Summary
Background and objectives Conflicting evidence exists with regard to the association of thyroid hormones and mortality in dialysis patients. This study assesses the association between basal and trimestral variation of thyroid stimulating hormone, triiodothyronine, and thyroxine and mortality.
Design, setting, participants, & measurements In 210 prevalent hemodialysis patients, serum triiodothyronine, thyroxine, thyroid stimulating hormone, and interleukin-6 were measured 3 months apart. Cardiovascular and non-cardiovascular deaths were registered during follow-up. Based on fluctuations along tertiles of distribution, four trimestral patterns were defined for each thyroid hormone: persistently low, decrease, increase, and persistently high. The association of baseline levels and trimestral variation with mortality was investigated with Kaplan–Meier curves and Cox proportional hazard models.
Results During follow-up, 103 deaths occurred. Thyroid stimulating hormone levels did not associate with mortality. Patients with relatively low basal triiodothyronine concentrations had higher hazards of dying than patients with high levels. Longitudinally, patients with persistently low levels of triiodothyronine during the 3-month period had higher mortality hazards than those having persistently high levels. These associations were mainly attributable to cardiovascular-related mortality. The association between thyroxine and mortality was not altered after adjustment for triiodothyronine.
Conclusions Hemodialysis patients with reduced triiodothyronine or thyroxine levels bear an increased mortality risk, especially due to cardiovascular causes. This was true when considering both baseline measurements and trimestral variation patterns. Our longitudinal design adds observational evidence supporting the hypothesis that the link may underlie a causal effect.
Introduction
A reduction in kidney function encompasses a number of alterations in thyroid hormone metabolism that renders a particular “non-thyroidal illness” syndrome (1) in patients with CKD, where low levels of both total triiodothyronine (T3) and free triiodothyronine (fT3) are hallmark findings. Recent reports suggest that as many as 70% of patients with ESRD present with low T3 levels (2–5) and as many as 20%–25% have subclinical hypothyroidism (6,7). The underlying pathophysiology of these derangements is likely multifactorial, involving iodine retention, altered serum protein binding capacity, systemic inflammation, and peripheral deiodinase activity (8,9).
These abnormalities have traditionally been considered as a physiologic mechanism to save energy in response to uremic wasting (10). However, growing evidence suggests participation of T3 in the pathophysiology of endothelial dysfunction (11), atherosclerosis (12), and cardiac abnormalities (13) of CKD patients. Some studies, with somewhat conflicting results, have also implicated thyroid alterations in the mortality risk of CKD patients (14–16). Zoccali et al. (14) were the first to report a direct association between low fT3 and all-cause mortality in stable hemodialysis (HD) patients, with similar findings in peritoneal dialysis patients (17). We could further confirm and expand these observations in a cohort of euthyroid incident HD patients, in which the variance of all-cause and cardiovascular (CV) mortality was better explained by T3 than by fT3 levels (16). In contrast, a study in 87 HD patients could not confirm these findings (15), and a retrospective report found that the association between fT3 and mortality was confounded by the nutritional status of the patients (5). Whereas existing evidence on the association between thyroid hormones and mortality is limited to studies considering single measurements of hormone levels, it is currently unknown whether thyroid hormone fluctuation over time affects mortality hazards. In apparently healthy individuals, thyroid hormone concentrations show a narrow intraindividual variation (18), although certain circadian and seasonal (mainly due to seasonal differences in iodine intake) fluctuations have been observed (19). In diseased individuals, however, thyroid hormone concentrations seem subjected to a higher fluctuation as influenced among others by wasting syndromes, persistent inflammation, and cardiac comorbidities (19). Against this background, the objective of our study was to evaluate the association between both basal levels and longitudinal (3-month) variation of T3, thyroxine (T4), and thyroid stimulating hormone (TSH) and all-cause and cause-specific mortality. We did so in a well-defined cohort of maintenance HD patients.
