table).^22,30

Thyroid hormone therapy was not associated with benefit for either of the 2 primary outcomes. Four studies including 796 participants evaluated general quality of life (SMD, −0.11; 95% CI, −0.25 to 0.03; I²=66.7%) (Figure 1).^12,26,27,31 It is estimated that on the Euro Quality of Life 5 Dimensions Questionnaire (range, −0.59 to 1.00; higher scores indicate better quality of life), this SMD would represent a difference of 0.02 (95% CI, −0.01 to 0.05) in favor of placebo. Four studies including 858 participants evaluated thyroid-related quality of life/hypothyroid symptoms (SMD, 0.01; 95% CI, −0.12 to 0.14; I²=0.0%) (Figure 1).^12,22,31,32 It is estimated that on the ThyPRO hypothyroid symptoms score (range, 0-100; higher scores indicate more hypothyroid symptoms), this SMD would represent a difference of 0.18 (95% CI, −2.10 to 2.45) in favor of levothyroxine. Similarly, thyroid hormone therapy was not associated with benefit regarding the secondary outcomes. For depressive symptoms (4 studies; 278 participants), the SMD was −0.10 (95% CI, −0.34 to 0.13; I²=0.0%) (Figure 1)^16,24,26,31; on the Hospital Anxiety and Depression Scale (range, 0-21; higher scores indicate worse depressive symptoms), this SMD would represent a difference of 0.28 (95% CI, −0.36 to 0.95) in favor of placebo. For cognitive function (4 studies; 859 participants), the SMD was 0.09 (95% CI, −0.05 to 0.22; I²=14.7%) (Figure 2)^12,16,25,26; on the Letter-Digit Coding Test (range, 0 or higher [no upper limit]; higher scores indicate better cognitive function), this SMD would represent a difference of 1.01 (95% CI, −0.56 to 2.46) in favor of levothyroxine. For muscle strength (2 studies; 695 participants), the SMD was 0.1 (95% CI, −0.1 to 0.2; I²=0.0%) (eTable 4 in Supplement 1)^12,31; in hand-grip strength, this SMD would represent a difference of 1.12 kg (95% CI, −1.12 to 2.24 kg) in favor of levothyroxine. Systolic blood pressure (8 studies; 1372 participants) was −0.7 mm Hg (95% CI, −2.6 to 1.2 mm Hg; I²=0.0%) (Figure 3).^{11,12,29,32-36} Body mass index (calculated as weight in kilograms divided by height in meters squared) (15 studies; 1633 participants) was 0.2 (95% CI, −0.4 to 0.8; I²=45.5%) (Figure 4).^{11,12,23,27-30,32-39} Only the TRUST trial (the largest included study, with 737 participants randomized) evaluated fatigue/tiredness, cardiovascular events, mortality, and adverse effects.¹² No beneficial or harmful association between thyroid hormone therapy and these outcomes was reported (Figure 1 and eTables 2 and 4 in Supplement 1). No study included pain as an outcome. Detailed results are summarized in eTables 2 and 4. Subgroup analyses were not performed because the number of studies for a single outcome was too small and/or there was low to moderate heterogeneity such that no exploration was indicated. The meta-analyses for general quality of life and body mass index showed moderate heterogeneity (I²=66.7% and I²=45.5%, respectively). Therefore, post hoc sensitivity analyses were performed excluding studies showing a statistically significant benefit of placebo.^23,27,30 Results remained similar, but heterogeneity was lower (for general quality of life, SMD, −0.08 [95% CI, −0.22 to 0.06; I²=34.7%]; for body mass index, −0.2 [95% CI, −0.6 to 0.2; I²=1.6%]). We did not formally assess publication bias. Based on the negative results, there was no indication that positive studies were published while negative studies remained unpublished.

