eFigure. Study Population Selection.
eTable 1. Clinically Relevant Medications That May Affect Thyroid Function.
eTable 2. TSH Levels Within 6 Weeks and 12 Months After Index Date Among Nonswitchers vs Switchers (Adjusted Regression Models).
eTable 3. Sensitivity Analysis Using TSH Levels Between 6 and 12 Weeks After Index Date Among Nonswitchers vs Switchers (Adjusted Regression Models).
eTable 4. Subgroup Analyses: Patients on Varying Doses of Levothyroxine (Adjusted Regression Models).
eTable 5. TSH Levels Within 6 Weeks and 12 Months After Index Date Among Nonswitchers vs Switchers (Using Robust Estimator of the Variance).
eTable 6. Sensitivity Analysis Using TSH Levels Between 6 and 12 Weeks After Index Date Among Nonswitchers vs Switchers (Using Robust Estimator of the Variance).
eTable 7. Subgroup Analyses: Patients on Varying Doses of Levothyroxine (Using Robust Estimator of the Variance).
Customize your JAMA Network experience by selecting one or more topics from the list below.
Brito JP, Deng Y, Ross JS, et al. Association Between Generic-to-Generic Levothyroxine Switching and Thyrotropin Levels Among US Adults. JAMA Intern Med. 2022;182(4):418–425. doi:10.1001/jamainternmed.2022.0045
Is switching among generic levothyroxine sodium products made by different manufacturers associated with changes in serum thyrotropin (TSH) levels?
In this large comparative effectiveness research study of 15 829 patients who filled a generic levothyroxine prescription, the proportion of patients with normal TSH levels was 83% among patients who continued to use the same product vs 85% among patients who switched between generic products, a nonsignificant difference.
Results of this study conflict with the current expert guideline recommendation that universally warns patients and clinicians about the possible changes in TSH levels associated with switching among levothyroxine products.
Switching among generic levothyroxine sodium products made by different manufacturers typically occurs at the pharmacy and may affect serum thyrotropin (TSH) levels.
To compare TSH levels between patients who continued taking the same sourced generic levothyroxine product and those who switched.
Design, Setting, and Participants
This comparative effectiveness research study with 1:1 propensity matching used data from OptumLabs Data Warehouse, a national administrative claims database linked to laboratory test results. Adults aged 18 years or older were included if they filled a generic levothyroxine prescription between January 1, 2008, and June 30, 2019, and had a stable drug dose, the same drug manufacturer, and a normal TSH level (0.3-4.4 mIU/L) for at least 3 months before either continuing to take the same product or switching among generic levothyroxine products (index date). Patients were excluded if they were pregnant, had diagnosed hypopituitarism or hyperthyroidism, or had a medical condition or used medications that could affect thyrotropin levels. They were also excluded if they filled a prescription for other forms of thyroid replacement therapy between 6 months before the index date and when the first TSH level was obtained 6 weeks to 12 months after the index date. Data were analyzed from December 1, 2019, to November 24, 2021.
Main Outcomes and Measures
Proportion of individuals with a normal (0.3-4.4 mIU/L) or markedly abnormal (<0.1 or >10.0 mIU/L) TSH level using the first available laboratory result 6 weeks to 12 months after the index date. A propensity score model was developed to minimize confounding using logistic regression with the binary outcome of continuing the same sourced levothyroxine product vs switching generic levothyroxine. Covariates were demographics, comorbidities, and baseline TSH level. The balance among the treatment groups was evaluated by comparing standardized mean differences of baseline covariates between the groups.
A total of 15 829 patients filled generic levothyroxine (mean [SD] age, 58.9 [14.6] years; 73.4% [11 624] were women; 4.5%  were Asian, 10.2%  were Black, 11.4%  were Hispanic, and 71.4% [11 295] were White individuals); of these patients, 56.3%  received a daily levothyroxine dose of 50 μg or less. A total of 13 049 patients (82.4%) continued taking the same sourced preparation, and 2780 (17.6%) switched among generic levothyroxine preparations. Among 2780 propensity-matched patient pairs, the proportion of patients with a normal TSH level after the index date was 82.7% (2298) among nonswitchers and 84.5% (2348) among switchers (risk difference, −0.018; 95% CI, −0.038 to 0.002; P = .07). The proportion of patients with a markedly abnormal TSH level after the index date was 3.1% (87) among nonswitchers and 2.5% (69) among switchers (risk difference, 0.007; 95% CI, −0.002 to 0.015; P = .14). The mean (SD) TSH levels after the index date were 2.7 (2.3) mIU/L among nonswitchers and 2.7 (3.3) mIU/L among switchers (P = .94).
