Long-Term Outcome in Patients With Heart Failure Treated With Levothyroxine

An Observational Nationwide Cohort Study

Mette Nygaard Einfeldt; Anne-Marie Schjerning Olsen; Søren Lund Kristensen; Usman Khalid; Jens Faber; Christian Torp-Pedersen; Gunnar H Gislason; Christian Selmer


J Clin Endocrinol Metab. 2019;104(5):1725-1734. 

In This Article

Materials and Methods

Setting and Data Sources

This study is based on comprehensive health care usage registers from Denmark. Use of the unique personal identification number, assigned to all Danish residents, enabled individual-level linkage between the national administrative registers.[18] Five of these registers were used in this study. First, the Danish National Patient Register holds records of all hospital admissions since 1977. All admissions have been registered with one main discharge diagnosis and, if applicable, one or more supplemental discharge diagnoses coded according to the International Classification of Diseases (ICD-8 until 1994 and ICD-10 from 1994).[19] Second, the Civil Registration system records deaths for all Danish citizens and provided information about vital status.[18] Third, the Danish Register of Causes of Death provided information about specific causes of death.[20] Fourth, the Danish Register of Medicinal Product Statistics, which keeps records on all claimed prescriptions [coded according to the international Anatomical Therapeutic Chemical (ATC) Classification] from pharmacies in Denmark since 1994, provided information about prescribed medication.[21] The register also holds information about quantity, strength, and date of dispensation as well as formulation and the affiliation of the physician issuing the prescription. Finally, annual incomes were retrieved from the Danish registers on personal income and transfer payments from the Danish Labor Market.[22] Socioeconomic status was defined as the average yearly gross household income in a 5-year period before inclusion in the study.

The Study Population

The study cohort comprised all Danish citizens aged $18 years with a first-time HF-related hospitalization in the period 1997 to 2012. The diagnosis of HF (ICD-10: I50) was validated with a specificity of 99%.[23] Additionally, a recent study confirmed that the validity of cardiovascular diagnoses in the Danish National Patient Register is high and sufficient for use in research.[24] To avoid selection bias in L-T4 exposure caused by the high mortality associated with HF, the cohort was restricted to patients still alive 30 days after discharge. Patients with a history of antithyroid drugs or amiodarone treatment were excluded from the study (Table 1).

Study Design

The study was a register-based historical cohort study of patients with HF diagnosed in the period between 1 January 1997 and 31 December 2012. We identified a population of all patients with HF and categorized the patients by L-T4 treatment status. We use the terms "treated" and "received treatment" to mean patients who were prescribed and claimed one or more prescriptions for L-T4 during the study period. The first group was defined as patients already receiving L-T4 treatment at baseline (the date they were diagnosed with HF). The second group consisted of patients who initiated L-T4 treatment at some point during follow-up (after diagnosis with HF) but were not treated at baseline. The final group consisted of patients who did not receive L-T4 treatment. Patients contributed with risk time in the untreated group until they were prescribed with L-T4, at which time they were moved from the untreated group to the treated group. Furthermore, patient groups were subcategorized by sex and age, (<65 years and >65 years). Patients were followed until death, emigration, prescription of amiodarone, or end of study, whichever came first.

Comorbidity and Concomitant Medical Therapy

Comorbidities such as ischemic heart disease, atrial fibrillation, diabetes, previous stroke, and MI were identified from the Danish National Patient Register (Table 1). Charlson Comorbidity Index was calculated on basis of prespecified diagnoses up to 5 years before cohort entry.[25,26] Concomitant treatment status was defined for β-blockers, angiotensin-converting enzyme (ACE) inhibitors, statins, aldosterone antagonists, digoxin, and loop diuretics.

Dose and Duration of L-T4 Treatment

The Danish Register of Medical Product Statistics does not include information on prescribed daily dosage of the medication. We therefore used an algorithm for L-T4 in which a minimum, maximum, and typical daily dosage range of the used medication were defined. For each patient, L-T4 treatment periods were calculated by dividing the number of tablets dispensed by the estimated daily dosage. The estimated daily dosage for each patient was calculated by comparing the cumulated dosage and the elapsed time between seven successive prescriptions for L-T4. This algorithm allowed the dosage to change when a new prescription was dispensed. This method, used to determine the average treatment time and dosage, has previously been described.[27,28]


The primary outcome of interest was all-cause mortality, with the secondary outcomes being MI, including both fatal and nonfatal MI; cardiovascular death, defined as any death related to the cardiovascular system; and MACE, defined as cardiovascular death, fatal or nonfatal MI, or stroke (Table 1).

Statistical Analysis

Baseline characteristics are presented as numbers with percentages for categorical variables and as means 6 SDs for continuous variables. Median follow-up time and average treatment time are reported with interquartile ranges (IQRs). Incidence rates (IRs) were calculated as number of events per 1000 person-years (py), stratified by L-T4 treatment status and average daily dosage. Incidence rate ratios (IRRs) with 95% CIs for each study outcome were estimated by time-dependent Poisson regression models adjusted for age, sex, and Charlson Comorbidity Index. The model therefore included two time scales: calendar time, with bands split in 1-year periods after 1 January 1997, and duration time after the HF diagnosis. Age was calculated at the beginning of each interval. Patients were censored at the time of death, emigration, or end of study (31 December 2012). A 95% significance level was used in all analyses, including the test of interactions.

A number of sensitivity analyses were performed to validate the primary findings. First we adjusted the main model for ischemic heart disease and loop diuretic status at baseline to account for previous heart disease and the potentially severe degree of HF indicated by the presence of a prescription for loop diuretics. Second, we adjusted the main model for socioeconomic status. Third, we made an identical sensitivity analysis including only patients with a Charlson Comorbidity Index equal to 0, corresponding to no known comorbidities. Then we modified the main analysis to also include patients who did not survive beyond the first 30 days after their first HF discharge. Finally, time was stratified on 3- and 6-month of follow-up by splitting the time into smaller intervals to ensure a constant rate in time division (bands were split every 3 and 6 months from 1 January 1997).

All statistical analyses were performed with the SAS Statistical Software package version 9.4 (SAS Institute Inc., Gary, NC) and Stata Software version 14 (StataCorp, College Station, TX).


Register-based studies do not require ethical approval in Denmark. The Danish Data Protection Agency approved this study (ref. no. 2007–58-0015/GEH-2014-018; I-Suite no. 02736), and data were made available for this study in an anonymized format preventing identification of individuals.