Serum 25-hydroxyvitamin D and peripheral thyroid hormone sensitivity in euthyroid type 2 diabetes: Linear estimates and low-range threshold

Table 4 Threshold effect of 25-hydroxyvitamin D on the FT3/FT4 ratio from two-piecewise linear regression

FT3/FT4	Adjusted β (95%CI)	P value
25(OH)D (model I)	0.088 (0.036-0.140)	0.0010
25(OH)D (model II)
Inflection point	14.7 (12.85-16.11)
25(OH)D < 14.7	0.61 (0.31-0.92)	< 0.0001
25(OH)D ≥ 14.7	0.04 (-0.02 to 0.10)	0.2048
Difference (segment 2-1)	-0.57 (-0.91 to -0.24)	0.0007
Log likelihood ratio		< 0.001

Two-piecewise (segmented) linear regression with an estimated inflection at 14.7 ng/mL (95%CI: 12.85-16.11) of 25-hydroxyvitamin D [25(OH)D]. Slopes (β) represent the change in FT3/FT4 × 100 per 1 ng/mL higher 25(OH)D below and at/above the inflection, reported as β (95%CI) with two-sided P values. “Difference (segment 2-1)” is the contrast between the slope at/above vs below the inflection. “Model I” denotes the single-line linear model; “model II” the two-piecewise model. Model comparison vs the single-line specification used the likelihood-ratio test. The inflection point and its 95%CI were obtained by bootstrap resampling. Both model I and model II were fitted in the same complete-case sample (n = 577). All models were fully adjusted for age, sex, body mass index, systolic blood pressure, diastolic blood pressure, glycated hemoglobin (Box-Cox), total cholesterol (Box-Cox), high-density lipoprotein cholesterol (Box-Cox), low-density lipoprotein cholesterol, triglycerides (Box-Cox), estimated glomerular filtration rate (Box-Cox), uric acid (Box-Cox), aspartate aminotransferase (Box-Cox), alanine aminotransferase (Box-Cox), C-reactive protein (Box-Cox), white blood cell count (Box-Cox), visceral fat area (Box-Cox), subcutaneous fat area (Box-Cox), current smoking, and current drinking. No adjustment for multiple comparisons was applied. The outcome was modeled as FT3/FT4 × 100 for numerical scaling; inference is unaffected. 25(OH)D: 25-hydroxyvitamin D.