Conclusions and Relevance
Results of this comparative effectiveness research study suggest that switching among different generic levothyroxine products was not associated with clinically significant changes in TSH level. These findings conflict with the current guideline recommendation that warns clinicians about potential changes in TSH level associated with switching among levothyroxine products sourced from different manufacturers.
Levothyroxine sodium is one of the most commonly prescribed drugs in the US, with approximately 25 million individuals using levothyroxine in 2016.1-3 Levothyroxine is available as a generic or brand-name product. Generic levothyroxine preparations are less expensive and have been rated by the US Food and Drug Administration (FDA) as bioequivalent to their brand-name reference-listed drugs.4 However, generic levothyroxine has been less widely prescribed than other generic pharmaceutical products.3,5 One potential reason for this may be the lingering concerns about an association between switching levothyroxine products sourced from different manufacturers and the stability of thyroid hormone levels.4,6 When clinicians prescribe a particular brand-name product (indicating dispense as written), that specific product is dispensed to a patient for the duration of the prescription. In contrast, when clinicians prescribe generic levothyroxine, pharmacists can substitute a variety of generic levothyroxine preparations made by different manufacturers (eg, Mylan, Sandoz, or Lannett) without requiring clinician notification or approval.
The 2014 American Thyroid Association guideline, citing concerns about the bioequivalence methods suggested by the FDA for levothyroxine products, specifically recommends avoiding switches between levothyroxine products sourced from different manufacturers.6 To maintain the same levothyroxine product, the guideline recommends that clinicians prescribe either the same identifiable product of generic levothyroxine (ie, sourced from the same manufacturer) or a specific brand of levothyroxine. Because the latter strategy is logistically easier, the guidelines imply that prescribing generic levothyroxine may be associated with changes in thyroid hormone levels or dosages required. This guideline recommendation is at odds with the FDA position that approved generic levothyroxine products are expected to be interchangeable without the need for serum thyrotropin (TSH) testing.7,8
Given these concerns, it is important to understand the implications of switching among different generic levothyroxine products for TSH levels. Accordingly, using national data from a large administrative claims database, we compared the effectiveness and safety of generic levothyroxine between patients who did and did not switch manufacturers. Specifically, we examined the implications of switching among generic levothyroxine products for levels of TSH, the widely accepted laboratory test used to assess thyroid status.
We conducted a retrospective comparative effectiveness research study of deidentified administrative claims data (January 1, 2008, to June 30, 2019) linked to laboratory results from OptumLabs Data Warehouse, a large database that includes commercially insured and Medicare Advantage enrollees throughout the US.9 Data were analyzed from December 1, 2019, to November 24, 2021. Data on race and ethnicity were obtained from the database, which contains longitudinal health information on enrollees representing a diverse mix of ages, races and ethnicities, and geographic regions across the US.10 The health plans provide comprehensive full insurance coverage for physician, hospital, and prescription drug services. Pharmacy claims include information on medications dispensed, including doses, amount dispensed, and dates of prescriptions. Laboratory data, available for a subset of the cohort on the basis of data sharing agreements, include test names, Logical Observation Identifiers Names and Codes, and test results. Study data were accessed using techniques compliant with the Health Insurance Portability and Accountability Act of 1996. Because this study involved analysis of preexisting, deidentified data and was not considered human participant research, the Mayo Clinic Institutional Review Board declared it exempt from board approval and waived the requirement for informed consent. This study followed the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) reporting guideline.
