Background
Critical illness is hallmarked by pronounced neuroendocrine alterations, with those within the thyroid axis, the somatotropic axis and the adrenal axis most extensively studied [
1]. The typical responses follow a biphasic pattern, distinguishing between the acute and prolonged phase of critical illness [
1,
2]. In the acute phase (first hours to days), the anterior pituitary is actively secreting hormones, with a transient rise in thyroid-stimulating hormone (TSH) and an increase in growth hormone [
1‐
5]. An acute rise in adrenocorticotropic hormone (ACTH) has been described in patients with sepsis or multiple-trauma [
6], but such rise was not observed in heterogeneous general ICU patients [
7,
8]. The active pituitary hormone secretion occurs in the face of altered peripheral hormone metabolism, altered target organ sensitivity, and altered hormone binding proteins [
4‐
13]. These alterations reduce the availability of most anabolic effector hormones, including triiodothyronine (T
3) and insulin-like growth factor-I (IGF-I), while the availability of the catabolic stress hormone cortisol increases. When illness is prolonged beyond the first few days, the neuroendocrine axes are uniformly suppressed, with low target organ hormone levels or, in case of cortisol, insufficiently elevated or normal levels [
1,
2,
5,
7,
10,
14]. This suppression during the prolonged phase of illness is of central/hypothalamic origin [
1,
2,
15].
Whereas the neuroendocrine responses in ICU have been well documented, data after ICU discharge are scarce and mostly limited to patients who suffered from brain damage either due to traumatic brain injury (TBI) or brain surgery. Although TBI-associated neuroendocrine disturbances often resolve, persistent hypopituitarism remains present in many patients up to years after the insult, associated with poor recovery and worse long-term outcome (e.g. cognitive impairment, decreased exercise capacity, poor quality-of-life) [
16‐
18]. Likewise, survivors of brain tumors show a high risk of hypopituitarism and need for hormone replacement therapy many years later [
19,
20]. Heterogeneous prolonged critically ill adult patients showed a uniform rise in ACTH and cortisol to supra-normal levels from ICU discharge to one week later [
14]. In children, salivary cortisol levels were normal months to years after ICU admission [
21,
22]. In the absence of other data, it remained unclear whether the neuroendocrine abnormalities that are present in general adult ICU patients recover in the long-term.
In this study of patients who were followed-up 5 years after ICU admission for heterogeneous diagnoses, we compared hormonal parameters within the thyroid axis, the somatotropic axis and the adrenal axis with those of demographically matched controls, and investigated whether any long-term neuroendocrine abnormalities in former ICU patients associate with long-term physical functional impairments.
Discussion
Patients admitted to the ICU develop typical neuroendocrine changes within the thyroid axis, the growth hormone axis and the adrenal axis in response to critical illness, with pronounced disturbances remaining present until the day of ICU discharge. In this study, we demonstrated that 5 years after critical illness, most of these neuroendocrine abnormalities had normalized, with the exception of rT3 concentrations that remained supranormal, resulting in persistently low T3/rT3 ratios, and of IGFBP3 concentrations that rose from subnormal levels in the ICU to supranormal levels at 5-year follow-up. The lower T3/rT3 ratios observed in former ICU patients were independently associated with several measures of worse long-term physical capacity, whereas the higher IGFBP3 concentrations were independently associated with better physical capacity (handgrip strength).
