Discontinuation of levothyroxine therapy in patients with subclinical hypothyroidism: a pilot randomized clinical trial

Abstract

Purpose:

Randomized clinical trials (RCTs) have shown no benefit of levothyroxine for subclinical hypothyroidism (SCH) in improving well-being, cardiovascular outcomes, or mortality. We aimed to evaluate study procedures’ feasibility, safety, and preliminary effects of levothyroxine discontinuation in adults with SCH.

Methods:

We conducted a pilot, double-blind, placebo-controlled RCT with 6-month follow-up at a Veterans Affairs Medical Center. Adults with SCH on levothyroxine ≤ 75 mcg daily were randomized to continue levothyroxine or switch to placebo. The primary outcome was feasibility.

Results:

Fifty participants were randomized (32% enrollment rate); five were excluded post-randomization due to unconfirmed SCH, yielding 45 participants (21 levothyroxine, 24 placebo). One patient in the placebo group withdrew for personal reasons (98% completion rate). Participants’ mean age was 68.2 years (SD 9.7); 80% were male, and 86.7% were White. At 6 months, there was no statistically significant difference between the placebo and levothyroxine groups in ThyPRO-Hypothyroid Symptoms [28.3 (22.8) vs. 22.9 (19.5)], Tiredness [27.6 (22.8) vs. 32.8 (22.1)], and EQ-5D score [0.750 (0.232) vs. 0.741 (0.180)]. The only notable adverse event was rib fractures in a placebo group participant (TSH 3.04 mIU/L at 6 months). Two participants in the placebo group restarted levothyroxine (n=1, TSH >10 mIU/L; n=1, fatigue).

Conclusion:

We demonstrated feasibility of study procedures for discontinuing levothyroxine in patients with SCH and obtained preliminary effects on well-being. The low occurrence of adverse events suggests that levothyroxine discontinuation may be well-tolerated. These findings support conducting a larger multi-site RCT to comprehensively assess the effects of levothyroxine discontinuation.

Introduction

Subclinical hypothyroidism (SCH) is a biochemical diagnosis based on an elevated thyroid stimulating hormone (TSH) level with free thyroxine (FT4) level in the normal reference range [1]. SCH is present in approximately 12% of the population and has been often treated with levothyroxine (LT4), a synthetic form of T4 [2, 3]. LT4 is one of the most prescribed medications in the United States (U.S.), with approximately 18 million Americans using it daily [4], predominantly for treatment of SCH [5].

Nine out of 10 patients who start LT4 for SCH continue treatment indefinitely [6]. However, most patients diagnosed with SCH have a small TSH level elevation that often normalizes without any intervention [7, 8]. In addition, randomized clinical trials (RCTs) involving >2,000 patients with SCH have demonstrated that LT4 treatment does not significantly improve patients’ health-related quality of life (QoL), thyroid-related symptoms, depressive symptoms, fatigue, or cognitive dysfunction [9]. Moreover, pooled data from two RCTs showed no major difference in treatment satisfaction between older adults with SCH receiving LT4 or placebo [10]. Due to the lack of benefit coupled with treatment burden and potential harms, recent guidelines strongly recommend against the routine use of LT4 for treatment of SCH [11].

Given this recommendation, there is an urgent need to deprescribe LT4 in patients with SCH [12], particularly in older adults [13]. One significant obstacle to discontinuing unnecessary medications that provide no benefit is the concern among patients and clinicians of potential negative consequences on the patients’ medical condition, such as worsening of symptoms and QoL [14–21]. Currently, there are no RCTs that have assessed how LT4 discontinuation affects QoL, thyroid-related symptoms, or other patient-important outcomes-representing a key barrier to deprescribing. To address this knowledge gap, we aimed to evaluate the feasibility of conducting an RCT to discontinue LT4 treatment in adults with SCH, and to explore the changes in thyroid-related symptoms, QoL measures, thyroid hormone levels, lipid levels, clinical parameters (including blood pressure, weight, and body mass index [BMI]), and adverse events. We hypothesized that such an RCT will be feasible and that preliminary effect estimates would support the safety of LT4 discontinuation. Completion of this novel study will lead to the definitive evaluation and eventual implementation of an evidence-based LT4 discontinuation intervention for patients with SCH.

Materials and Methods

Study setting and participants

This study was conducted at the Central Arkansas Veterans Healthcare System (CAVHS) and the trial protocol was approved by the CAVHS Institutional Review Board. Written informed consent was obtained from the participants.