The overall quality of the 21 included studies was good, with only 9 of 126 items judged to be at high risk of bias (eTable 3 in Supplement 1); 2 trials had low risk of bias for all criteria,^12,22 including the largest and most recent trial,¹² and only 1 trial, the second largest and second most recent, had a high risk of bias in 3 of 6 domains.¹¹ Accordingly, the quality of evidence assessed with the GRADE tool was high regarding the main outcomes of general quality of life and thyroid-related symptoms, as well as regarding muscle strength, blood pressure, and body mass index (eTable 4 in Supplement 1). The quality of evidence was moderate for depressive symptoms, fatigue/tiredness, cognitive function, and adverse effects, whereas it was low for cardiovascular events and mortality (eTable 4).

In this systematic review and meta-analysis of RCTs in nonpregnant adults with subclinical hypothyroidism, thyroid hormone therapy was not associated with benefit regarding general quality of life, thyroid-related symptoms, depressive symptoms, fatigue/tiredness, cognitive function, muscle strength, blood pressure, or body mass index.

Compared with prior systematic reviews and meta-analyses published between 2007⁸ and 2015,⁹ this meta-analysis included 2 recent randomized trials, which were the largest trials published to date on this topic.^11,12 Overall, the quality of evidence reported herein was moderate to high. Quality of evidence was high regarding the primary outcomes of this review (general quality of life and thyroid-related symptoms). Results of this review consistently demonstrated no association of thyroid replacement therapy with improved outcomes, including a relatively large number of diverse outcomes. Most outcomes, except cardiovascular events and mortality, had narrow confidence intervals. In addition, this meta-analysis focused on patient-centered outcomes such as quality of life and fatigue, which are the most common symptoms that prompt therapy in general practice.⁵⁶

Although current guidelines are at first sight cautious with treatment recommendations, more than 90% of persons with subclinical hypothyroidism and a thyrotropin level of less than 10 mIU/L would actually qualify for treatment.^6,10,57 However, results of this meta-analysis are not consistent with these guideline recommendations. In addition to absence of an association of thyroid hormone therapy with improved outcomes, thyroid hormone therapy is associated with adverse effects when overtreatment occurs.^5,58,59

This study has several limitations. First, the RCTs included in this meta-analysis used different questionnaires and/or tests for a given outcome in combination with different treatment durations (eg, 4 different cognitive tests in the 4 studies examining cognitive function, with treatment durations ranging from 3 to 18 months). However, little heterogeneity across the study results was observed except for general quality of life and body mass index. For these outcomes, heterogeneity resulted from 3 studies that showed a statistically significant benefit of placebo.^23,27,30 After excluding these studies in post hoc sensitivity analyses, thyroid hormone therapy remained unassociated with benefit for general quality of life and body mass index, and heterogeneity was lower. Therefore, it seems unlikely that this meta-analysis missed a potential beneficial association between thyroid hormone therapy and any outcome analyzed due to inappropriate pooling of overly heterogeneous studies.

Second, only 1 RCT reported on major adverse cardiovascular events. Therefore, definitive evidence is lacking regarding the association of therapy for subclinical hypothyroidism with reduced cardiovascular event rates.¹² Third, RCTs that reported results only qualitatively were excluded from analyses. Fourth, mean thyrotropin values at baseline were less than 7.0 mIU/L in 11 of 21 included RCTs, and only 2 RCTs examined participants with a mean baseline thyrotropin level higher than 10 mIU/L.^22,23 Therefore, the current findings may not be generalizable to people with subclinical hypothyroidism and a thyrotropin level higher than 10 mIU/L. Fifth, the highest mean age in the included studies was 74 years.^12,16 Therefore, these results may not be generalizable to people older than 80 years. Sixth, only 7 of 21 trials (33%) reported hypothyroid symptoms at baseline, and the burden of symptoms was mild to moderate in these trials. The other 14 trials did not describe symptoms at baseline. It is possible that the subgroup of people with subclinical hypothyroidism and a high burden of symptoms would still benefit from treatment. Seventh, patients with subclinical hypothyroidism and “severe” symptoms of hypothyroidism may be underrepresented in clinical trials because they may be treated immediately with levothyroxine and not included in clinical trials.⁶⁰ Therefore, results reported herein may not be generalizable to patients with subclinical hypothyroidism who have severe symptoms. Eighth, 2 RCTs (n = 831) included participants with a mean age older than 65 years.^12,16 Their mean thyrotropin level at baseline was between 6.0 and 7.0 mIU/L. Given the possibility that the upper thyrotropin reference limit may increase with age,⁶¹ the 2 studies may have included older individuals with mildly elevated thyrotropin levels who do not represent subclinical hypothyroidism, although current international guidelines do not use different thyrotropin levels according to age to define subclinical hypothyroidism.^6,10,62 However, this phenomenon may have biased the results toward the null. Ninth, it is possible that thyroid hormone therapy is associated with benefit regarding outcomes that were not examined in this meta-analysis (eg, carotid intima-media thickness, various lipid fractions). Tenth, it is possible that treatment of subclinical hypothyroidism may be beneficial in study populations not included in these analyses (eg, patients with subclinical hypothyroidism and renal impairment). Eleventh, the largest RCT to date¹² contributed substantially to the results of this meta-analysis because of the large sample size relative to the other trials (737 of 2192 participants [33.6%]). However, the mean age of participants in the largest trial was 74 years, while the mean age of participants in the studies included herein ranged from 32 to 74 years.

