Hypothyroidism in hospitalized elderly patients: a sign of worse prognosis

Overt hypothyroidism has adverse clinical consequences and might worsen prognosis in critically ill elderly patients. However, the difficult interpretation of thyroid function tests (TFT) due to non-thyroidal illness (NTI) has led to discouragement of screening for thyroid dysfunction. Our aim was to determine the prevalence of TFT compatible with hypothyroidism and to study its influence on mortality among hospitalized elderly patients. In this prospective study we consecutively included all patients ≥60 years admitted by the Internal Medicine Department to the hospital ward (n = 451) of the Cesar Milstein Hospital in Buenos Aires, Argentina. TFT were done on day 1 and 8. Thyroid function categories were defined as overt and subclinical hypothyroidism, overt and subclinical hyperthyroidism, euthyroidism and NTI. Stage of chronic kidney disease (CKD), Adult Comorbidity Evaluation (ACE)-27, and intra-hospital mortality were recorded. The association between mortality and TFT categories was studied by Cox regression. Out of 451 patients (77.0 ± 7.9 years, 54% females) 76% were categorized as NTI, 4% as overt hypothyroid, 10% as subclinical hypothyroid, 1% as subclinical hyperthyroid and 9% as euthyroid. Overt hypothyroid patients showed significantly higher mortality than the rest of the groups (25%, p < 0.05) while ACE-27 was similar among all of them (p = 0.658). In addition, patients within the overt hypothyroid category showed a higher mortality rate than NTI in a model adjusted by Stage 5-CKD, ACE-27, sex and age [HR 3.1 (1.14–8.41), p < 0.026]. Overt hypothyroidism during hospitalization was associated with elevated mortality. Further studies would reveal if TFT alterations compatible with hypothyroidism should be diagnosed/treated in hospitalized elderly patients.


Introduction
According to population studies, the values of TSH slightly increase as age advances [1,2].Nevertheless, this has not been associated to a deficit of thyroid hormones but to a possible adaptation of the thyroid axis in the elderly.Major controversies arose from several reports of lack of adverse effects [3][4][5] and prolonged life span [6,7] in old age individuals with abnormally high levels of TSH.Such results have led to important changes in the treatment recommendations for these patients.Most of the treatment guidelines [8][9][10] recommend against treating this special segment of the population when TSH values are below 10 mU/L and

Abstract
Purpose Overt hypothyroidism has adverse clinical consequences and might worsen prognosis in critically ill elderly patients.However, the difficult interpretation of thyroid function tests (TFT) due to non-thyroidal illness (NTI) has led to discouragement of screening for thyroid dysfunction.Our aim was to determine the prevalence of TFT compatible with hypothyroidism and to study its influence on mortality among hospitalized elderly patients.Methods In this prospective study we consecutively included all patients ≥60 years admitted by the Internal Medicine Department to the hospital ward (n = 451) of the Cesar Milstein Hospital in Buenos Aires, Argentina.TFT were done on day 1 and 8. Thyroid function categories were defined as overt and subclinical hypothyroidism, overt and subclinical hyperthyroidism, euthyroidism and NTI.Stage of chronic kidney disease (CKD), Adult Comorbidity Evaluation (ACE)-27, and intra-hospital mortality were Electronic supplementary material The online version of this article (doi:10.1007/s40618-017-0690-2) contains supplementary material, which is available to authorized users.free T4 is within the reference range.Such recommendation is issued despite the lack of randomized controlled trials assessing the cost-effectiveness of levothyroxine replacement in elderly population.
On the other hand, it is widely recognized that with age the prevalence of thyroid autoimmunity raises [11], with the consequent increase in hypothyroidism and a detrimental impact in the cardiovascular and neurocognitive areas [8].Considering the influence of medications and comorbidities in this period of life together with the above explained changes in the thyroid profile, it is quite understandable why it is so difficult to interpret thyroid function tests (TFT) and to identify true thyroid pathology in the elderly.
In elderly hospitalized patients, the scenario is even worse due to the use of drugs that interfere on the measurement of TFT and, most importantly, to the presence of the non-thyroidal illness (NTI) [12,13].In particular, the use of contrast solutions, amiodarone, heparin, corticoids and of other medications that interfere with the absorption of levothyroxine, all hamper the evaluation of thyroid function in hospitalized patients [14] Therefore, the diagnosis of hypo or hyperthyroidism in this particular population is a difficult task; thus, the screening for thyroid pathology is discouraged in these patients [15].Nevertheless, this strategy underestimates the potential detrimental effect of a true state of hyper or hypothyroidism on the evolution of hospitalized patients.
There are scarce data on the possible effect of hypothyroidism on the evolution of elderly in-patients.A former retrospective study has shown that higher baseline TSH values during hospital admission were related to a worse long-term prognosis in older patients [16].In addition, Maldonado et al. [17] reported in a series of 116 critically ill patients that high TSH was independently associated with mortality.These findings challenge the notion of a possible protective effect of hypothyroidism in this period of life [6,7] and also contrast with other studies that did not report excess mortality in hypothyroid patients [15,18,19].
Our aim was to evaluate the prevalence of TFT compatible with hypothyroidism and to study its influence on mortality in a population of hospitalized elderly patients.