Participants were identified by reviewing electronic medical records of Veterans taking low dose LT4 (≤ 75 mcg) from the CAVHS pharmacy. This dosing range aligns with typical management of SCH in the U.S., where LT4 doses of 25 to 75 mcg/day are commonly used depending on the degree of TSH elevation and clinical judgment, in accordance with American Thyroid Association clinical practice guidelines [22]. The inclusion criteria were age ≥18 years, on treatment with LT4 and documented SCH prior to LT4 initiation, defined as elevated TSH with a normal FT4 level based on institutional reference ranges. We excluded patients who had medical conditions, clinical factors, or exposure to medications that could affect their thyroid function or increase their risk for overt hypothyroidism. The exclusion criteria were: TSH >10 mIU/L at any point; LT4 dose more than 75 mcg daily; use of antithyroid medications, liothyronine, desiccated thyroid extract, amiodarone, tyrosine kinase inhibitors or lithium; history of thyroidectomy or radioactive iodine therapy; thyroid cancer, goiter, pregnancy or plans for pregnancy in the next 6 months; an unstable medical condition that would jeopardize safety or interfere with study participation (e.g., active malignancy on chemotherapy); hospitalization for major illness in the last 4 weeks; severe hypothyroidism-related symptoms (Thyroid-specific Patient-Reported Outcome [ThyPRO] Hypothyroid Symptoms scale score 100 out of 100); strong family history of hypothyroidism (>3 first-degree relatives); severe dyslipidemia (LDL>190 mg/dL); acute coronary artery syndrome, acute myocarditis, pancarditis or stroke within the previous 12 months; grade IV New York Heart Association (NYHA) heart failure; receiving services from hospice; lack of decision-making capacity; terminal medical condition for which life expectancy would be less than 6 months; not willing to stop LT4; self-reported non-adherence to LT4 therapy; abnormal TSH at the time of screening for participation (elevated indicating a need to continue LT4, and suppressed suggesting overtreatment-both scenarios making randomization unsafe).

Trial design and procedures

Figure 1 describes the trial’s flow chart. At the baseline visit, patients consented to and completed blood tests (TSH, FT4, and lipids) and baseline patient-reported scales (see outcomes section). Demographics, including self-reported race and ethnicity, vitals, comorbidities and LT4 dose and duration were recorded. Eligible patients who were confirmed not to have any exclusion criteria were randomized in a 1:1 ratio, with stratification according to their pre-study daily LT4 dose (≤25 mcg, 26–50 mcg, 51–75 mcg) and the use of randomly permuted blocks, to either continue their usual (pre-study) dose of LT4 (sham discontinuation- control) or placebo (real discontinuation- intervention). The randomization schedule was created by the study biostatistician, independent of the remaining study team. Participants, their clinicians, and the research team, except for the research pharmacist and biostatistician, were blind to treatment allocation. Study pill preparation involved over-encapsulation with the placebo capsule containing filler (microcrystalline cellulose) to further minimize the risk of unblinding.

Figure 1.

Trial’s flow chart.

The first follow-up visit was scheduled 6–8 weeks after the baseline visit/randomization, which is the usual timeframe in clinical practice to evaluate LT4 treatment changes [22]. Vital signs were recorded, participants completed questionnaires regarding QoL, and blood was drawn to measure TSH, FT4, and thyroid peroxidase antibody (TPOAb) levels. Study investigators were blinded to the laboratory results. The final follow-up visit was scheduled 6 months after the baseline visit/randomization, which is adequate time to detect any significant changes in thyroid function, symptoms, and QoL [23, 24]. Vital signs were recorded, participants completed questionnaires regarding QoL, and blood was drawn to measure TSH, FT4, and lipid levels. Reverse or free T3 were not included, as they do not correlate with patient outcomes and are not recommended in clinical practice for assessing hypothyroidism [22, 25].

Development of overt hypothyroidism, TSH > 10 mIU/mL, subclinical hyperthyroidism, or overt hyperthyroidism was an indication to discontinue the study medication and return to care under treating clinicians. After discontinuing the study medication, participants continued to be followed per protocol unless they chose to withdraw from the study.

The self-reported adherence to treatment (LT4/placebo) was assessed monthly through phone encounters and at participants’ scheduled follow-up visits. In addition, at each study visit, all participants were asked to bring their medication bottles and their pill diary, and a pill count was performed to assess adherence. Adherence was classified as good if a participant was taking the study medication between 80% and 100% of the time based on pill counts [26]. At the final follow-up visit, all participants were asked to describe any difficulties they faced in understanding and participating in the study.

Outcomes

The primary outcome of this pilot trial was feasibility of study procedures. Secondary outcomes assessed at the baseline, 6–8 weeks, and 6-month visits included changes in: thyroid function tests (TSH and FT4); thyroid-related symptoms measured with the Thyroid-specific Patient-Reported Outcome (ThyPRO) Hypothyroid Symptoms scale score (4 items) and Tiredness scale score (7 items). Each ThyPRO scale score ranges from 0 to 100, with higher scores indicating more symptoms and tiredness, respectively [27]. ThyPRO is a validated questionnaire that is responsive to changes in thyroid status and has been found in a systematic review to be the best disease-specific QoL measure for use in clinical trials of patients with hypothyroidism [28]; general QoL measured with the EuroQoL (EQ) Group 5-Dimension Self-Report Questionnaire (EQ-5D). Scores on the EQ-5D descriptive index range from −0.59 to 1.00, and scores on the EQ visual analogue scale (EQ VAS) range from 0 to 100, with higher scores indicating better health state [29]; comprehensive thyroid-related QoL measured with the ThyPRO-39 score (includes the 11 items of the Hypothyroid Symptoms/Tiredness scales and 28 additional items), a shorter version of the ThyPRO measure [30], (assessed at baseline and 6-month visit only); blood pressure (systolic and diastolic); pulse; weight; and BMI.