Materials and Methods
The Medical Ethics Committee at Karolinska University Hospital and Uppsala University approved the study protocol, and patients consented to participate. This is a post hoc analysis from a cohort of prevalent HD patients in the Stockholm region, described elsewhere in more detail (20,21). The inclusion period ranged from October 2003 until September 2004. Of 224 patients originally included in this cohort, TSH, T3, and T4 were not measured in six individuals due to limited plasma. The patients were on angiotensin-converting enzyme inhibitors and∕or angiotensin II receptor antagonists (n=72), β-blockers (n=109), calcium-channel blockers (n=57), diuretics (n=104), statins (n=72), antiplatelet drugs (n=65), erythropoiesis-stimulating agents (n=211), and iron substitution (n=144). Sixteen patients were receiving levothyroxine, and one of these additionally used amiodarone. No patients were receiving lithium or propranolol. Comorbidity was classified according to Davies et al. (22) on a 7-point scale that was simplified into a three-risk-category scale. Nutritional status was evaluated with the subjective global assessment.
From inclusion onward, patients were followed for the occurrence of fatal events. Causes of death were retrieved from death certificates and classified as CV or non-CV. CV mortality was defined as death due to myocardial ischemia or infarction, cardiac arrest or unknown sudden death, acute and chronic heart failure, cerebrovascular accidents, cerebral hemorrhage, and ruptured aortic aneurysm. Non-CV death was defined as that not attributable to a CV origin. Individuals with unknown causes of death (n=13) were grouped within the non-CV group.
Biochemical Methods
Venous blood samples were drawn and stored at −70°C if not analyzed immediately. Concentrations of high-sensitivity C-reactive protein (CRP) and serum albumin (bromcresol purple) were determined using routine methods at the Department of Laboratory Medicine, Karolinska University Hospital (Huddinge, Sweden). Serum IL-6 concentrations were quantified in an automated Immulite analyzer (Siemens Healthcare Diagnostics, Los Angeles, CA). Thyroid hormones levels were assessed on an Immulite system, using commercially available immunometric assays for T3 (analytical sensitivity [AS], 0.54 nmol/L; total coefficient of variation [CoV], 13.2% and 5.4% at the levels of 0.95 and 6.02 nmol/L), T4 (AS, 5 nmol/L; total CoV, 8.4% and 6.3% at the levels of 49 and 167 nmol/L), and TSH (AS, 0.004 mIU/L; total CoV, 12.5% and 4.6% at the levels of 0.016 and 1.3 mIU/L). Results are expressed as the average of two measurements.
Statistical Analyses
Low T3 or T4 categories were defined as those below the 66th percentile of distribution (higher tertile as the reference group). For the analysis of trimestral thyroid hormone variation, we classified patients according to their shift along tertile distribution at each time point, a strategy described in more detail elsewhere (23). From the nine possible combinations, four groups were created by clustering patients with changes in the same direction: (1) individuals who showed a change to lower tertiles were classified as the “decrease” group, (2) individuals showing an increase to upper tertiles were assigned to the “increase” group, (3) individuals with both values within the highest tertile of distribution were labeled as a “persistently high” group, and (4) individuals with both values within the lower or the middle tertile were labeled as a “persistently low” group. Differences between groups were tested by means of parametric (independent sample t test, one-way ANOVA), non-parametric (Mann–Whitney U and Kruskal–Wallis test), and chi-squared tests as appropriate. Correlation was assessed by means of Pearson or Spearman tests, as appropriate. Survival during follow-up was analyzed by the Kaplan–Meier method, and hazard ratios (HRs) were calculated with Cox proportional hazard models with different degrees of adjustment. Age, sex, comorbidities, protein-energy wasting, albumin, smoking, levothyroxine prescription, dialysis vintage, and IL-6 were considered possible confounders. T3, IL-6, and vintage on dialysis were logarithmically transformed because of non-normal distribution. In order to test whether the effect of T4 levels at baseline and variation groupings on mortality was due to the effect of T3, we performed an additional adjustment for log T3 or δ T3, respectively.
As a sensitivity analysis, Cox analyses were repeated in biochemically euthyroid patients only, defined as presenting with normal TSH (reference range 0.1–4.5 mIU/L) and T4 (reference range 57.9–169.9 nmol/L). Cox analyses were adjusted by CRP instead of IL-6 as a surrogate of inflammation. HRs with 95% confidence intervals (95% CIs) not including one were considered to be statistically significant. P values <0.05 were considered statistically significant. SPSS version 17.0 (SPSS Inc., Chicago, IL) was used to analyze the study material.