Studies investigating neuroendocrine function after recovery from critical illness outside the setting of TBI or brain surgery are scarce if not absent. It has been shown that former ICU patients who were transferred to long-term care facilities still reveal the typical non-thyroidal illness syndrome, which is not unexpected given that such patients have not fully recovered [
29]. However, information on the thyroid axis in fully recovered former ICU patients was hitherto lacking. We here observed normal serum TSH, T
4, T
3, and TBG concentrations 5 years after ICU admission, whereas rT
3 remained elevated, resulting in an abnormally low T
3/rT
3 ratio. The underlying mechanism of this long-term abnormality within the thyroid axis remains unclear, but could involve persisting changes in the expression or activity of the deiodinases that control the conversion of T
4 to T
3 and rT
3. We also demonstrated that in the former ICU patients assessed 5 years later, serum growth hormone, IGF-I and IGFBP1 had normalized, whereas IGFBP3 was increased reaching supranormal levels. Increases in IGF-I and IGFBP3 (and T
4) have also been documented over a 2-year time-period in children after severe burn injury, after they had acutely dropped in the critical phase of the injury, but comparison with healthy children was not performed [
30]. IGFBP3 is a binding protein that positively regulates tissue availability of IGF-I [
31]. Its expression is regulated by growth hormone, insulin, androgens, and vitamin D, among others, with involvement of DNA methylation and transcriptional, posttranscriptional and translational control [
32]. Whether alterations in these regulators play a role in the supranormal IGFBP3 levels of former ICU patients remains unknown. Although we have previously shown that 1 week after ICU discharge of prolonged critically ill adult patients ACTH and cortisol levels had risen to supranormal levels [
14], the current data suggest that the adrenal axis recovers thereafter. This finding is consistent with normal salivary cortisol levels observed in former critically ill children, assessed months to years after pediatric-ICU admission [
21,
22]. Interestingly, however, in acute-respiratory-distress-syndrome survivors, an inverse correlation has been reported between long-term basal serum cortisol level and increasing traumatic ICU memories [
33]. Unlike in human patients, long-term perturbations of the adrenal axis have been observed in rodent models of sepsis. In mice, increased stress-induced corticosterone and increased adrenal weights have been observed 2 to 7 weeks after induction of sepsis via cecal ligation and puncture [
34]. In rats injected with lipopolysaccharide, desensitization of the adrenal axis has been described weeks later [
35].
Observing the residual long-term neuroendocrine changes evidently raises questions about their physiological relevance. Reverse T
3 has long been considered an inactive metabolite of thyroid hormone, considering its weak affinity for nuclear thyroid hormone receptors [
36]. Recently, however
, in vitro studies have suggested an active role for rT
3 through interaction with extranuclear receptors, though physiological relevance hereof remains to be established [
36]. In ICU, higher rT
3 and lower T
3/rT
3 ratios have been associated with worse outcome of critically ill patients [
37,
38]. In a rat stroke model, however, a protective effect of rT
3 has been suggested [
39]. Altered IGFBP3 levels may affect IGF-I transport, bioavailability and activity [
31]. However, also pleiotropic IGF-independent actions of IGFBP3 have been described, regulating gene transcription with effects on cell growth, survival and apoptosis [
31]. To explore any potential physiological relevance of the documented long-term neuroendocrine abnormalities in former ICU patients, we studied associations with long-term physical function. Interestingly, we observed an independent association of a persisting low T
3/rT
3 ratio with decreased physical performance, including lower handgrip strength and shorter 6-min walk distance. For handgrip strength, the effect size appeared clinically relevant, considering 5 kg as minimal clinically important difference [
40,
41] was reached with a two-standard-deviations change in the T
3/rT
3 ratio. The effect size for the 6-min walk distance was far below the minimal clinically important difference of 14–30 m [
42]. Nevertheless, the combination of all data suggests that the residual abnormality within the thyroid axis may confer an increased risk of long-term physical impairment, and would thus be a harmful long-term consequence of critical illness, at least for a subgroup of patients. Such an interpretation is plausible as thyroid hormone affects diverse aspects of skeletal muscle physiology, being key in the regulation of the muscle’s contractile function, energy metabolism, myogenesis and muscle regeneration [
43]. Thus, thyroid hormone action plays an important role in the maintenance of muscle strength and physical functioning. Outside the context of critical illness, patients with newly diagnosed thyroid disease complain about weakness, fatiguability, muscle pain, stiffness and cramps, which usually resolve after treating the thyroid disease [