Safety and adverse events

Adverse events were assessed, managed, recorded, reported, and analyzed in accordance with the U.S. Food and Drug Administration (FDA) regulations (21 CFR Parts 312 and 812) and the International Council for Harmonisation (ICH) E6(R2) Good Clinical Practice guidelines. An independent safety monitor oversaw the study’s conduct and monitored for significant adverse events. Adverse events of special interest included incidence of overt hypothyroidism and hyperthyroidism (confirmed by laboratory thyroid function tests), death, cardiovascular events (atrial fibrillation confirmed by electrocardiography, acute myocardial infarction, stroke, acute coronary syndrome, and heart failure identified through medical records and clinical diagnosis), and fractures (per patient report and ascertained via medical records and imaging reports when possible).

Statistical analysis

For the primary outcome, feasibility of study procedures, we calculated participants’ enrollment rate (percent of eligible patients approached who consented to participate), recruitment rate (number of patients randomized divided by the length of the recruitment period, i.e., from the date that recruitment opened to the date of the last randomization), and study completion rate.

Continuous variables were summarized as mean (standard deviation; SD) and categorical variables as % (n). Group comparisons were conducted using Welch’s two-sample t-test for continuous variables and Fisher’s exact test for categorical variables. Analysis of covariance (ANCOVA) adjusting for gender and baseline value of the variable was performed for the secondary outcomes. In addition, generalized estimating linear mixed effects models were employed to evaluate longitudinal changes from baseline to 6–8 weeks and to 6 months in the thyroid function tests, ThyPRO Hypothyroid Symptoms and Tiredness scale scores, and EQ-5D scores. We also explored differences in outcomes by duration of LT4 treatment before randomization (<2 years vs. ≥2 years). Analyses comparing groups with respect to adverse event rates were performed using Fisher’s exact tests.

For the ThyPRO Hypothyroid Symptoms and Tiredness scale scores, we conducted a noninferiority analysis to assess whether stopping LT4 leads to significantly worse symptoms compared to continuing treatment. We used a one-sided Welch’s two-sample t-test to determine whether the average symptom scores at 6 months in the placebo (real discontinuation) group were at most 14 points higher than in the LT4 group, which we set as the noninferiority limit based on our sample size. A 9-point difference in ThyPRO scales has been identified in prior research as the minimal clinically important difference —the smallest change that patients would likely notice [31]. However, due to sample size constraints, we were only able to evaluate noninferiority using a 14-point threshold. We repeated the noninferiority analysis for the ThyPRO Hypothyroid Symptoms and Tiredness scale scores while adjusting scores for the baseline values by taking the difference between 6-month and baseline scores (exploratory noninferiority analysis). We used a statistical significance level (α) of 0.05 for each ThyPRO scale. If the 2-sided 90% confidence interval (CI) of the difference included the noninferiority limit, we could not rule out the possibility that symptoms in the discontinuation group were meaningfully worse.

All analyses were based on the intention-to-treat principle. Analyses were repeated in the per-protocol population, which included participants who continued to take the LT4 treatment per the trial protocol. Statistical analyses were conducted using SAS 9.4 (SAS Institute, Cary, NC).

Results

Trial population and feasibility

Between March 2021 and April 2022, we contacted 177 patients who appeared eligible based on electronic medical record review to confirm their eligibility and assess their interest in participating in the trial. Of these, 10 (6%) were ineligible, 59 (33%) were willing to participate, and 108 (61%) declined. The primary reasons for declining were distance or time constraints (49/108, 45%), concerns about LT4 discontinuation (30/108, 28%), health concerns (14/108, 13%), general opposition to research (9/108, 8%), and needing more time to consider participation (6/108, 6%).

Among the 59 patients who were willing to participate, five were later deemed ineligible, and two could not be enrolled due to enrollment closure upon reaching the target number of participants. Ultimately, we enrolled 52 participants, achieving an enrollment rate of 32% (52/162 eligible patients) and a recruitment rate of four patients per month. Two participants were excluded due to abnormal TSH levels at screening, and 50 were randomized: 24 to the LT4 group and 26 to the placebo group. Post-randomization, five participants were excluded due to unconfirmed SCH diagnosis, resulting in a final sample size of 45 participants (21 in the LT4 group and 24 in the placebo group). Participants were on a stable LT4 dose for at least 3 months prior to randomization.

One patient in the placebo group withdrew 2 weeks after enrollment for personal reasons (98% completion rate); repeat TSH and FT4 levels were normal at the time of the participant’s withdrawal. The remaining 44 participants completed 100% of the study procedures. The numbers of participants who were included in the analyses are presented in Figure 2. Participants self-reported 100% adherence to the study medication. Pill counts at study visits confirmed adherence rates ranging from 91% to 100%, with 88% of participants maintaining adherence above 98%. None of the participants reported difficulties in understanding and participating in the study.

Figure 2.

Randomization and follow-up.