Results
Basal Thyroid Hormone Levels and Its Association with Mortality
In all 218 individuals, median (interquartile range; IQR) levels of TSH and T3 were 1.35 (0.84–2.52) mIU/L and 0.82 (0.67–1.03) nmol/L, respectively. Basal T4 levels were on average (SD) 70.44 (24.65) nmol/L and positively related to T3 levels (ρ=0.529, P<0.001). Thyroid hormone disorders on the basis of circulating hormones were common: 146 patients were classified as having a low T3 syndrome, 11 patients were classified as having overt hypothyroidism, and an additional 4 patients were classified as having subclinical hypothyroidism. General and thyroid-specific characteristics of studied patients are presented in Table 1, according to low levels of T3 (≤66th percentile; ≤0.94 nmol/L) and T4 (≤66th percentile; ≤77.2 nmol/L). Patients with low T3 presented with a higher prevalence of comorbidities, increased IL-6 levels, and lower T4 concentrations. Patients with low T4 were more often male, malnourished, and presenting with lower T3 levels. In univariate analyses, IL-6 positively associated with TSH (Spearman rank test ρ=0.16, P=0.01) and negatively with T3 (ρ = −0.17, P=0.009). CRP levels positively associated with T4 (ρ=0.19, P=0.003). During a median (IQR) follow-up period of 38.2 (17.6–45.1) months, 103 deaths occurred (40 CV and 63 non-CV) of which 77 occurred in the low T3 group and 26 in the high T3 group. As shown in Table 2, the crude and adjusted hazards of dying (by all causes) were 1.8 and 1.6 times higher, respectively, in patients with low T3 levels than in those with high T3 levels. This was mainly due to CV deaths, where HRs were considerably higher in magnitude (crude HR [95%CI]: 2.8 [1.2–6.4]) and remained so after adjustment for confounders (adjusted HR: 2.7 [1.2–6.3]). A similar analysis was performed for T4. A statistically significant association between low T4 levels and an increased all-cause and CV mortality was observed after adjustment for confounders (adjusted HR: 3.0 [1.3–6.8]). Further adjustment for log-T3 levels did not materially alter the strength of these associations, although a 16% reduction in magnitude was observed for CV mortality. Neither low T3 nor low T4 levels were significantly associated with non-CV mortality. TSH levels did not associate with any outcome measure (data not shown).
Baseline characteristics according to T3 and T4 dichotomization in 218 prevalent hemodialysis patients
Mortality HRs and 95% CIs according to basal T3 and T4 in 218 prevalent hemodialysis patients
Trimestral Thyroid Hormone Variation and Its Association with Mortality
Thyroid hormones were assessed again after 3 months in 210 patients. Median (IQR) coefficients of intrapatient variation were as follows: T3, 7.8% (2.6–12.5); T4, 7.1% (4.0–12.0); and TSH, 14.0% (5.4–24.8). In Table 3, general and thyroid-specific characteristics are depicted across four different T3 variation groups (Materials and Methods), not observing major differences among these groups. The same was true for T4 variation groups (data not shown). In univariate correlation analyses, δ TSH associated with δ IL-6 (ρ=0.17, P=0.01), whereas δ T3 associated with δ serum albumin (ρ=0.22, P=0.01). Figure 1 shows the Kaplan–Meier curves for T3 and T4 trimestral variation, both being associated with patient outcome. Table 4 shows crude and adjusted Cox models. Patients with persistently low T3 levels presented the highest hazard for all-cause mortality (HR: 2.7 [1.5–5.0]) compared with that of subjects having persistently high levels. This was particularly true for the association with CV mortality (HR: 4.0 [1.3–11.7]), whereas no association between T3 variation and non-CV mortality was observed (data not shown). Adjustment for confounders did not alter either the magnitude of the association or the statistical significance. Of note, elevated HRs, albeit statistically non-significant, were observed for patients showing increases or decreases of T3 levels during the 3-month follow-up compared with those of patients having persistently high concentrations.