44]. In middle-aged and older euthyroid subjects, a higher free T
3 level has been independently associated with a higher handgrip strength and physical function, and with an attenuated decline in handgrip strength over time, whereas no association was found for free T
4 or TSH [
45,
46]. Another study evaluating only TSH and free T
4 found an independent association of low-normal TSH with lower handgrip strength in elderly euthyroid men, but not in post-menopausal women [
47]. In young, euthyroid men, rT
3 has been inversely associated with lean body mass, as have thyroid hormones [
48]. In independently living elderly men, high supranormal rT
3 levels, but also higher free T
4 levels within the normal range, were independently associated with worse physical performance and lower muscle strength (handgrip, leg extensor), whereas an isolated low T
3 level remarkably was associated with better physical performance [
49]. For IGFBP3, association with long-term physical outcome of former ICU patients was less clear than for T
3/rT
3, as such association was only found for handgrip strength. Considering a higher IGFBP3 was associated with better handgrip strength, the supranormal levels of IGFBP3 years after critical illness could be interpreted as a beneficial, compensatory response and thus would not explain long-term physical impairment after critical illness. Positive associations of IGFBP3 with physical outcome in aging have previously been documented for activities of daily living (ADL) in women but not men, for handgrip strength in a cohort of 89-year-old women, and for get-up-and-go times in a mixed gender historical cohort [
50,
51]. Most studies in middle-aged to elderly people, however, failed to independently associate IGFBP3 with functional performance measures such as walking speed, grip-strength, or ADL [
50,
52‐
55].
The identification of long-term abnormalities in the thyroid axis as a potential contributor to the long-term physical legacy after critical illness is important, as pathophysiological insight in the long-term physical impairments in ICU survivors is scarce [
56‐
58]. Many prolonged critically ill patients who developed critical illness polyneuropathy showed signs of chronic partial denervation up to 5 years after ICU discharge, whereas persisting evidence of myopathy in patients who developed critical illness myopathy appeared unusual [
59,
60]. One small study in prolonged critically ill patients with persistent weakness as assessed 6 months after ICU discharge suggested normalization of proteolysis, autophagy, inflammation and mitochondrial content in muscle, but persistence of impaired regenerative capacity [
61,
62]. Studies in mice suggested involvement of sustained mitochondrial dysfunction in chronic sepsis-induced muscle weakness [
63], and also showed that engraftment of mesenchymal stem cells improved muscle regeneration and strength after sepsis [
64].
This study has limitations to consider. First, we analyzed single samples, whereas several hormones show pulsatile patterns [
1,
2]. Second, no direct measurements of free, bioavailable target hormones were performed. Indeed, as the use of heparinized lines in-ICU interferes with free thyroid hormone measurements [
65] we also did not measure free T
4 and T
3 in the follow-up samples, and the complex, time-consuming methodology to measure bioavailable IGF-I [
66] and free cortisol [
13] does not allow analysis of such a large number of samples. Third, we have no information on ACTH concentrations as blood samples were not immediately stored on ice after collection, which precludes reliable ACTH measurements. Fourth, tissue hormone concentrations or metabolizing enzymes could not be evaluated. Fifth, our search for independent determinants of hormonal parameter concentrations at follow-up was only of exploratory nature and should not be overinterpreted. Of importance here, no information was available about the participants’ chronic nutritional status, whereas this could affect hormone concentrations as well [
5,
38]. Finally, the studied patient cohort may be prone to selection bias. Non-survivors obviously could not be studied, whereas they are generally more severely ill, have a more complicated ICU trajectory, and overall show worse neuroendocrine disturbances in ICU than survivors [
37,
67‐
70]. However, the three studied neuroendocrine axes also showed severe disturbances in the present cohort of survivors while in the ICU, representative of the impact of critical illness. The studied cohort was relatively enriched in sicker, long-stay patients as compared with the total EPaNIC cohort. Nevertheless, exclusion of patients with disabilities potentially confounding morbidity endpoints in the follow-up study, as well as availability of blood samples only from former ICU patients who were able to come to the hospital for participation in the study, may have introduced bias toward those with better physical capacity. Also, only a small number of patients were studied at the intermediate time points, which may have biased findings. That is why we only included these data for visual purposes without statistical analyses.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.