The characteristics at baseline were similar between the LT4 and placebo groups (Table 1) except for low-density lipoprotein cholesterol [77.5 (SD 34.5) vs. 104.2 (SD 28.4) mg/dL, p=0.0078], high-density lipoprotein cholesterol [42.4 (SD 11.5) vs. 53.3 (SD 19.8) mg/dL, 0.0276], and total cholesterol [148.1 (SD 41.9) vs. 179.8 (SD 31.3) mg/dL, p=0.0074] levels, respectively. The mean age of the participants was 68.2 years (SD 9.7), 36 participants (80%) were male, and 39 participants were White (87%). Most participants were treated with LT4 for ≥ 2 years (73.3%). Among those treated for less than 2 years, the duration ranged from 9 to 23.3 months. Eight participants were TPOAb positive (18%). At baseline, the participants’ mean Hypothyroid Symptoms score was 24.3 (SD 19.5) and the mean Tiredness score was 32.2 (SD 23.5).

Table 1.

Demographic and baseline medical characteristics.

Variable		Levothyroxine (N=21)	Placebo (N=24)	P-value
Age (yrs)		71.0 (7.9)	65.7 (10.6)	0.0646
Male		85.7% (18)	75.0% (18)	0.4689
Race				>0.99
Caucasian		85.7% (18)	87.5% (21)
African American		9.5% (2)	8.3% (2)
American Indian		0.0% (0)	4.2% (1)
≥ 2 races		4.8% (1)	0.0% (0)
Hispanic ethnicity		0.0% (0)	4.0% (1)
Weight (lbs)		220.4 (58.4)	201.0 (61.5)	0.2841
Body mass index (kg/m2)		32.5 (7.7)	29.7 (7.8)	0.2405
Blood pressure (mm Hg)
Systolic		141.1 (19.7)	137.5 (16.8)	0.5176
Diastolic		80.1 (8.7)	80.2 (11.7)	0.9814
Pulse (beats/min)		68.3 (15.7)	66.8 (9.4)	0.6891
Comorbidities
IHD		42.9% (9)	8.3% (2)	0.0132
AFIB		19.0% (4)	8.3% (2)	0.3955
HTN		90.5% (19)	50.0% (12)	0.0042
DM		42.9% (9)	12.5% (3)	0.0406
LT4 dose prior to enrollment				0.9247
25 μg		33.3% (7)	29.2% (7)
50 μg		47.6% (10)	54.2% (13)
75 μg		19.0% (4)	16.7% (4)
LT4 treatment duration				>0.99
<2 yrs		28.6% (6)	25.0% (6)
2 to 7 yrs		47.6% (10)	50.0% (12)
>7 yrs		23.8% (5)	25.0% (6)
TSH (mIU/L)		3.19 (1.42)	3.08 (0.99)	0.7721
FT4 (ng/dL)		0.9 (0.1)	0.9 (0.1)	0.0841
TPOAb positivity		19.0% (4)	17.4% (4)	>0.99
Lipids (mg/dL)
LDLc		77.5 (34.5)	104.2 (28.4)	0.0078
HDLc		42.4 (11.5)	53.3 (19.8)	0.0276
Total cholesterol		148.1 (41.9)	179.8 (31.3)	0.0074
Triglycerides		141.5 (68.0)	111.4 (48.0)	0.0997
Hypothyroid Symptoms Score		21.1 (18.4)	27.1 (20.4)	0.3091
Tiredness score		31.1 (23.9)	33.2 (23.7)	0.7733
ThyPRO-39 score		20.5 (14.8)	18.4 (11.2)	0.6092
EQ-5D descriptive index		0.758 (0.193)	0.746 (0.229)	0.8429
EQ-5D VAS score		72.0 (15.0)	77.3 (14.4)	0.2348

Continuous variables summarized as mean (standard deviation). Welch’s two-sample t-tests were used to compare groups.

Categorical variables summarized as % (n) and groups were compared using Fisher’s exact tests.

IHD; ischemic heart disease, AFIB; atrial fibrillation, HTN; hypertension, DM; diabetes mellitus, LT4; levothyroxine, TSH; thyroid stimulating hormone, FT4; free thyroxine, TPOAb; thyroid peroxidase antibody, LDLc; low-density lipoprotein cholesterol, HDLc; high-density lipoprotein cholesterol, ThyPRO; Thyroid-specific Patient-Reported Outcome, EQ-5D; EuroQoL Group 5-Dimension Self-Report Questionnaire, VAS; visual analogue scale

Thyroid function tests

At baseline, the participants’ mean TSH level was 3.1 mIU/L (SD 1.2) and FT4 level 0.9 ng/dL (SD 0.1). The TSH levels increased from baseline to a greater extent in the placebo group than in the LT4 group at both time points of analysis. At 6–8 weeks after randomization, the mean TSH was 3.07 mIU/L (SD 1.39) in the LT4 group as compared with 5.06 mIU/L (SD 2.04) in the placebo group, with a mean between-group difference of −2.10 mIU/L (95% CI −3.19 to −1.01) (P=0.0004) (Table 2). At 6 months after randomization, the mean TSH was 3.49 mIU/L (SD 1.45) in the LT4 group as compared with 5.15 mIU/L (SD 2.20) in the placebo group, resulting in a between-group difference of −1.74 mIU/L (95% CI −2.83 to −0.66) (P=0.0024) (Table 3). There was no significant longitudinal change in TSH levels from the visit at 6–8 weeks to the visit at 6 months after randomization for either the LT4 group (P=0.06) or placebo group (P=0.85).