Among nonpregnant adults with subclinical hypothyroidism, the use of thyroid hormone therapy was not associated with improvements in general quality of life or thyroid-related symptoms. These findings do not support the routine use of thyroid hormone therapy in adults with subclinical hypothyroidism.

Corresponding Author: Martin Feller, MD, MSc, Department of General Internal Medicine, Inselspital, Bern University Hospital, University of Bern, Bern 3010, Switzerland (martin.feller@insel.ch).

Accepted for Publication: August 30, 2018.

Author Contributions: Drs Feller and Dekkers had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs Feller and Dekkers had final responsibility for the decision to submit for publication.

Concept and design: Feller, Snel, Ford, Jacobijn, Stott, Westendorp, Rodondi, Dekkers.

Acquisition, analysis, or interpretation of data: Feller, Snel, Moutzouri, Bauer, de Montmollin, Aujesky, Ford, Kearney, Mooijaart, Quinn, Stott, Westendorp, Rodondi, Dekkers.

Drafting of the manuscript: Feller, Snel, Quinn, Dekkers.

Critical revision of the manuscript for important intellectual content: Moutzouri, Bauer, de Montmollin, Aujesky, Ford, Jacobijn, Kearney, Mooijaart, Quinn, Stott, Westendorp, Rodondi.

Statistical analysis: Feller, Ford, Quinn, Dekkers.

Obtained funding: Stott, Westendorp, Rodondi.

Administrative, technical, or material support: Feller, Snel, Moutzouri, de Montmollin, Aujesky, Rodondi.

Supervision: Feller, Jacobijn, Kearney, Mooijaart, Stott, Westendorp, Rodondi, Dekkers.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Quinn reported being an investigator on one of the trials included in this review (funded by the European Union 7th Framework Program scheme) and also an investigator on an RCT in which the study drug was provided by Merck. Dr Stott reported receipt of a research grant from the European Union 7th Framework Program scheme. No other disclosures were reported.

Funding/Support: The Swiss National Science Foundation funded this study (grant SNSF 320030_172676 to Dr Rodondi).

Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.

Previous Presentations: This article was presented orally in part at the European Congress of Endocrinology in Barcelona, Spain, May 19-22, 2018, and at the annual meeting of the Swiss Society of General Internal Medicine in Basel, Switzerland, May 30–June 1, 2018.

Data Sharing Statement: See Supplement 2.

Additional Contributions: We thank Fahime Ghanbari, MD, Department of General Internal Medicine, Inselspital, Bern University Hospital, University of Bern, Switzerland, for translating an article written in Persian. We thank Jan W. Schoones, MA, Leiden University Medical Center, Leiden, the Netherlands, the librarian responsible for the literature search. We also thank Salman Razvi, MD, Institute of Genetic Medicine, Newcastle University, Newcastle Upon Tyne, England, who provided data for a study included in the analysis. None of the persons mentioned received compensation.