We included adults aged 18 years or older who filled generic levothyroxine preparations from any of the 3 most common manufacturers (Mylan, Sandoz, and Lannett) between January 1, 2008, and June 30, 2019. The cohort included new and prevalent users. First, we required patients to have been taking levothyroxine for at least 12 months after the first fill date in our study period. For patients who continuously used the same sourced generic levothyroxine preparation, the index dates were randomly assigned to a fill date within 1 year of the first fill date after having been taking the same levothyroxine dose and preparation for at least 3 months. For patients who switched among levothyroxine preparations within 1 year of the first fill in the study period, we assigned an index date based on the date of the first switch—that is, the date of the fill for levothyroxine sourced from a different manufacturer than the original levothyroxine preparation used.
Once we identified the index date, we limited the study to patients who had continuous medical and pharmacy benefits for 6 months before the index date. This criterion ensured that at least 6 months of information on patients’ medical history and medication use were available for purposes of identifying eligible patients and capturing baseline characteristics for the matching process. Within this 6-month period before the index date, patients were required to have at least 3 months of constant generic levothyroxine use (stable dose and same manufacturer) and at least 1 available TSH level in the normal range (0.3-4.4 mIU/L).11 After the index date, patients were required to have at least 12 months of continuous medical and pharmacy benefits to capture the TSH levels available within 1 year and to adhere to the same dose and to use a preparation from the same manufacturer until the first TSH test result, which was obtained between 6 weeks and 12 months after the index date. Adherence was defined as having a proportion of days covered of more than 0.8.
We excluded adults who were pregnant, who had diagnosed hypopituitarism or hyperthyroidism, and who had a medical condition or used medications that could affect TSH levels (the conditions and medications are listed in eTable 1 in the Supplement). In addition, we excluded adults who filled other forms of thyroid replacement therapy—including thyroid extracts, or triiodothyronine therapy, such as liothyronine, thyroid desiccated or extracts, and the brands Cytomel, Armour Thyroid, or Nature-Throid—between 6 months before the index date and when the first TSH test result was obtained 6 weeks to 12 months after the index date.
Baseline patient characteristics included age, sex, race and ethnicity, census region, physician specialty (general practitioner, endocrinologist, or other specialist), length of stable levothyroxine dose before the index date, health plan type (commercial or Medicare Advantage), Charlson Comorbidity Index score (estimated using International Classification of Diseases, Ninth Revision, Clinical Modification, and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision diagnosis codes included in administrative claims), TSH level, levothyroxine dose (calculated based on fill data), conditions that may increase the risk of levothyroxine malabsorption (inflammatory bowel disease, anemia as surrogate of iron use) within 180 days of the index date, and use of estrogen within 90 days of the index date.12,13
We used First Databank to categorize fills as generic and to identify the specific levothyroxine manufacturer. First Databank categorizes pharmacy products as generic if they are sold under a generic pharmacy label, which includes authorized generics. The levothyroxine generic manufacturers included Mylan (56.8% [8983 of 15 829]), Sandoz (8.9% [1403 of 15 829]), and Lannett (34.4% [5443 of 15 829]). These 3 generic products are the most commonly prescribed in the US market.
We examined the proportion of individuals with normal, abnormal, and markedly abnormal TSH levels using the first TSH level available 6 weeks to 12 months after the index date. We used the following definitions of these outcomes: (1) normal TSH level (0.3-4.4 mIU/L), (2) abnormal TSH level (<0.3 or >4.4 mIU/L), and (3) markedly abnormal TSH level (<0.1 or >10 mIU/L) because these levels are associated with clinically significant harms owing to thyrotoxicosis or hypothyroidism.6,11 We also evaluated mean TSH levels after the index date and the mean change in level after the index date vs the baseline level.
The implications of switching generic products for thyroid hormone levels could be altered by the presence of endogenous production of thyroid hormone (patients with a functional thyroid gland).14 Therefore, we examined whether TSH levels differed among patients with and without endogenous thyroid hormone production. We categorized patients as not having endogenous thyroid production if they had a history of total thyroidectomy or history of thyroid cancer and if they received a high dose of thyroid hormone (>100 μg daily) that suppresses endogenous thyroid production.15
In addition, given that the TSH level is reliable as a treatment marker only after at least 6 weeks following initiation of therapy or change in treatment dose, we used the first available TSH test result between 6 weeks and 12 months after the index date for the primary analysis and the first available TSH test result between 6 and 12 weeks after the index date for a sensitivity analysis.6 We reestimated the propensity score using the same baseline variables and rematched the cohorts for all sensitivity analyses.