Table 2.

Comparison of treatment groups with respect to outcome measures 6–8 weeks after randomization. The reported levothyroxine – placebo difference and the 95% confidence interval (CI) are from an analysis of covariance (ANCOVA) model that adjusts group differences for gender and the baseline measure of the variable. The P-value testing for group differences is also from the ANCOVA analysis.

Variable	Levothyroxine (N=21)	Placebo (N=23)	Difference 95% CI	P-value
Weight (lbs)	220.6 (59.2)	198.8 (60.4)	0.2 −2.4 to 2.8	0.8850
Body mass index (kg/m2)	32.5 (7.8)	29.3 (7.5)	0.0 −0.3 to 0.4	0.8360
Systolic blood pressure (mm Hg)	140.3 (21.1)	137.1 (15.9)	0.5 −8.4 to 9.4	0.9155
Diastolic blood pressure (mm Hg)	79.2 (12.4)	80.0 (11.5)	−0.4 −6.6 to 5.8	0.8956
Pulse (beats/min)	65.6 (9.0)	68.3 (9.6)	−3.1 −8.1 to 1.8	0.2091
TSH (mIU/L)	3.07 (1.39)	5.06 (2.04)	−2.10 −3.19 to −1.01	0.0004
FT4 (ng/dL)	0.9 (0.2)	0.8 (0.2)	0.1 0.0 to 0.2	0.0223
SCH Status@	9.5% (2)	26.1% (6)	--	0.2451
Hypothyroid Symptoms score	22.0 (17.7)	25.3 (21.9)	0.8 −9.4 to 10.9	0.8793
Tiredness score	33.0 (23.6)	31.1 (24.3)	2.3 −10.1 to 14.7	0.7104
EQ-5D descriptive index	0.759 (0.190)	0.786 (0.214)	−0.020 −0.076 to 0.036	0.4684
EQ VAS score	78.0 (13.8)	77.2 (19.6)	7.0 −0.4 to 14.4	0.0643

^@Fisher’s exact test was used to compare groups with respect to SCH Status.

TSH; thyroid stimulating hormone, FT4; free thyroxine, SCH; subclinical hypothyroidism, EQ-5D; EuroQoL Group 5-Dimension Self-Report Questionnaire, VAS; visual analogue scale

Table 3.

Comparison of treatment groups with respect to outcome measures 6 months after randomization. The reported levothyroxine – placebo difference and the 95% confidence interval (CI) are from an analysis of covariance (ANCOVA) model that adjusts group differences for gender and the baseline measure of the variable. The P-value testing for group differences is also from the ANCOVA analysis. (Intent-to-Treat)

Variable	Levothyroxine (N=21)	Placebo (N=23)	Difference 95% CI	P-value
Weight (lbs)	218.2 (57.1)	196.9 (61.6)	−0.2 −5.4 to 5.0	0.9399
Body mass index (kg/m2)	32.1 (7.0)	29.0 (7.6)	0.1 −0.7 to 0.8	0.8869
Systolic blood pressure (mm Hg)	144.9 (25.1)	140.2 (23.6)	0.1 −12.0 to 12.3	0.9821
Diastolic Blood Pressure (mm Hg)	81.2 (9.4)	80.7 (12.0)	−0.0 −6.0 to 5.9	0.9909
Pulse (beats/min)	68.5 (15.1)	67.2 (10.2)	0.6 −4.9 to 6.1	0.8291
LDLc (mg/dL)	76.5 (28.6)	100.2 (32.3)	−6.8 −22.2 to 8.6	0.3805
HDLc (mg/dL)	42.7 (16.5)	51.5 (16.2)	1.7 −4.44 to 7.9	0.5724
Total cholesterol (mg/dl)	145.3 (32.6)	177.4 (39.6)	−12.0 −29.3 to 5.3	0.1695
Triglycerides (mg/dL)	137.5 (66.8)	139.7 (68.4)	−17.8 −59.4 to 23.9	0.3939
TSH (mIU/L)	3.49 (1.45)	5.15 (2.20)	−1.74 −2.83 to −0.66	0.0024
FT4 (ng/dL)	0.9 (0.2)	0.8 (0.1)	0.2 0.2 to 0.3	0.0008
SCH status@	9.5% (2)	34.8% (8)	--	0.0725
Hypothyroid Symptoms score	22.9 (19.5)	28.3 (22.8)	−2.3 −14.0 to 9.3	0.6860
Tiredness score	32.8 (22.1)	27.6 (22.8)	5.4 −5.9 to 16.7	0.3381
ThyPRO-39 score	16.6 (12.0)	16.7 (15.0)	−2.2 −8.2 to 3.8	0.4567
EQ-5D descriptive index	0.742 (0.180)	0.750 (0.232)	0.003 −0.102 to 0.107	0.9585
EQ VAS score	68.4 (21.2)	79.9 (12.9)	−6.2 −15.3 to 2.9	0.1742

^@Fisher’s exact test was used to compare groups with respect to SCH Status.

