Systemic Safety Evaluation of Levothyroxine Therapy in Thyroid Cancer Survivors: A Nationwide 20-Year Pharmacoepidemiologic Study = 갑상선암 생존자에서 레보티록신 치료의 전신적 안전성 평가: 전국 단위 20년 약물역학 연구

다국어 초록 (Multilingual Abstract)

Background: Levothyroxine (LT4) therapy is a cornerstone of long-term management in thyroid cancer survivors, serving both as thyroid hormone replacement and, in selected patients, as thyroid-stimulating hormone (TSH) suppression to reduce recurrence risk. With the rapidly increasing incidence of differentiated thyroid cancer and excellent long-term survival, the population of thyroid cancer survivors receiving lifelong LT4 therapy has grown substantially. As a result, concerns have emerged regarding the potential long-term systemic effects of chronic LT4 exposure, particularly when administered at suppressive or near-suppressive doses over decades. Previous studies evaluating the safety of LT4 therapy have reported inconsistent findings, often focusing on single outcomes, short-term exposure, or baseline dose categories. Moreover, many investigations have not adequately addressed time-varying exposure, cumulative hormonal burden, or differences related to surgical extent, such as lobectomy versus total thyroidectomy. Consequently, the long-term, systemic safety profile of LT4 therapy across multiple disease domains remains incompletely characterized.
Objective and Hypothesis: The primary objective of this study was to comprehensively evaluate the long-term systemic safety of LT4 therapy in thyroid cancer survivors using a nationwide population-based cohort with extended follow-up. We aimed to assess associations between LT4 exposure and a broad spectrum of metabolic, cardiovascular, skeletal, and oncologic outcomes, while explicitly accounting for time-varying exposure, cumulative dose, cumulative duration, and average daily dose. We hypothesized that (1) the long-term effects of LT4 therapy are outcome-specific rather than uniformly harmful, (2) cumulative exposure metrics are more strongly associated with adverse outcomes than short-term or average dose intensity, and (3) exposure–response patterns differ according to surgical extent, with total thyroidectomy patients exhibiting greater vulnerability to long-term cumulative LT4 exposure.
Materials and Methods: This nationwide retrospective cohort study was conducted using the Korean National Health Insurance database, which provides near-complete coverage of the Korean population. Adults diagnosed with thyroid cancer who underwent thyroid surgery between 2002 and 2022 were identified and followed longitudinally. Patients were classified according to surgical extent (lobectomy or total thyroidectomy), and LT4 exposure was ascertained using prescription claims. LT4 use was modeled as a time-varying exposure. Three complementary exposure metrics were constructed: cumulative dose, cumulative duration, and average daily dose. These metrics were updated over follow-up and categorized into quartiles based on the distribution among LT4 users, with additional analyses using group-specific quartiles to account for differences in hormonal requirements between surgical groups. Primary outcomes included diabetes mellitus, dyslipidemia, hypertension, cardiovascular disease, cerebrovascular disease, arrhythmia, osteoporosis, fractures, breast cancer, and prostate cancer. Cox proportional hazards models with time-varying covariates were used to estimate adjusted hazard ratios (HRs) and 95% confidence intervals (CIs), controlling for demographic factors, comorbidities, and healthcare utilization. Restricted cubic spline models were applied to evaluate nonlinear exposure–response relationships. Sensitivity analyses were performed for sex- and age-specific subgroups, including premenopausal and postmenopausal women.