We used propensity score matching to minimize confounding. We developed a propensity score model using logistic regression with the binary outcome of continuing the same sourced levothyroxine product vs switching generic levothyroxine products.
Covariates included in the model were participant demographic characteristics, comorbidities, and the baseline TSH level, as shown in Table 1. Variables within these models were selected based on clinical relevance and evidence from previous studies. For each cohort, we evaluated the balance among the treatment groups by comparing standardized mean differences of baseline covariates between the groups. A baseline characteristic was considered balanced if the maximum standardized mean difference was less than 10%. Switchers were matched 1:1 to nonswitchers using nearest neighbor matching with a caliper of 0.2 of the SD of the logit of the propensity score. Using this matched cohort, we then tested for differences between patients who continued taking the same sourced levothyroxine products vs those who switched using χ2 tests for categorical outcomes and an unpaired t test for continuous outcomes as our primary analysis method. In addition, we used a robust estimator of variance to address possible concerns about the sample clustering within matched sets. To this end, we conducted a sensitivity analysis using simple linear regression for continuous outcomes and logistic regression for categorical outcomes comparisons with robust sandwich estimates. Finally, to account for a possible decrease in sample size at the moment of matching, we conducted a sensitivity analysis with all the data and with covariate adjustment. We considered a 2-sided P < .05 to be statistically significant. All analyses were conducted using SAS, version 9.4 (SAS Institute Inc) and Stata, version 14.1 (StatCorp, LLC).
We included 15 829 patients who filled generic levothyroxine prescriptions between January 1, 2008, and June 30, 2019 (eFigure in the Supplement). Of these patients (mean [SD] age, 58.9 [14.6] years), 73.4% (11 624) were women and 26.6% (4205) were men; 4.5% (705) were Asian, 10.2% (1617) were Black, 11.4% (1801) were Hispanic, and 71.4% (11 295) were White individuals (2.6%  had unknown race and ethnicity); and 56.3% (8905) received a daily levothyroxine dose of 50 μg or less. A total of 13 049 (82.4%) continued taking the same sourced preparation, and 2780 (17.6%) switched among generic levothyroxine preparations. Patients who continued taking the same preparation and those who switched were generally similar with respect to demographic and clinical characteristics at baseline except that patients who switched generic levothyroxine manufacturers had a shorter duration on a stable dose of levothyroxine therapy before the index date compared with nonswitchers (226 days vs 271 days). After 1:1 propensity score matching, no significant differences were detected with respect to all of the measured variables between the 2 groups of patients who continued taking the same preparation and those who switched between generic levothyroxine preparations (Table 1). For example, the duration on a stable dose of levothyroxine therapy before the index date was no longer different between the groups (223 days for nonswitchers and 225 days for switchers; standard mean difference, 0.02).
Among all 2780 propensity score–matched pairs of patients, the mean (SD) time from the index date to the first TSH test result was 162 (84) days for patients who did not switch and 157 (86) days for patients who switched generic levothyroxine products. Among matched patients, the proportion of patients with a normal TSH level between 6 weeks and 12 months after the index date was similar among nonswitchers (82.7% ) and switchers (84.5% ) (risk difference [RD], −0.018; 95% CI, −0.038 to 0.002; P = .07). The proportion of patients with markedly abnormal TSH levels between 6 weeks and 12 months after the index date was also similar among nonswitchers (3.1% ) and switchers (2.5% ) (RD, 0.007; 95% CI, −0.002 to 0.015; P = .14; Table 2). Mean (SD) TSH levels after the index date among nonswitchers and switchers were 2.7 (2.3) mIU/L and 2.7 (3.3) mIU/L, respectively (P = .94). Absolute mean change (SD) in TSH level was 0.51 (2.28) among nonswitchers and 0.53 (3.25) among switchers (P = .84).