TSH; thyroid stimulating hormone, FT4; free thyroxine, LDLc; low-density lipoprotein cholesterol, HDLc; high-density lipoprotein cholesterol, SCH; subclinical hypothyroidism, ThyPRO; Thyroid-specific Patient-Reported Outcome, EQ-5D; EuroQoL Group 5-Dimension Self-Report Questionnaire, VAS; visual analogue scale

The mean FT4 level was 0.1 ng/dL (95% CI 0.0 to 0.2) and 0.2 ng/dL (95% CI 0.2 to 0.3) higher in the LT4 group than in the placebo group at 6–8 weeks (P=0.0223) and at 6 months (P=0.0008) after randomization, respectively (Tables 2 and 3). There was no significant longitudinal change in FT4 from the visit at 6–8 weeks to the visit at 6 months after randomization for either the LT4 group (P=0.70) or placebo group (P=0.09).

At 6–8 weeks after randomization, 19/21 participants (90.5%) in the LT4 group and 17/23 participants (73.9%) in the placebo group were euthyroid (P=0.25). By the end of the follow-up at 6 months after randomization, 19/21 participants (90.5%) in the LT4 group and 15/23 participants (65.2%) in the placebo group were euthyroid (P=0.07).

Thyroid-specific quality of life and other outcome measures

At 6–8 weeks and 6 months after randomization, no significant differences were found between groups in weight, BMI, blood pressure, heart rate, lipid profile (assessed only at 6 months) and thyroid-specific and general health QoL measures after adjusting group differences for gender and the baseline measure of the variable (Table 2 and Table 3).

Noninferiority analysis for the ThyPRO Hypothyroid Symptoms score at 6 months after randomization showed that the difference of the mean scores between the LT4 and placebo (LT4 discontinuation) groups could exceed the −14 noninferiority limit (p=0.0913). Noninferiority analysis for the ThyPRO Tiredness score at 6 months after randomization showed that the difference of the mean scores between the LT4 and placebo (LT4 discontinuation) groups did not exceed the −14 noninferiority limit (p=0.0036) (Table 4). In an exploratory analysis where ThyPRO Hypothyroid Symptoms and Tiredness scores were adjusted for baseline values, both the Hypothyroid Symptoms (p=0.0108) and Tiredness scores (p=0.0016), on average, did not exceed the −14 noninferiority limit (Supplementary Table 1).

Table 4.

Noninferiority tests comparing the placebo group to the levothyroxine group with respect to 6-month Hypothyroid Symptoms and Tiredness scores. Differences in means are calculated as the levothyroxine mean minus the placebo mean. The difference in means is compared to a noninferiority limit of Δ = −14 using a one-sided Welch’s two-sample t-test. (Intent-to-Treat: Participants analyzed in the group to which they were randomized, regardless of whether protocol deviations occurred.)

Variable	Levothyroxine (N=21)	Placebo (N=23)	Difference 90% Conf. Interval	P-value@
Hypothyroid Symptoms score	22.9 (19.5)	28.3 (22.8)	−5.3 −16.1 to 5.4	0.0913
Tiredness score	32.8 (22.1)	27.6 (22.8)	5.2 −6.2 to 16.6	0.0036

^@A one-sided Welch’s two-sample t-test was used to test the following hypotheses:

H₀: μ_Lev - μ_Plac ≤ Δ = −14 versus H₁: μ_Lev - μ_Plac > −14.

A P-value less than 0.05 indicates the placebo mean fell within the noninferiority limit.

We explored the effect of pre-study LT4 therapy duration on outcomes (Table 5). The sample size of the group with duration of LT4 therapy less than 2 years was small (N=12) which did not allow further stratification by treatment status.

Table 5.

Assessing the effect of the duration LT4 therapy on outcomes. No adjustment for multiple comparisons were made to the reported P-values.

Variable	Duration of LT4 Therapy		P-value
	<2 yrs. (N=12)	≥2 yrs. (N=32)
Weight (lbs)	210.0 (63.6)	206.0 (59.3)	0.8463
Body mass index (kg/m2)	30.5 (8.4)	30.5 (7.1)	0.9941
Systolic blood pressure (mm Hg)	141.1 (29.8)	142.9 (22.2)	0.8237
Diastolic blood pressure (mm Hg)	81.9 (12.8)	80.6 (10.0)	0.7267
Pulse (beats/min)	74.8 (18.4)	65.2 (8.7)	0.0225
LDLc (mg/dL)	96.2 (27.4)	86.2 (34.3)	0.3712
HDLc (mg/dL)	45.3 (15.5)	46.7 (18.4)	0.8221
Total cholesterol (mg/dl)	169.9 (35.4)	158.6 (40.9)	0.4039
Triglycerides (mg/dL)	142.0 (40.4)	133.3 (77.3)	0.7130
TSH (mIU/L)	4.34 (1.79)	4.37 (2.15)	0.9687
FT4 (ng/dL)	0.8 (0.2)	0.9 (0.2)	0.4616
Hypothyroid Symptoms score	22.4 (15.2)	27.0 (23.2)	0.5324
Tiredness score	29.2 (23.9)	30.5 (22.2)	0.8661
ThyPRO-39 score	16.3 (11.0)	16.8 (14.5)	0.9129
EQ-5D descriptive index	0.778 (0.225)	0.734 (0.202)	0.5352
EQ VAS score	74.7 (17.0)	74.3 (18.8)	0.9548