Population
The study was conducted in a hospital for the elderly in Buenos Aires during a 12-month period, from January 2013 to January 2014.Eligible subjects were 504 patients who were consecutively admitted to the clinical ward of our institution by the Internal Medicine Department.Exclusion criteria were: age <60 years, myxedema coma, thyroid cancer, incomplete clinical or biochemical data and refusal to participate in the study.The study population comprised 451 patients.The study protocol was approved by the Ethics Committee of the Cesar Milstein Hospital.

Informed consent
"Informed consent was obtained from all individual participants included in the study."

Clinical evaluation
The main cause of admission and the medications taken before admission were registered.At day 1 of the hospital stay, the usage of drugs that may interfere with TFT was registered.Comorbidities at day 1 were assessed by the Adult Comorbidity Evaluation (ACE-27) [20].Geriatric assessment was obtained with the Karnofsky performance scale (KPS) that allows patients to be classified according to their functional impairment [21].During the hospital stay, changes in medications, intensive care unit (ICU) requirement, hospital discharge, and intra-hospital mortality were recorded.
Overt hyperthyroidism was diagnosed when TT4 was over 13 µg/dL and/or T3 over 190 ng/dL and TSH below 0.3 mU/L.Subclinical hyperthyroidism was diagnosed when TT4 (4.5-13 µg/dL) and T3 (80-190 ng/dL) were normal and TSH below 0.3 mU/L.Euthyroidism was diagnosed when all hormones were within their normal range: TSH 0.3-5 mU/L, TT4 4.5-13 µg/dL, and T3 80-190 ng/dL.Overt hypothyroidism was diagnosed when TT4 was below 4.5 µg/ dL and TSH over 5 mU/L.Subclinical hypothyroidism was diagnosed with normal levels of TT4 4.5-13 µg/dL and T3 80-190 ng/dL with TSH over 5 mU/L.NTI was diagnosed when T3 levels were below 80 ng/dL and the rest of the hormones within their normal range or in any case where the combination of hormones would not fit in any of the previously defined categories.
General biochemical parameters were evaluated by standardized methods.The stage of chronic kidney disease (CKD) of each patient was evaluated using age and serum creatinine to calculate the glomerular filtration rate (CKD-EPI formula).CKD stages were defined according to international guidelines.

Clinical outcomes
The primary outcome was defined as intra-hospital mortality.