Results: The study included a large nationwide cohort of thyroid cancer survivors with up to 20 years of follow-up, yielding extensive person-time for outcome assessment. Overall, LT4 therapy was not associated with a uniform increase in adverse outcomes; instead, risk patterns varied substantially by outcome category, exposure metric, and surgical extent. Metabolic outcomes, including diabetes mellitus and dyslipidemia, demonstrated largely neutral or modestly reduced risks among LT4 users compared with non-users. These patterns were consistent across exposure metrics and were more pronounced among total thyroidectomy patients. Spline analyses revealed flat or slightly downward-sloping risk curves, indicating no evidence of harmful dose–response relationships for these outcomes. Cardiovascular outcomes showed a distinct pattern. Higher cumulative dose and longer cumulative duration of LT4 therapy were associated with increased risks of cardiovascular disease, cerebrovascular disease, arrhythmia, and hypertension, particularly among total thyroidectomy patients. In contrast, average daily dose showed weaker or non-monotonic associations. Restricted cubic spline analyses demonstrated gradual and monotonic risk increases with prolonged cumulative exposure, suggesting that long-term hormonal burden, rather than short-term dosing intensity, is the primary driver of cardiovascular risk. Skeletal outcomes exhibited divergent findings. Osteoporosis risk increased in a cumulative, exposure-dependent manner, especially among total thyroidectomy patients and postmenopausal women. In contrast, fracture risk remained largely neutral across exposure metrics and subgroups, with no consistent exposure–response gradients observed. Cancer outcomes, including breast and prostate cancer, showed predominantly null associations with LT4 exposure across all analytic approaches. Sensitivity analyses in sex- and age-specific subgroups confirmed the robustness of these findings, with stronger exposure–response relationships observed for osteoporosis in postmenopausal women and for cardiovascular outcomes in long-term, high cumulative exposure settings.
Conclusions: In this nationwide 20-year pharmacoepidemiologic study, long-term LT4 therapy in thyroid cancer survivors exhibited a largely favorable and differentiated safety profile rather than broad systemic harm. Potential safety concerns were outcome-specific and primarily associated with prolonged cumulative exposure, particularly for cardiovascular and skeletal outcomes among total thyroidectomy patients. In contrast, metabolic and secondary cancer outcomes remained largely neutral. These findings underscore the importance of cumulative exposure assessment and support a risk-adapted approach to long-term LT4 management in thyroid cancer survivorship. Periodic reassessment of TSH targets and LT4 dosing intensity, guided by recurrence risk and cumulative exposure burden, may help optimize long-term outcomes. More broadly, this study highlights the value of advanced longitudinal exposure modeling and real-world data in evaluating the long-term safety of chronic hormone therapy.

국문 초록 (Abstract)

배경: 레보티록신(levothyroxine, LT4) 치료는 갑상선암 생존자에서 갑상선호르몬 보충 요법의 핵심이며, 일부 환자에서는 재발 위험 감소를 목적으로 한 갑상선자극호르몬(thyroid-stimulating hormone, TSH) 억제 요법으로 사용된다. 분화갑상선암의 발생률 증가와 우수한 장기 생존율로 인해, 평생 LT4 치료를 받는 갑상선암 생존자 인구는 지속적으로 증가하고 있다. 이에 따라 수십 년에 걸친 장기적 LT4 노출이 전신 건강에 미치는 잠재적 영향에 대한 우려가 제기되고 있다. 기존 연구들은 LT4 치료의 안전성을 평가함에 있어 단일 질환, 단기간 노출, 혹은 기저 용량 분류에 초점을 맞춘 경우가 많았으며, 시간에 따라 변화하는 노출, 누적 호르몬 부담, 수술 범위(엽절제술 대 전절제술)에 따른 차이를 충분히 고려하지 못한 한계가 있었다. 그 결과 LT4 장기 치료의 전신적 안전성에 대한 종합적이고 체계적인 평가는 아직 충분히 이루어지지 않았다.
목적 및 가설: 본 연구의 목적은 전국 단위 인구 기반 코호트와 장기간 추적 자료를 활용하여, 갑상선암 생존자에서 LT4 치료의 장기적 전신 안전성을 종합적으로 평가하는 것이다. 이를 위해 시간가변 노출 모형과 누적 노출 지표(누적 용량, 누적 치료 기간, 평균 일일 용량)를 적용하여, LT4 노출과 대사성, 심혈관계, 골격계 및 종양성 질환 발생 위험 간의 연관성을 분석하였다. 본 연구는 다음과 같은 가설을 설정하였다. 첫째, LT4 장기 노출의 영향은 전신적으로 균일한 위해 효과를 나타내기보다는 질환별로 상이할 것이다. 둘째, 단기 또는 평균 용량보다 누적 노출 지표가 장기 부작용과 더 강한 연관성을 보일 것이다. 셋째, 수술 범위에 따라 노출–반응 양상이 달라지며, 전절제술 환자에서 누적 LT4 노출에 따른 위험 증가가 더 뚜렷할 것이다.
