Studies included in the meta-analysis must meet the following criteria: 1) Subjects must be diagnosed with SCH that was defined as elevated TSH and normal FT4. 2) Subjects must be older than or equal to 60 years of age. 3) Interventions should be thyroid hormone therapy (e.g. levothyroxine).

Outcomes can be various aspects of older SCH patients, such as quality of life, cognitive function, depression, lipids profile, renal function, cardiovascular events, etc. We combined the same outcomes for meta-analysis.

We searched Embase, Web of Science, Cochrane Library, Pubmed and chinese databases including Wanfang, Weipu and China National Knowledge Infrastructure (CNKI) from inception until December 21, 2021. In order to ensure the quality of Chinese literature, we included articles from Chinese core journals and dissertation, which have undergone rigorous peer review and have a high degree of recognition. Subjects were human, and the published languages mainly included English and Chinese. To reduce selection bias, we also searched for articles published in German, Spanish, French, Korean and Japanese. The subject words included Hypothyroidism, and Thyroxine. The random words included Subclinical, Mild, Aged, Elder, Older, randomized controlled trial and so on ( Supplementary Text S1 ). In order not to miss possible literature, we manually searched references of relevant papers obtained from systematic searches, as well as relevant conferences and registered clinical trials.

Two reviewers independently screened eligible articles and extracted relevant data from the included studies. If there was disagreement, a third reviewer would join to conclude the final results through a discussion. The first step was to screen the articles based on their titles and abstracts, and the second step was to read the full-text of selected articles for the final eligible literature.

The information extracted from each eligible literature included authors, year of publication, country, number and age of participants, follow-up time, outcomes and so on. For the ages of participants in some included literature, the mean and standard deviation were not used to record, we will use mathematical methods to convert. If a study measured outcomes at different follow-up time points, we would extract data from the time point most frequently used in the other studies with the same outcomes, or the data closest to that time point. For unavailable result data, we would get the correct data by sending an email to the author. Quality assessment was conducted according to the Cochrane Handbook for Systematic Reviews of Interventions Version 5.0.1. (Higgins JPT and Green S, n.d.).

The software RevMan5.3 was used to perform statistical analysis and generate forest plots in this meta-analysis study. The outcomes of some articles were continuous variables, such as quality of life, depression, etc., and standardized mean difference (SMD) was used as a summary analysis when measured with different tools, or using mean difference (MD) when the same tool was used. When the outcomes were dichotomous variables, such as cardiovascular events, odds ratio (OR) was used. The chi-square test and I² test were used to assess the magnitude of the heterogeneity among the included studies. A fixed effects model was used if heterogeneity is not significant (P ≥ 0.1 or I² ≤ 50%), otherwise, a random effects model was applied (P < 0.1 or I² > 50%). We considered the effect to be statistically significant when the P value was less than 0.05.

A subgroup analysis was conducted. On the one hand, both prospective randomized controlled trials and retrospective case-control studies were included in the meta-analysis, and on the other hand, some included studies had significantly different follow-up times. In order to analyze the effects of the above factors, we excluded those retrospective studies and studies with large differences in follow-up time to analyze the outcomes again, and saw if there were differences.

After systematic search of 7 databases and manual search, a total of 2640 papers were yielded. Three hundred and twenty-one duplicates were removed, 2150 papers were removed by title and abstract, 156 papers were removed after reading the full text, and finally 13 eligible papers were included ( Figure 1 ). About 5000 participants were included in the meta-analysis and systematic review, all of whom were over 60 years of age. Individuals in the Stuber’s study (15) were all from the Stott’s study and had no new findings, so Stuber’s study were excluded. Individuals in the Gencer’s (18), Wildisen’s (16) and Gonzalez Rodriguez’s (12) studies, although also all from the Stott’s study, were continued to be included because these three articles analyzed different results. Individuals in the Mooijaart’s study (14) were only partially from the Stott’s study and another part from the Du Puy’s study (27), which was not an exact duplicate of the Stott’s study, so it was continued to be included in this mata-analysis. A total of 4 Chinese literatures (28–31) were included, of which 3 (28, 30, 31) were from Chinese core journals and the other (29) was a master’s dissertation, which was also included due to its low risk of bias. Except for English and Chinese literature, literature published in other languages (including German, Spanish, French, Korean and Japanese) in the above databases did not meet the inclusion criteria. In addition, no studies in the database of Cochrane library that met the inclusion criteria had unpublished data. The characteristics of the included studies were summarized in Supplementary Table S1 .