Statistical analysis
Data were expressed as mean and standard deviation or median and interquartile range (IQR) as appropriate according to data distribution.ANOVA or the Kruskall-Wallis test was used to compare data between the groups.To compare categorical variables between the groups a Z-test with the Bonferroni correction for multiple comparisons was employed.Associations among TFT categories and mortality were evaluated by COX regression.All variables included in the model were graphically tested to meet the proportional hazard assumption.The group of patients with subclinical hyperthyroidism was excluded from COX regression analysis due to its low number of patients.Similarly, euthyroid patients were excluded from COX analysis because no deaths were registered in this group.Potential confounders/predictors that were included in the COX model were: sex, age (divided by tertiles), ACE-27 score, CKD stages and TFT categories.P < 0.05 was used to consider statistical significance.SPSS ® 17.0 software (Chicago, Ill) was used in all the statistical analyses.
Clinical characteristics and outcomes according to TFT categories are shown in Table 1.Patients classified as overt hypothyroidism were older and presented higher ICU requirement and mortality than the other TFT categories.The prevalence of history of hypothyroidism and levothyroxine use was higher in the patients categorized as subclinical hypothyroid.Most likely, these results suggest sub medication as a leading cause of high TSH values among patients within this group (Table 1).TSH and thyroid hormones results and the proportion of positive TPOab patients according to TFT categories are shown in Supplementary Table 1.Significantly higher prevalence of positive TPOab was observed in the subclinical hypothyroid group.The rest of significant differences among TFT categories were in line with the selection criteria employed.
Causes of admission according to TFT categories are shown in Table 2.The leading cause of hospitalization was infectious disease (22%), followed by respiratory (19%) and hematological disease (18%).Differences in the causes of admission among TFT categories were only significant for hematological and tumoral diseases (Table 2).Overt hypothyroid and euthyroid groups showed a larger number of patients with hematological diseases than the NTI category.In addition, in the subclinical hypothyroid group a higher number of patients were admitted by tumoral diseases in comparison with the NTI and euthyroid groups.
In regards to geriatric assessment no differences in functional status were found among the groups with the KPS.The possible presence of cognitive impairment and consequent potential onset of delirium (a well-known risk factor for short term mortality), was assessed as part of the ACE score and no differences were observed when hypothyroid patients were compared with the rest of the groups.
We also compared the number of patients in each CKD stage among the groups and found that euthyroid patients presented a lower number of patients at stage 5-CKD (Table 1).

Changes in categories in day 8
Out of 451 patients, 194 remained at the hospital and had their TFT reevaluated at day 8. Figure 1 shows TFT categories at day 1 and day 8 according to baseline TSH values.Those patients whose baseline TSH level was 0.3-5.0mU/l as well as those classified as NTI and Author's personal copy 1 3 subclinical hypothyroidism in day 1 were the ones in which most of the TFT category switches took place.It was remarkable that most of the patients initially classified as overt hypothyroid remained in the same TFT category by day 8.

Mortality according to TFT groups
Mortality was compared among the NTI, overt hypothyroid and subclinical hypothyroid categories.None of the euthyroid patients had died during their hospital stay; thus, this group was not included in the analysis.1 3 Regarding the potential interference of severe pathological conditions, we accounted for the presence of important clinical parameters such as albuminemia, natremia, glomerular filtration rate.
As euthyroid patients presented a lower number of patients at stage 5-CKD, the stages of CKD were included in the survival analysis.However, CKD-stage as an ordinal variable (HR 1.14, 95% CI 0.88-1.48,p = 0.314) and stage 5-CKD as a dichotomic variable (yes/no) (HR 2.5, 95% CI 0.995-6.27,p = 0.051), were not significantly associated with mortality either.
Patients classified as overt hypothyroid presented significantly higher mortality hazard than NTI patients (Fig. 2 and Supplementary Table 2).On the other hand, no significant differences were observed between NTI and subclinical hypothyroid patients.Such result remained statistically significant even when the model was adjusted by age, sex, ACE-27 score and Stage 5-CKD (Fig. 2).