General characteristics according to T3 trimestral variation in 210 prevalent hemodialysis patients
Kaplan–Meier curves for T3 and T4 trimestral variation. Kaplan–Meier survival curves for all-cause mortality according to T3 (A) and T4 (B) trimestral variation in 210 prevalent hemodialysis patients.
Mortality HRs and 95% Cis according to T3 and T4 trimestral variation in 210 prevalent hemodialysis patients
Patients with persistently low T4 levels showed elevated hazards for all-cause mortality compared with those of patients having persistently high T4 levels, an association that reached statistical significance after adjustment for confounders. Higher hazards were observed for prediction of CV mortality (crude HR: 2.6 [1.0–7.0]). In both cases, further adjustment for trimestral T3 variation (as a continuous variable) did not significantly affect the results. Of note, a decrease in T4 was also associated with elevated hazards for CV mortality. No association was observed between T4 variation patterns and non-CV mortality.
Trimestral TSH variation patterns were not associated with outcome. In a sensitivity analysis, all baseline and longitudinal Cox analyses were repeated in biochemically euthyroid patients only (n=210 at baseline, n=202 with available data at both time points). Additionally, Cox adjustment was done with CRP instead of IL-6. In both cases, results were not different (data not shown).
Discussion
In this study, we show that both basal levels and trimestral variation of T3 and T4 are associated with increased mortality, particularly due to CV causes. Adjusting in the causal pathway suggests that the mechanisms associating low T4 levels with an elevated mortality rate may be, at least in part, independent of T3 levels.
Our observation linking low basal T3 levels with increased all-cause mortality is in agreement with most (5,14,16,17) but not all (15) preceding literature. Whereas earlier studies (5,14,17) only studied the association between T3 and fT3 and all-cause mortality, our study adds that this mortality prediction is mainly attributable to CV causes. The lack of mortality prediction observed by Fernández-Reyes et al. (15) could relate to the selection of patients having survived at least 12 months, the low number of events encountered, and the strict inclusion criteria that excluded, among others, patients with abnormal fT3 and free T4 (fT4) levels. In this sense, our analysis shows that patients with both baseline and longitudinal low T3 levels have a relatively shorter time to death, possibly suggesting a stronger impact of thyroid hormones on short-term outcome. The study from Ozen et al. (5) showed that adjustment for CRP and s-albumin abrogated the association between fT3 and all-cause mortality, and the authors concluded that poor nutritional status confounded this association. The acute-phase reactant nature of s-albumin makes these results difficult to contextualize (24,25), and in our study, adjustment for inflammation (IL-6) and malnutrition did not affect this association. Furthermore, Zoccali et al. (14) reported that as adjustment for fT3 abrogated the association between inflammation and patient outcome, fT3 may be an intermediate pathologic indicator of systemic inflammation. In our data, as well as in Ozen et al. (5), adjustment for inflammation did not affect the link between T3 and mortality. Also, associations between trimestral thyroid hormone changes and variation of inflammatory markers were somewhat weak and inconsistent in our study. One limitation that makes these studies not fully comparable is that whereas Zoccali et al. (14) and Ozen et al. (5) measured free fractions, we report on total T4 and T3 levels only. Although adjustment for s-albumin (one of the main transporters of T3 in plasma) did not modify the observed associations, we should acknowledge that fT3 is not tantamount to T3. Several studies have, nevertheless, suggested anti-atherosclerotic effects of fT3 on the vascular bed via its effect on mitochondrial oxidative systems, induction of vasodilatory molecules, inhibition of angiotensin II receptor expression and inhibition of downstream signal transduction, mechanisms all of which do not necessarily involve the inflammatory cascade (26–28).
In support of the association between T3 and cardiovascular death in our study, a previous report (11) in patients with moderate to severe kidney disease reported that T3 levels inversely associated with flow-mediated vasodilation and that this observation was dependent on multivariate adjustment for asymmetric dimethylarginine levels. At a cardiac level, low T3 levels in CKD patients are associated with reduced left ventricular function, increased left ventricular mass (13), and elevated intima–media thickness (12). Our study also adds novel evidence regarding the association between trimestral T3 variation and all-cause and CV mortality in ESRD. Patients with persistently low T3 levels exhibited the highest hazards of dying. Notably, also increased variability in T3 (e.g., both increases and decreases in concentration) resulted in elevated hazards albeit statistically not significant, leading us to hypothesize that T3 fluctuations may link to adverse outcomes. This finding accords with a previous study in critically ill patients (29), in which a fast decline in T3 and T4 (without a concomitant rise in TSH levels) was observed prior to death. Our longitudinal analysis, by including two observations 3 months apart in the same individual, represents a step forward by virtue of observing the subject-specific temporal order of events. What factors drive this, and whether uremic thyroid fluctuation is higher than in other diseases remains, for now, unknown.