대상 및 방법: 본 연구는 국민건강보험공단 자료를 이용한 후향적 전국 단위 코호트 연구로 수행되었다. 2002년부터 2022년까지 갑상선암 진단 후 수술을 받은 성인을 연구 대상자로 선정하였으며, 엽절제술과 전절제술에 따라 수술 범위를 구분하였다. LT4 노출은 처방 기록을 기반으로 확인하였다. LT4 사용은 시간가변 노출로 모형화하였으며, 세 가지 상호보완적인 노출 지표를 구축하였다: 누적 용량, 누적 치료 기간, 평균 일일 용량. 각 지표는 추적 기간 동안 갱신되었으며, LT4 사용자 전체 분포를 기준으로 사분위수로 범주화하였다. 또한 수술 범위에 따른 호르몬 요구량 차이를 고려하기 위해 군별 사분위 분석을 추가로 수행하였다. 주요 연구 결과는 당뇨병, 이상지질혈증, 고혈압, 심혈관질환, 뇌혈관질환, 부정맥, 골다공증, 골절, 유방암 및 전립선암 발생이었다. 시간가변 공변량을 포함한 Cox 비례위험 모형을 이용하여 보정 위험비(hazard ratio, HR)와 95% 신뢰구간을 산출하였으며, 인구학적 요인, 동반질환, 의료 이용 지표를 보정하였다. 제한적 삼차 스플라인(restricted cubic spline) 모형을 통해 비선형 노출–반응 관계를 평가하였고, 폐경 전·후 여성 등 성별 및 연령별 민감도 분석을 수행하였다.
결과: 본 연구는 최대 20년에 이르는 장기 추적 기간을 포함한 대규모 갑상선암 생존자 코호트를 분석하였으며, 충분한 관찰 인년을 확보하였다. 전반적으로 LT4 치료는 모든 질환에서 일관된 위험 증가를 보이지 않았으며, 질환 유형, 노출 지표, 수술 범위에 따라 상이한 양상을 보였다. 당뇨병과 이상지질혈증 등 대사성 질환에서는 LT4 사용자에서 비사용자 대비 위험이 중립적이거나 소폭 감소하는 경향을 보였으며, 이러한 양상은 특히 전절제술 환자에서 뚜렷하였다. 스플라인 분석에서도 명확한 용량–반응 관계는 관찰되지 않았다. 반면 심혈관계 질환에서는 다른 양상이 관찰되었다. 누적 용량과 누적 치료 기간이 증가할수록 심혈관질환, 뇌혈관질환, 부정맥, 고혈압 발생 위험이 증가하였으며, 이러한 경향은 전절제술 환자에서 더욱 뚜렷하였다. 평균 일일 용량은 상대적으로 약하거나 비일관된 연관성을 보였다. 스플라인 분석 결과, 장기간 누적 노출에 따라 위험이 점진적으로 증가하는 단조 증가 패턴이 확인되었다. 골격계 결과에서는 이질적인 양상이 나타났다. 골다공증은 누적 노출 증가에 따라 위험이 상승하였으며, 특히 전절제술 환자와 폐경 후 여성에서 뚜렷한 노출–반응 관계가 관찰되었다. 반면 골절 위험은 대부분의 노출 지표와 하위군에서 중립적인 결과를 보였다. 유방암과 전립선암 등 이차 암 발생은 모든 분석에서 LT4 노출과 유의한 연관성을 보이지 않았다.
결론: 본 전국 단위 20년 추적 연구에서, 갑상선암 생존자의 장기 LT4 치료는 전신적 위해를 초래하는 일관된 위험 증가와는 거리가 먼, 질환별로 구분되는 안전성 프로파일을 보였다. 잠재적 위험 증가는 주로 심혈관계 및 골격계 질환에 국한되었으며, 특히 전절제술 환자에서 장기간 누적 노출과 연관되어 나타났다. 반면 대사성 질환과 이차 암 발생 위험은 전반적으로 중립적이었다. 본 연구 결과는 갑상선암 생존자에서 장기 LT4 치료 관리 시 누적 노출을 고려한 위험 적응형 접근의 중요성을 시사한다. 재발 위험과 누적 호르몬 부담을 함께 고려하여 TSH 목표와 LT4 용량을 주기적으로 재평가하는 전략이 장기 예후 최적화에 기여할 수 있을 것이다. 더 나아가, 본 연구는 실세계 자료와 고급 노출 모형을 활용한 장기 약물 안전성 평가의 중요성을 강조한다.