LT4; levothyroxine, TSH; thyroid stimulating hormone, FT4; free thyroxine, LDLc; low-density lipoprotein cholesterol, HDLc; high-density lipoprotein cholesterol, ThyPRO; Thyroid-specific Patient-Reported Outcome, EQ-5D; EuroQoL Group 5-Dimension Self-Report Questionnaire, VAS; visual analogue scale

Per-protocol analyses showed similar results (data not shown).

Serious adverse events

One patient in the placebo group (LT4 discontinuation) reported rib fractures after falling. Computed tomography of the chest 8 days later showed multiple chronic rib fractures, but no acute fracture. None of the study participants developed overt hypothyroidism, hyperthyroidism, cardiovascular events (atrial fibrillation, acute myocardial infarction, stroke, acute coronary syndrome, heart failure) or died.

Two participants in the placebo group (LT4 discontinuation) restarted LT4. The first patient had a TSH level of 7.22 mIU/L with a normal FT4 level at 6–8 weeks after the randomization visit. The patient’s ThyPRO Hypothyroid Symptoms score increased from 18.8 to 37.5, and the Tiredness score increased from 3.6 to 28.6, indicating more hypothyroid symptoms and tiredness, but the EQ-5D descriptive index improved from 0.827 to 0.86. Two weeks later, due to ongoing concern of tiredness, the patient was restarted on LT4 (patient preference). Three weeks later the patient was diagnosed with metastatic rectal adenocarcinoma. The second patient had a TSH level of 11.40 mIU/L with a normal FT4 level at 6–8 weeks after the randomization visit. Although the patient did not endorse any significant change in their health state, LT4 was started per study protocol due to TSH>10 mIU/L.

Discussion

In this pilot, double-blind, randomized, placebo-controlled, parallel-group trial involving adults with SCH, LT4 discontinuation and implementation of study procedures was feasible. Preliminary findings suggest that LT4 discontinuation did not increase adverse events or result in deterioration of overall QoL, weight, BMI, or lipid levels. LT4 discontinuation did not meet the noninferiority margin for hypothyroid symptoms, meaning we cannot rule out the possibility of a clinically meaningful worsening of symptoms. However, for tiredness, LT4 discontinuation met the noninferiority margin, suggesting no significant difference compared to continued LT4 use. An exploratory analysis adjusting for baseline symptoms indicated that initial symptom severity may have influenced these findings.

For the primary outcome, we have successfully confirmed the feasibility of the study by meeting enrollment and recruitment goals despite conducting the RCT during the COVID-19 pandemic with enrollment and recruitment rates better than that described in the literature for clinical trials at VA sites [32–34]. The main reason for declining participation was time or distance constraints, which needs to be considered in the design of future studies. The completion rate of the study procedures was 98% and none of the participants reported difficulties in understanding and participating in the study. For the secondary outcomes, although the study did not have sufficient statistical power to identify any significant differences between the participants who continued LT4 and those randomized to placebo (LT4 discontinuation) in terms of tiredness, overall QoL measures, weight, BMI, blood pressure, heart rate, lipid profile, and incidence of adverse events, there was no signal of harm. Although noninferiority analysis showed that we cannot rule out the possibility of a clinically meaningful worsening of hypothyroid symptoms after LT4 discontinuation, an exploratory analysis adjusting for baseline symptoms indicated that initial symptom severity may have influenced these findings. A high symptom burden has been defined as ThyPRO Hypothyroid Symptoms score >30 and Tiredness score >40 [35]. Notably our study participants had a mean ThyPRO Hypothyroid Symptoms score of 24.3 (SD 19.5) and Tiredness score of 32.2 (SD 23.5), showcasing a diverse range of symptom burden.

After six months of follow-up, we found that 65% of participants who had discontinued LT4 (placebo) had normal thyroid function tests, while 35% exhibited SCH, and none developed overt hypothyroidism. These findings indicate that these patients maintained a euthyroid state regardless of LT4 use, suggesting overtreatment and unnecessary exposure to the treatment burden and potential harm. Notably, most participants had been on LT4 therapy for at least two years. These data are consistent with several studies showing that many adults with biochemical results that are consistent with SCH will revert to a euthyroid state if they are followed-up without initiating LT4 treatment [31, 36, 37]. Similarly to our findings, a systematic review and meta-analysis including 1103 patients from 17 observational studies found that 36% (95% CI: 8–63%) of patients with SCH remain euthyroid after LT4 discontinuation [12].