Discussion
Our study showed that TFT compatible with hypothyroidism is a frequent finding in the elderly in-patient population and that it can go unnoticed during the hospital stay.This finding is clinically relevant as patients classified as hypothyroid showed an increased mortality in comparison to patients with other thyroid function alterations, even when adjusted by age, sex, stage 5-CKD and a comorbidity score.
In regards to hospitalized patients, the role of hypothyroidism has recently become a controversial issue.Indeed, TF testing is discouraged unless there is a clear suspicion of thyroid dysfunction [8,9].These recommendations are based on the interference created by drugs and the presence of NTI that hamper the interpretation of TFT in this setting [15].Although this strategy can be cost effective, it is at the expense of under-diagnosing true thyroid dysfunction.In fact, in a retrospective study it was observed that out of 56 adult patients with newly diagnosed primary overt hypothyroidism admitted from the Emergency Department, only in 12 (21%) there was a correct initial impression of primary overt hypothyroidism [22].
A few studies that have explored TFT in admitted elderly patients confirmed that NTI is the most frequent cause of TFT abnormalities.Iglesias et al. [18] found about three quarters of patients with alterations in TFT and the reduction of FT3 values was a powerful predictor for mortality during hospitalization in their cohort of elderly patients.In another study that excluded patients admitted to ICU, NTI prevalence was only 31.9% and in this group mortality rate was also significantly higher [19].Similarly, De Alfieri et al. [23] included 450 very old (mean age: 84 years) hospitalized patients admitted for an acute disease in their study, and reported that low T3 syndrome was associated with an excess mortality risk during the hospital stay.Although all these studies are in coincidence with the ample evidence about low T3 and T4 values as predictors of poor survival in critically ill patients [16,24], the clinical significance of impaired TSH values at admission still remains undetermined.In the present study, in line with Iglesias et al. [25], the large majority of the population had NTI.However, patients with overt hypothyroidism presented an even worse prognosis during hospitalization than NTI.A possible interpretation to these data is that we excluded ICU patients, therefore, sicker patients with low T3 values.Moreover, our group of overt hypothyroid patients was larger in comparison with that found in the study of Iglesias et al. [25].In the study by De Alfieri et al. [23], hypothyroidism did not reveal any difference in survival compared to the rest of the population.However, these authors defined hypothyroidism as patients taking replacement therapy, and included patients with subclinical hypothyroidism apart from purely overt hypothyroid patients as in the present study.
With regards to the rest of the studies that have explored the effects of TSH alterations as predictors of survival in hospitalized patients, an earlier one [26] performed in 200 patients admitted to the ICU showed that subnormal TSH values (<0.4 mU/L) were associated to much higher mortality rates than patients with TSH serum levels >5 mU/L.These results differ to the present findings since every patient with high TSH in the study of Rothwell et al. [26] had T4 values within the normal range and most of these patients returned to normal TSH values on repeated measurements.Although doubts had been casted on the diagnostic value of TSH levels <10 mU/L [15], we observed that most of the patients initially classified as hypothyroid still remained in the same category by day 8 when the TFT were reassessed.Therefore, it can be possible that elderly in-patients with a minor increase of TSH levels (TSH: 5-10 mU/L) might present a true hypothyroidism in this particular setting.
On the other hand, in another study where only ICU patients were included a high mortality rate (55%) was found among the patients who were retrospectively found to be hypothyroid [17].Furthermore, in the mentioned study, higher TSH values were still associated to mortality after excluding the group of hypothyroid patients.In line with these data, we previously found in a retrospective cohort of elderly admitted patients, that hypothyroidism was associated to increased morbi-mortality.Nonetheless, this study was limited to a population of elderly patients with a clinical suspicion of thyroid dysfunction [16].In a recent study of 1662 hospitalized patients with type 2 diabetes, hypothyroidism was related to increased cerebrovascular disease compared with euthyroidism after adjustment for age and gender [27].In further agreement with our findings, Laulund et al. [28] also reported excess mortality for hospitalized patients with overt hypothyroidism.However, this association disappeared when patients with no registered comorbidities were only included for the analysis reflecting that lack of thyroid hormones might have a detrimental influence particularly in those patients who are critically ill.
Another interesting finding of our study is related to subclinical hypothyroid patients.Although all had higher than normal TSH values, 32% of these patients were on levothyroxine treatment at admission in coincidence with history of previous hypothyroidism in almost half of this group.During their stay in the hospital some of these patients developed overt hypothyroidism and it was also observed that 40% of patients who were taking levothyroxine on day 1 were not taking it on day 8.These findings are in line with those of Vaisman et al. [29], showing that a significant number of hypothyroid patients taking thyroid hormones are not in the therapeutic range and with those of Imberti et al. [30] who reported a series of cases on hypothyroid in-patients whose thyroid status worsened due to unintended discontinuation of levothyroxine or malabsorption secondary to other drugs frequently administered during hospitalization.
Spencer et al. [15] found the TSH concentration to be above 20 mU/L in only 1.6% of unselected in-patients.During follow-up they showed that in only 50% of these cases, thyroid disease could be diagnosed.Since overt hypothyroid patients in our study did not exhibit very high TSH values despite low T3 and T4 values, miscategorization of NTI as hypothyroidism cannot be discarded.It could be speculated that this group of hypothyroid patients may also be under the effects of NTI, thus explaining the low levels of T4 and T3 with TSH levels that do not escalate accordingly.However, it has to be acknowledged that the effect of glucocorticoids or dopamine, drugs often used during hospitalization are a possible explanation to this discrepancy between thyroid hormones and TSH [14].Another plausible explanation rely in an age-effect as it was reported that hypothyroid elderly subjects show lower TSH values than younger ones at similar T4 values [31].In addition, it could be possible that during the study our hypothyroid patients presented a diminished TSH secretion during acute stress as it has been previously reported [32].
In the present cohort, from the 13 patients classified as hypothyroid at day 8, eight patients were already identified as hypothyroid in day 1.Such result supports that thyroid dysfunction found early during hospitalization in elderly patients may be persistent and of clinical relevance.In this regard, a limitation of our study is the lack of intervention to observe possible effects of levothyroxine treatment in the patients found to be hypothyroid.However, this was an observational study and the results of TFT were blinded until the end of the recruitment phase.