목차 (Table of Contents)

• Abstract i
• Chapter 1. Introduction １
• 1.1. Study Background １
• 1.1.1. Thyroid cancer epidemiology and survivorship context １
• 1.1.2. Levothyroxine therapy after thyroidectomy and clinical rationale ３
• 1.1.3. Systemic physiological consequences of chronic LT4 exposure ３
• 1.2. Prior research and conceptual framework ４
• 1.2.1. IARC Group I pharmaceutical program ４
• 1.2.2. From carcinogenic risk to systemic safety ５
• 1.3. Rationale for outcome selection ５
• 1.3.1. Overview ５
• 1.3.2. Metabolic diseases ６
• 1.3.3. Cardiovascular and cerebrovascular diseases ６
• 1.3.4. Musculoskeletal diseases ６
• 1.3.5. Sex-specific cancers ７
• 1.4. Objectives and hypothesis ７
• 1.4.1. Study objectives ７
• 1.4.2. Central hypothesis ８
• 1.4.3. Hypothesis 1: Ever-use vs non-use ９
• 1.4.4. Hypothesis 2: Exposure metrics １０
• 1.4.5. Hypothesis 3: Modeling coherence and temporal validation １１
• 1.4.6. Analytical implications and anticipated contributions １２
• Chapter 2. Materials and methods １４
• 2.1. Data source １４
• 2.2. Study population １５
• 2.2.1. Identification of the study population １５
• 2.2.2. Exclusion criteria １５
• 2.2.3. Sex-specific analyses and surgical subgroup classification １６
• 2.3. Time origin, follow-up, and immortal-time control １８
• 2.4. Exposure ascertainment and quantification １９
• 2.4.1. Conceptual dimensions and coding １９
• 2.4.2. Operational equations ２０
• 2.5. Outcomes and competing events ２１
• 2.6. Covariates and confounding control ２４
• 2.7. Primary exposure contrasts and analytic subgroups ２６
• 2.8. Statistical modeling strategy ２７
• 2.8.1. Cox proportional hazards model ２７
• 2.8.2. Restricted cubic splines for dose–response ２８
• 2.8.3. Sequence symmetry analysis ２８
• 2.9. Subgroups, sensitivity analyses, and diagnostics ２９
• 2.10. Variable coding, scaling, and outlier handling ３０
• 2.11. Software and reproducibility ３１
• Chapter 3. Results ３２
• 3.1. Study flow ３２
• 3.2. Baseline characteristics ３５
• 3.2.1. Baseline characteristics for metabolic diseases ３５
• 3.2.1.1. DM ３５
• 3.2.1.2. HTN ３６
• 3.2.1.3. DL ３８
• 3.2.2. Baseline characteristics for circulatory system diseases ４０
• 3.2.2.1. CVD ４０
• 3.2.2.2. Arrhythmia ４１
• 3.2.2.3. CeVD ４３
• 3.2.3. Baseline characteristics for musculoskeletal diseases ４５
• 3.2.3.1. OP ４５
• 3.2.3.2. Fx ４８
• 3.2.4. Baseline characteristics for sex-specific cancers ５０
• 3.2.4.1. BC ５０
• 3.2.4.2. PC ５１
• 3.3. Sequence symmetry analyses ６６
• 3.3.1. Overview of SSA findings ６６
• 3.4. Summary of effect estimates ６９
• 3.4.1. Effect estimates of metabolic diseases ６９
• 3.4.1.1. DM ６９
• 3.4.1.2. HTN ７２
• 3.4.1.3. DL ７６
• 3.4.2. Effect estimates of circulatory system diseases ７９
• 3.4.2.1. CVD ７９
• 3.4.2.2. Arrhythmia ８３
• 3.4.2.3. CeVD ８６
• 3.4.3. Effect estimates of musculoskeletal diseases ８９
• 3.4.3.1. OP ８９
• 3.4.3.2. Fx ９７
• 3.4.4. Effect estimates of sex-specific cancers １０３
• 3.4.4.1. BC １０３
• 3.4.4.2. PC １０５
• 3.5. Summary of restricted cubic spline １９２