Sensitivity analysis using TSH test results obtained between 6 and 12 weeks after the index date did not change these estimates (Table 3). In the subgroup of 364 patients receiving a levothyroxine dose of more than 100 μg daily, the proportion of nonswitchers vs switchers with normal (70.9%  vs 76.6% ; RD, −0.058; 95% CI, −0.122 to 0.006), abnormal (29.1%  vs 23.4% ; RD, 0.058; 95% CI, −0.006 to 0.122), and markedly abnormal (8.8%  vs 7.4% ; RD, 0.014; 95% CI, −0.026 to 0.053) TSH levels was not significantly different (Table 4). The study was not powered to conduct a sensitivity analysis in patients with total thyroidectomy and thyroid cancer because of the small sample size. The results of the analysis conducted using all data with covariate analysis and analysis using robust estimator of variance were the same as those of the primary analysis (eTables 2-7 in the Supplement).
Using national data from a large administrative claims database, we found that the proportion of patients with normal TSH levels did not differ between patients who continued to use the same generic levothyroxine product and patients who switched among different generic levothyroxine products. Moreover, we found no significant differences in the proportion of patients with markedly abnormal TSH levels or in the mean change in TSH levels among patients who continued taking the same generic levothyroxine vs those who switched. These results are reassuring with respect to the effectiveness of generic levothyroxine products, even when switches occur between preparations sourced from different manufacturers.
The questions about bioequivalence of generic and brand-name levothyroxine products and whether they are interchangeable have been raised for a quarter century.16 Levothyroxine was first introduced to the market in 1962 and did not require a new drug application by the FDA because it was not considered a new drug.17 The lack of a new drug application process meant that there were no clear standards for bioequivalence between levothyroxine products. As a result, the market was dominated mainly by the first synthetic version of levothyroxine, Synthroid.18 In 1997, Dong et al19 published a single-blind, randomized, 4-way crossover trial that included 22 women with hypothyroidism. Using standard bioequivalence criteria, the study showed that 4 generic and brand-name levothyroxine products were equivalent and interchangeable. In 1997, the FDA, citing potency stability, consistency, and bioavailability issues, concluded that a new drug application would be required to bring new levothyroxine products to the market after August 2000, including the levothyroxine products included in the study by Dong et al.20 At that time, the FDA also provided guidelines for levothyroxine bioequivalence. Since then, several levothyroxine products have been approved and are used in clinical practice. Professional societies, however, disagreed with some of the FDA levothyroxine bioequivalence guidelines, citing that they would not be able to detect clinically important differences among the products.4 In fact, in the guideline statements, experts recommend maintaining the same sourced levothyroxine preparation throughout treatment.6 Given that generic levothyroxine products can be switched at the pharmacy level without notifying or requiring approval from clinicians or patients, this recommendation may have indirectly fueled the preferential use of brand-name levothyroxine, particularly among endocrinologists. In 2016, approximately 40% of prescriptions for levothyroxine written by endocrinologists were for brand-name levothyroxine vs 15% of those prescribed by primary care practitioners.3
The present study does not address the issue of therapeutic equivalence among levothyroxine products. Instead, it focuses on the association between different generic levothyroxine products and TSH levels, and the results conflict with the current expert guideline recommendation, which universally warns patients and clinicians about possible changes in TSH level associated with switching among levothyroxine products sourced from different manufacturers.
Our findings may help curtail the relative underuse of generic levothyroxine in the US because one of the main concerns of prescribing generic levothyroxine is the lack of control over switching among different generic products. Generic levothyroxine use, although increasing during the past decade, continues to be low compared with use of other generic drugs, which usually dominate the market. From 2007 to 2016, approximately 74% of thyroid hormone prescriptions were for generic levothyroxine in the US compared with 97% of other drugs that had both brand-name and generic options available.3,21 The underuse of generic levothyroxine has considerable financial implications for patients and health care systems. Brand-name levothyroxine is several times more expensive than generic levothyroxine; information available from the OptumLabs Data Warehouse suggests that in 2016, the median out-of-pocket cost of a 30-day prescription for branded thyroid hormone was $25 to $31 for insured patients, substantially more than the $2 to $10 median out-of-pocket cost for a 30-day prescription for generic thyroid hormone products.