Author's personal copy
It has been previously recommended against screening for thyroid dysfunction in hospitalized patients since NTI, although associated to higher mortality, is usually not worth of treatment.Moreover, hypo or hyperthyroidism are difficult to identify in this context.However, according to our findings we would suggest to reconsider screening the elderly population in whom hypothyroidism was frequently found and was also associated to a worse prognosis.Although it seems reasonable that those hypothyroid patients identified through screening deserve levothyroxine treatment, we still need a RCT proving the efficacy of levothyroxine replacement in hospitalized elderly patients.
In conclusion, we found a large proportion of overt hypothyroid in-patients with a worse prognosis when compared to the rest of the population admitted to our clinical ward.These findings warn about the misdiagnosis of potentially treatable cases of thyroid dysfunction and also show that hypothyroidism may serve as a marker of bad prognosis among hospitalized elderly patients.

1 2 3
Your article is protected by copyright and all rights are held exclusively by Italian Society of Endocrinology (SIE).This e-offprint is for personal use only and shall not be selfarchived in electronic repositories.If you wish to self-archive your article, please use the accepted manuscript version for posting on your own website.You may further deposit the accepted manuscript version in any repository, provided it is only made publicly available 12 months after official publication or later and provided acknowledgement is given to the original source of publication and a link is inserted to the published article on Springer's website.The link must be accompanied by the following text: "The final publication is available at link.springer.com".

Fig. 1 Fig. 2
Fig. 1 Sequence of patients in each TFT category from day 1 to day 8 according to TSH values in day 1 (N = 194).NTI non-thyroidal illness, oh overt hypothyroidism, sch subclinical hypothyroidism, SCH subclinical hyperthyroidism, eu euthyroid