• 3.5.1. RCS of metabolic diseases １９２
• 3.5.1.1. DM １９２
• 3.5.1.2. HTN １９３
• 3.5.1.3. DL １９４
• 3.5.2. RCS of circulatory system diseases １９５
• 3.5.2.1. CVD １９５
• 3.5.2.2. Arrhythmia １９７
• 3.5.2.3. CeVD １９８
• 3.5.3. RCS of musculoskeletal diseases １９９
• 3.5.3.1. OP １９９
• 3.5.3.2. Fx ２０２
• 3.5.4. RCS of sex-specific cancers ２０５
• 3.5.4.1. BC ２０５
• 3.5.4.2. PC ２０６
• Chapter 4. Discussion ２２２
• 4.1. Principal findings ２２２
• 4.1.1. Overall safety profile of LT4 in thyroid cancer survivors ２２２
• 4.1.2. Outcome-specific patterns ２２６
• 4.1.2.1. Cardiovascular disease, arrhythmia, and cerebrovascular disease ２２６
• 4.1.2.2. Metabolic outcomes: diabetes, hypertension, and dyslipidemia ２２７
• 4.1.2.3. Bone outcomes: osteoporosis and fracture ２２８
• 4.1.2.4. Secondary malignancies: breast and prostate cancer ２２９
• 4.1.3. Temporal patterns and surgery-related subgroups ２３０
• 4.1.3.1. Temporal patterns according to cumulative exposure and duration ２３０
• 4.1.3.2. Differential patterns by extent of surgery ２３１
• 4.2. Comparison with previous literature ２３２
• 4.2.1. Pharmacologic safety studies of long-term LT4 therapy ２３３
• 4.2.1.1. Cardiovascular and arrhythmia safety ２３３
• 4.2.1.2. Skeletal safety: osteoporosis and fracture ２３５
• 4.2.1.3. Metabolic outcomes and cardiometabolic risk factors ２３６
• 4.2.1.4. Subsequent primary cancers and other long-term outcomes ２３７
• 4.2.2. Agreement with or divergence from study hypotheses ２３８
• 4.2.2.1. Agreement with hypotheses: cumulative exposure and outcome specificity ２３９
• 4.2.2.2. Refinement of hypotheses: daily dose versus cumulative burden ２４０
• 4.2.2.3. Agreement with hypotheses: surgery-related risk stratification ２４１
• 4.2.2.4. Divergence from expectations: metabolic and oncologic neutrality ２４２
• 4.3. Biological mechanisms underlying the observed associations ２４３
• 4.3.1. Cumulative thyroid hormone exposure as a driver of long-term risk ２４３
• 4.3.2. Cardiovascular system: autonomic tone, myocardial remodeling, and vascular effects ２４４
• 4.3.3. Skeletal system: bone remodeling imbalance and duration dependence ２４５
• 4.3.4. Metabolic outcomes: divergent effects on glucose and lipid metabolism ２４６
• 4.3.5. Cancer-related outcomes: proliferation signaling versus clinical neutrality ２４６
• 4.3.6. Surgical extent and loss of endogenous buffering ２４７
• 4.4. Methodological considerations ２４８
• 4.4.1. Limitations ２４８
• 4.4.2. Strengths ２５１
• 4.5. Clinical and public health implications ２５３
• 4.6. Future research directions ２５６
• 4.6.1. Establishing causality through advanced observational and experimental designs ２５６
• 4.6.2. Biological mechanisms and tissue-level effects ２５７
• 4.6.3. Heterogeneity in LT4 response and precision medicine approaches ２５８
• 4.6.4. Long-term survivorship and multimorbidity frameworks ２５８
• 4.6.5. Global, population-level evidence and health policy ２５９
• Chapter 5. Conclusion ２６１
• 5.1. Summary ２６１
• 5.2. Concluding remarks ２６３
• Bibliography ２６５
• 요약(국문초록) ２７５