These included studies came from different countries, including the Switzerland, Ireland, Netherlands, United Kingdom, China, and Israel. Follow-up times ranged from 6 months to over 60 months, and most studies had a follow-up time of about one year. The main outcomes of eligible studies included bone density, tiredness, hypothyroid symptoms, quality of life, body mass index (BMI), cognitive function, depression, blood pressure, renal function, lipid profile and cardiovascular events. The criteria for SCH varied slightly from study to study, and some studies did not limit the upper limit of TSH. The risk of bias for all included studies were detailed in Supplementary Table S2 . The overall risk of bias was “Low” for most studies (12–14, 16–18, 29). Two studies (19, 25) were retrospective case-control studies, and two other studies (24, 31) did not use randomization schemes, so their risk of bias was “High”. The remaining two articles (28, 30) did not clearly describe the experimental protocol, so the risk of bias was “Unclear”. Due to the small number of included studies for different outcomes, no funnel plot was drawn to describe publication bias.

Among the continuous variable results, there were not significant differences regarding the effects of levothyroxine on bone mineral density (p = 0.99; MD = 0.00; 95% CI, -0.03 to 0.03; I² = 0%) (12, 28) ( Figure 2A ), fatigue (p = 0.97; MD = -0.05; 95% CI, -2.73 to 2.63; I² = 0%) (13, 14) ( Figure 2B ), hypothyroidism symptoms (p = 0.77; MD = 0.35; 95% CI, -2.00 to 2.70; I² = 0%) (13, 14) ( Figure 2C ), quality of life (p = 0.95; MD = 0.06; 95% CI, -1.88 to 2.00; I² = 0%) (13, 14) ( Figure 2D ), BMI (p = 0.97; MD = -0.02; 95% CI, -0.85 to 0.82; I² = 82%) (13, 14, 19, 28, 30) ( Figure 3A ), cognitive function (p = 0.10; MD = 0.78; 95% CI, -0.14 to 1.71; I² = 81%) (17, 29) ( Figure 3B ), depression (p = 0.12; SMD = 0.35; 95% CI, -0.09 to 0.78; I² = 71%) (16, 17) ( Figure 3C ), serum creatinine (p = 0.08; MD = -13.75; 95% CI, -28.94 to 1.43; I² = 80%) (24, 30) ( Figure 3D ), systolic blood pressure (p = 0.73; MD = -0.27; 95% CI, -1.81 to 1.27; I² = 0%) (13, 14, 19, 30) ( Figure 4A ), diastolic blood pressure (p = 0.91; MD = -0.05; 95% CI, -0.90 to 0.80; I² = 0%) (13, 14, 19, 30) ( Figure 4B ), fasting blood glucose (p = 0.59; MD = 0.05; 95% CI, -0.14 to 0.24; I² = 0%) (30, 31) ( Figure 4C ), high-density lipoprotein cholesterol (HDL-C) (p = 0.46; MD = -0.04; 95% CI, -0.16 to 0.07; I² = 62%) (28, 30) ( Figure 4D ) and apolipoprotein A (ApoA) (p = 0.45; MD = -0.03; 95% CI, -0.11 to 0.05; I² = 0%) (28, 31) ( Figure 4E ). In the above analyses, both fatigue and hypothyroid symptoms were assessed using the Thyroid-Related Quality-of-Life Patient-Reported Outcome scale (ThyPRO), quality of life was assessed using the EuroQoL visual analogue scale (EQ VAS), and cognitive function was assessed using the Mini-Mental State Examination scale (MMSE). However, in 2 different studies, the Hospital Anxiety and Depression Scale (HADS) and 15-item Geriatric Depression Scale (GDS-15) were used to assess depression, respectively.