This study has limitations. The data set included only patients who were commercially insured or used Medicare Advantage, which raises concerns about the representativeness of the sample used because it did not include patients with Medicaid or patients without insurance. Yet the distribution of age, sex, and minority groups in the OptumLabs Data Warehouse population is largely consistent with the general US population, and therefore, findings are likely generalizable to most adults.9 The present study was underpowered to assess the implications of generic-to-generic switching for TSH levels among patients without any residual endogenous thyroid function (eg, patients after thyroidectomy). However, most patients in the US taking levothyroxine have partial thyroid dysfunction (eg, subclinical hypothyroidism) and may maintain some endogenous thyroid hormone production.15,22 Furthermore, in the data set, approximately 40% of patients had a link to laboratory results based on a specific laboratory vendor, which limited the sample size. Yet because the use of a specific laboratory vendor is likely to be random, we do not believe this introduced substantial bias in the results.
The present study did not address the implications of switching among brand levothyroxine products and of switching from brand to generic levothyroxine. Furthermore, the study did not examine the implications of the direction of generic levothyroxine manufacturer switching (eg, switching from specific manufacturer to another manufacturer) on thyroid hormone levels. Instead, we used a more pragmatic assessment of the consequences of switching for the most common generic levothyroxine products on the market. The analysis was observational, and residual confounding may exist despite the use of propensity score matching. Because the exposure (nonswitching vs switching) is likely a random event, with little influence from patient or clinician factors, the remaining risk factors of confounding and selection bias are likely small.
This comparative effectiveness research study suggests that switching among generic levothyroxine products sourced from different manufacturers was not associated with differences in achieving a normal TSH level or with changes in TSH levels compared with continuing to take the same sourced levothyroxine. These results may reassure patients and clinicians that switching among generic levothyroxine products is unlikely to have substantial implications for treatment effects. The results conflict with the current expert guideline recommendation that warns patients and clinicians about possible consequences of switching among levothyroxine products for TSH levels.
Accepted for Publication: December 3, 2021.
Published Online: February 28, 2022. doi:10.1001/jamainternmed.2022.0045
Corresponding Author: Juan P. Brito, MD, MSc, Division of Endocrinology, Diabetes, Metabolism, and Nutrition, Department of Internal Medicine, Mayo Clinic, 200 First St SW, Rochester, MN 55905 (email@example.com).
Author Contributions: Drs Brito and Qiang 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.
Concept and design: Brito, Ross, Qiang, Wang, Zhao, Shah, Lipska.
Acquisition, analysis, or interpretation of data: Brito, Deng, Ross, Choi, Graham, Qiang, Rantou, Lipska.
Drafting of the manuscript: Brito.
Critical revision of the manuscript for important intellectual content: Deng, Ross, Choi, Graham, Qiang, Rantou, Wang, Zhao, Shah, Lipska.
Statistical analysis: Deng, Choi, Rantou.
Obtained funding: Ross.
Administrative, technical, or material support: Qiang, Wang, Shah.
Supervision: Brito, Wang, Zhao.
Conflict of Interest Disclosures: Dr Ross reported receiving research support through Yale University from the US Food and Drug Administration (FDA), Johnson & Johnson, Medtronic, Medical Devices Innovation Consortium, Blue Cross Blue Shield Association, Centers for Medicare & Medicaid Services (CMS), Agency for Healthcare Research and Quality (AHRQ), National Institutes of Health (NIH) National Heart, Lung, and Blood Institute (NHLBI), and the Laura and John Arnold Foundation; he reported serving as an expert witness at the request of a relator’s attorneys at the Greene Law Firm in a qui tam suit alleging violations of the False Claims Act and Anti-Kickback Statute against Biogen Inc. Dr Shah reported receiving grants from the FDA and research support through the Mayo Clinic from CMS, AHRQ, NIH NHLBI, National Science Foundation, and Patient-Centered Outcomes Research Institute (PCORI). Dr Lipska reported receiving grants from the National Institute on Aging and PCORI and receiving suppport from CMS to develop and evaluate publicly reported quality measures. No other disclosures were reported.
Funding/Support: This project was supported by grant U01FD005938 from the FDA of the US Department of Health and Human Services (HHS) as part of a financial assistance award (Center of Excellence in Regulatory Science and Innovation grant to Yale University and Mayo Clinic) totaling $158 430, with 100% funded by the FDA and HHS.
Disclaimer: The views expressed here are those of the authors and do not necessarily reflect the position or policy of the FDA or HHS or the US government.