Thirteen studies were included with 247 participants. Three studies (23%) were RCTs with 81 participants.
Total cholesterol:
The mean decrease in the serum total cholesterol concentration (13 studies) was -0.20 mmol/L (-7.9 mg/dL; 95% CI: -0.09, -0.34) using the fixed-effect model, and -0.24 mmol/L (95% CI: -0.06, -0.42) using the random-effects model. The decline in serum total cholesterol was directly proportional to its baseline concentration. The test for heterogeneity suggested that these studies might be too dissimilar to combine.
Studies enrolling hypothyroid participants receiving suboptimal T4 doses (number of studies not stated) reported significantly larger decreases in serum total cholesterol after thyroid-stimulating hormone normalisation than those enrolling previously untreated individuals with mild thyroid failure: -0.44 mmol/L (-17 mg/dL; 95% CI: -0.18, -0.70) versus -0.14 mmol/L (-5.6 mg/dL; 95% CI: -0.01, - 0.28); P=0.05.
LDL cholesterol:
The change in the serum LDL cholesterol concentration (9 studies) using a fixed-effect model was -0.26 mmol/L (-10 mg/dL; 95% CI: - 0.12, -0.41). The results were more heterogeneous than would be expected by chance (P=0.02), but the reduction was similar using a random-effects model, -0.30 mmol/L (95% CI: -0.01, -0.54).
HDL cholesterol:
The change in the serum HDL cholesterol concentration (10 studies) was an increase after therapy by 0.08 mmol/L (3.2 mg/dL; 95% CI: 0.04, 0.13) using the fixed-effect model; although statistically significant, the results were driven by one study. These results were also more heterogeneous than would be expected by chance (P<0.001). A similar reduction, which was statistically non significant, was obtained using a random-effects model, 0.02 mmol/L (95% CI: -0.09, 0.12).
Triglycerides:
The mean change in serum triglyceride concentration (12 studies) was -0.01 mmol/L (95% CI: -0.08, 0.06), which was not statistically significant.
The 3 RCTs were compared with the controlled studies, but the two groups were not statistically different. Division of studies into two groups according to the average age of participants, participant pre-treatment TSH level, participant final TSH level, overall study score, gender representation in the study, and use of free T4 testing for diagnosis, still did not separate the studies into statistically different groups.
No study fulfilled all 13 criteria relating to internal and external validity. Three to 8 criteria, with a median of 6 criteria, were met. Only 3 studies were randomised, all with placebo controls; 3 studies did not adequately characterise the selection of participants, and 7 of the studies neither discussed losses to follow-up nor gave sufficient information from which this could be determined.