It was worth noting that, except for the above results, levothyroxine could have a significant effect on the lipid profile in older SCH patients, as shown below. Three articles (28, 30, 31) analyzed the effects of levothyroxine on cholesterol (TC) (p < 0.00001; MD = -0.92; 95% CI, -1.19 to -0.66; I² = 25%) ( Figure 5A ) and triglyceride (TG) (p < 0.00001; MD = -0.34; 95% CI, -0.49 to -0.19; I² = 0%) ( Figure 5B ) and showed significant differences in both. Levothyroxine also significantly reduced low-density lipoprotein cholesterol (LDL-C) (p = 0.03; MD = -0.54; 95% CI, -1.03 to -0.06; I² = 89%) ( Figure 5C ) and apolipoprotein B (ApoB) (p < 0.00001; MD = -0.24; 95% CI, -0.33 to -0.15; I² = 51%) ( Figure 5D ) in older SCH patients compared to controls, according to the analysis of 3 (18, 28, 30) and 2 (28, 31) articles, respectively.

Regarding adverse events that occurred during the study period, this meta-analysis did not find significant differences in the levothyroxine treatment group compared with the control group regarding fatal or nonfatal cardiovascular event (p = 0.15; OR = 1.13; 95% CI, 0.96 to 1.34; I² = 0%) (13, 14, 19, 25, 28) ( Figure 6A ), cardiovascular death (p = 0.22; OR = 0.80; 95% CI, 0.56 to 1.14; I² = 0%) (13, 14, 19) ( Figure 6B ), all-cause death (p = 0.07; OR = 0.84; 95% CI, 0.69 to 1.02; I² = 45%) (13, 14, 19) ( Figure 6C ), atrial fibrillation (p = 0.53; OR = 1.08; 95% CI, 0.85 to 1.36; I² = 0%) (13, 14, 19, 25) ( Figure 6D ), heart failure (p = 0.91; OR = 1.07; 95% CI, 0.31 to 3.66; I² = 77%) (13, 14, 25) ( Figure 6E ), and number of patients with ≥1 serious adverse event (including new atrial fibrillation, heart failure, fracture, and new diagnosis of osteoporosis) (p = 0.08; OR = 0.78; 95% CI, 0.58 to 1.03; I² = 32%) (13, 14) ( Figure 6F ) in older SCH patients.

To reduce the effect of different study protocols and different follow-up times on the results, we reanalyzed the results after removing retrospective case-control studies and studies with differences in follow-up times greater than six months. If these influences were not present, no re-analysis is required. Results of the reanalysis included BMI (p = 0.45; MD = -0.31; 95% CI, -1.10 to 0.48; I² = 66%) (13, 14, 28, 30) ( Supplementary Figure S1A ), systolic blood pressure (p = 0.63; MD = -0.53; 95% CI, -2.68 to 1.63; I² = 0%) (13, 14, 30) ( Supplementary Figure S1B ), diastolic blood pressure (p = 0.24; MD = -0.81; 95% CI, -2.15 to 0.54; I² = 0%) (13, 14, 30) ( Supplementary Figure S1C ), fatal or nonfatal cardiovascular event (p = 0.33; OR = 0.78; 95% CI, 0.47 to 1.29; I² = 0%) (13, 14, 28) ( Supplementary Figure S2A ), cardiovascular death (p = 0.92; OR = 1.09; 95% CI, 0.19 to 6.46; I² = 0%) (13, 14) ( Supplementary Figure S2B ), all-cause death (p = 0.15; OR = 1.85; 95% CI, 0.80 to 4.27; I² = 0%) (13, 14) ( Supplementary Figure S2C ), atrial fibrillation (p = 0.61; OR = 0.84; 95% CI, 0.42 to 1.67; I² = 0%) (13, 14) ( Supplementary Figure S2D ), and heart failure (p = 0.24; OR = 0.55; 95% CI, 0.20 to 1.48; I² = 0%) (13, 14) ( Supplementary Figure S2E ). These results did not change significantly from the original results.