LT4 treatment is associated with potential patient burden, harm, and healthcare system costs. LT4 treatment requires modification of daily habits, e.g., dosing 30–60 min before a meal, monitoring of effects, frequent dose titration, clinic and laboratory visits, and financial costs to patients and society [1]. Indeed, increased utilization of LT4 has increased U.S. healthcare expenditures associated with LT4 from $1.1 billion in 1997 to $3.2 billion in 2016, and an average cost of $1600/year for each patient (including the costs of medical visits and laboratory tests), and up to $4,100/year for patients requiring 3 or more LT4 dose adjustments [38, 39]. Two-thirds of pregnant women with hypothyroidism report significant LT4-related treatment burdens such as financial hardship and difficulty in taking a pill [40]. Adding to this patient burden, half of adults ≥65 years who take LT4 are subjected to overtreatment during their follow-up, resulting in symptoms of hyperthyroidism and an increased risk of experiencing arrhythmia, cardiovascular events, osteoporosis, fractures, cognitive impairment, and dementia [41].

Due to treatment burden, harms, and lack of benefit, a recent multidisciplinary panel employing a novel metric to identify low-value prescribing practices assigned the highest low-value prescription score to the initiation of LT4 therapy in patients without overt hypothyroidism [13]. However, despite no proven health benefits and clinical practice guidelines against its routine use, a substantial portion of patients with SCH is still receiving LT4. LT4 deprescription in patients with SCH, particularly in older patients, has emerged as a vital intervention to mitigate negative effects related to burdens, harms, and societal costs [12]. Yet, an important barrier to LT4 deprescription is the uncertainty shared by patients, caregivers, and clinicians about potential negative impacts on the patient’s health [21]. In fact, this concern of negative consequences has been identified as an important barrier to the deprescription of multiple long-term medications in primary care for older adults, in whom most LT4 deprescribing is likely to occur [14–20]. A recent study reported barriers to thyroid hormone deprescribing in older adults at multiple levels including patient factors (e.g., concerns about potential side effects related to thyroid hormone dose reduction), physician and system factors (e.g., clinic visit time constraints, physician inertia and lack of knowledge about deprescribing) [21]. Addressing these barriers requires a better understanding of the impact of LT4 deprescription on patient well-being, which is currently limited to observational studies with high risk of bias that do not routinely measure thyroid-related symptoms and QoL [12]. Therefore, there is an urgent need for scientifically rigorous clinical studies assessing if LT4 can be discontinued safely (without affecting well-being) among patients who have SCH in order to support the adoption of LT4 discontinuation. Likewise, the impact of LT4 discontinuation on downstream LT4-associated costs remains unknown. Understanding these costs is crucial to provide insights into the economic implications of LT4 discontinuation that could support discontinuation strategies. This pilot RCT provided preliminary data on effect sizes and demonstrated study feasibility that can inform the design of a subsequent, full-scale effectiveness clinical trial with adequate power to assess changes on patient-important outcomes after LT4 discontinuation in patients with SCH.

Our trial has certain limitations. First, since participants were excluded if TSH level had exceeded 10 mIU/L or they were receiving LT4>75 mcg/day, we cannot generalize our results to this subgroup of patients with SCH. Second, 80% of participants were male, reflecting the demographics of the Veterans Health Administration where approximately 93% of patients are male, compared to about 45% in the general U.S. population [42]. In a retrospective cohort study conducted at CAVHS-the site of our trial- we identified 229 Veterans newly diagnosed with SCH between 2016 and 2018, 90% of whom were male, consistent with our trial population [43]. Since hypothyroidism is more prevalent in women, this gender imbalance may limit the generalizability of our findings to broader populations, particularly to female patients with SCH. Moreover, a small proportion of participants had TPOAb positivity, likely due to male predominance in our study. TPOAb-positive patients are more likely than TPOAb-negative patients to have progressive hypothyroidism and therefore an indication for LT4 treatment [1]. In addition, since the most commonly reported reasons for declining participation were time constraints and transportation issues (e.g., driving distance), it is possible that socioeconomic or logistical factors influenced patients’ decisions to participate, potentially introducing selection bias into our sample. We were unable to determine whether participants experienced hypothyroid-related symptoms or impaired QoL prior to initiating LT4 therapy, as this information was not reliably documented in the electronic medical record. This limits our ability to determine whether discontinuing LT4 had any impact on symptoms or QoL that may have been abnormal or impaired before treatment was initiated. Additionally, individuals who perceived subjective benefit from LT4 may have been less likely to enroll in the study, potentially introducing selection bias. Finally, our trial was underpowered to detect any statistically significant effect on QoL, lipid levels, and incidence of adverse events. Therefore, we cannot exclude the possibility that LT4 discontinuation may be beneficial or harmful. However, our trial has several strengths. We used validated measures of thyroid-specific QoL that have been shown to be sensitive to change [27, 30], as well as a range of secondary outcomes of clinical relevance and patient importance.

In conclusion, this pilot trial indicated the feasibility of study procedures for discontinuing LT4 in patients with SCH and provided preliminary effects on well-being. The low occurrence of adverse events suggests that LT4 discontinuation may be well-tolerated in this setting. These findings support proceeding with a larger multi-site RCT to comprehensively assess the effects of LT4 discontinuation.