Another novelty in our analysis is the association between baseline and trimestral variation of T4 levels with (CV) mortality. Although this is the first study to report so in a dialysis population, the findings are in line with previous evidence in non-renal patients with non-thyroidal illness (29,30). An interesting aspect of our analysis is that the strength of the association between low T4 levels and mortality was not fully affected by adjustment for T3. This may suggest that the effect of T4 on outcome is, at least in part, not dependent on its metabolite and that both may participate in the increased CV risk. According to recent literature, however, adjusting in the causal pathway may not be so straightforward as initially thought (31), and caution is needed in the interpretation of this finding, which warrants confirmation in further studies. Nevertheless, our previous report in incident dialysis patients also observed, by means of receiving operator characteristics analysis, a significant association between T4 levels and mortality (16). The study by Takamura et al. (32) in euthyroid patients demonstrated that carotid intima–media thickness was inversely and independently associated with T4 and fT4, suggesting an increased CV risk in subjects with low T4 even within the normal reference range. Differential effects of T4 and T3 on immune cells have also been reported: T4 stimulated whereas T3 inhibited peripheral lymphocyte proliferation (33).
In the interpretation of our results, some additional limitations must be addressed: Causes of death were extracted by death records and not confirmed by autopsies, which could result in misclassification. As unknown causes of death were denoted as non-CV, this could translate, in any case, into an underestimation of the observed effect toward the null hypothesis, and the true hazards may possibly be larger. The inclusion of prevalent dialysis patients may infer into a survival bias, although preceding studies contemplate also prevalent patients in their designs. Finally, ours is an observational study, and confounding by unmeasured factors cannot be ruled out.
To conclude, patients with reduced T3 and/or T4 levels bear an increased mortality risk, especially due to CV causes. This was true when considering both baseline measurements and trimestral variation patterns. Our longitudinal design adds important observational evidence—although nondecisive—that the link may underlie a causal effect.
Disclosures
B.L. is employed by Baxter Healthcare Corporation.
Acknowledgments
We thank the patients and personnel involved in the creation of this cohort. Also, we are indebted to our research staff at KBC (Annika Nilsson, Ann-Christin Emmoth, and Ulrika Jensen) and KFC (Bjorn Anderstam, Monica Eriksson, and Ann-Christin Bragfors-Helin).
The collection of this cohort was supported by a grant from Amgen. We acknowledge support from the Loo and Hans Ostermans’ Foundation, the Westman’s Foundation, and the Swedish Research Council. Baxter Novum is the result of a grant from Baxter Healthcare Corporation to the Karolinska Institute.
Footnotes
Published online ahead of print. Publication date available at www.cjasn.org.
See related editorial, “Thyroid Function and Clinical Outcomes in Kidney Failure,” on pages 12–14.
Access to UpToDate on-line is available for additional clinical information at www.cjasn.org.
- Received May 31, 2011.
- Accepted October 3, 2011.
- Copyright © 2012 by the American Society of Nephrology
References
In this issue
Jump to section
More in this TOC Section
Original Articles
ESRD and Chronic Dialysis
Cited By...
- Low Triiodothyronine Syndrome and Long-Term Cardiovascular Outcome in Incident Peritoneal Dialysis Patients
- Low-T3 Syndrome in Peritoneal Dialysis: Metabolic Adaptation, Marker of Illness, or Mortality Mediator?
- The impact of low triiodothyronine levels on mortality is mediated by malnutrition and cardiac dysfunction in incident hemodialysis patients
- Hypothyroidism and Mortality among Dialysis Patients
- Thyroid Function and Clinical Outcomes in Kidney Failure