Low body temperature and mortality in critically ill patients with coronary heart disease: a retrospective analysis from MIMIC-IV database

Even though significant advancements in care have rendered declined life lost, coronary heart disease (CHD) continues to be a leading cause of death and premature death, resulting in great healthcare expenditure [1]. Critically ill patients in the intensive care unit (ICU) often have complex underlying causes, with cardiac surgery for CHD being the primary reason for ICU admission in over 45% of cases. Furthermore, the mortality rate for patients experiencing exacerbation of chronic cardiovascular disease within 1 year after admission was found to be 16.1%, making it the second highest cause of death following malignant tumors [2]. Therefore, the role of predictive factors, biomarkers, and scores in the development and prognosis of CHD has been widely acknowledged [3‐7]. Although these factors are of good efficiency in predicting, some of them are too complicated to be routinely used and postponed awaiting laboratory tests.

Body temperature, a crucial physiological parameter frequently measured in ICU, affects inflammation and immune function, related to various diagnoses caused by infectious or not [8, 9]. In the operating room setting, rectal, bladder, esophageal, and nasopharyngeal probes are preferred for monitoring core temperature, while the invasive measurement of temperature in the pulmonary artery, considered the gold standard, is rarely employed. The skin temperature detected by infrared thermometers, which is highly susceptible to changes in ambient air temperature, thereby is not commonly used in ICU [10]. Compared with that, alternative peripheral temperature measurements, such as oral temperature, demonstrate comparable precision to the nasopharyngeal one (P = 1.00) with better acceptance in ICU [11]. An evidence-based guideline [12] also suggested that the study involving patients with CHD measuring peripheral temperature to predict morbid cardiac events by multivariate analysis was of good quality. In addition to that, another guideline [13] from The American Society of PeriAnesthesia Nurses (ASPAN) showed strong evidence for oral temperature measurements, with a recommendation class of Class I, Level B.

Whether temperature abnormalities have an influence on CHD patients remains unknown yet, so we hypothesized low body temperature was linked with worse outcome based on similar investigations aiming at patients undergoing coronary artery bypass grafting (CABG) [14, 15] and other cardiac surgeries [16]. And we tested this in the Medical Information Mart for Intensive Care (MIMIC)-IV database in a pre-specified manner.

After screening of critically ill patients diagnosed as CHD with complete records of body temperature during ICU stay and other admission data, 8577 patients were eventually enrolled in the present analysis cohort with the mean age of 69 years. Of these, 65% were men. Enrolled patients were distributed into three groups (℃): hypothermia (31.8–36.5), normothermia (36.6–37.4), and hyperthermia (37.5–39.6). The distribution of temperature was illustrated (Additional File 1: Figure S1). The in-hospital, 28-day, 90-day, 180-day and 1-year overall mortality rate were 10.9%, 16.7%, 21.5% and 30.4%, respectively.

To our knowledge, this is the first study to report the association of low body temperature and worse mortality in critically ill patients with coronary heart disease. In this retrospective cohort study, we analyzed 8577 patients and divided them into three groups: hypothermia (31.8–36.5), normothermia (36.6–37.4), and hyperthermia (37.5–39.6). The methods employed in this study to presage the disease outcomes are based on previous articles [23, 24]. Observing that the hypothermia group had a higher mortality risk over two other groups, we further perceived body temperature as a continuous variable and found a negative relationship between temperatures and worse outcomes. The predictive power of the full model has been examined based on the DCA model, and it was found that body temperature could serve as a convenient outcome predictor for critically ill patients.

Regarding of the intergroup heterogeneity showcased in baseline characteristics table which may cause confounding bias to KM curves, we further adjusted that through PSM. Though post-PSM results correct the probable fallacious outcome in hyperthermia group to some extent (the protective effect of hyperthermia for patients admitted to ICU had been described in former paper [25]), they did not present a statistically higher mortality in hypothermia group.

Intriguingly, in subgroup analysis, we found interactions between body temperature and age, heart arrest or cardiogenic shock. However, only those with cardiogenic shock had both statistically significant simple effect and statistically significant interaction effect in terms of 90-day mortality. Thus, higher proportion of patients with cardiogenic shock history was one of the crucial reasons of why hypothermia group experienced increased mortality especially in long term (the same result illustrated in the adjusted KM curves).

Studies have demonstrated that hypothermia happened due to the changes of thermoregulation system under anesthesia (including abolished behavioral responses, compromised homeostasis and reduced thresholds of vasoconstriction and shivering) [26]. Apart from this, refrigerated liquid drugs used and excessive blood loss in operations and poor physical quality at an older age can lead to hypothermia as well [27]. Systemic hemodynamic depression realized by catecholamine usage in ICU worked together with the thermoregulatory system to cause a lower temperature [28].

It has been investigated that body temperature at admission was related to the outcome and hypothermia appeared to be a significant and independent indicator of increased mortality rates both during ICU stays and over the long term [28]. In addition to that, numerous previous clinical studies have been carried on to research into the association of low body temperature with mortality and morbidity of cardiac patients. According to DeFoe et al. [14], regardless of any sites of core body temperature (nasopharyngeal, esophageal, bladder or rectal), patients undergoing isolated on-pump coronary artery bypass grafting (CABG) surgery have consistently higher in-hospital mortality rates with increasing colder temperatures; moreover, lower temperature groups were found to exhibit greater myocardial injury as assessed by myocardial markers. Likewise, Nam et al. [20] reported that the all-cause mortality of moderate-to-severe hypothermia was more than two times of that in normothermia for off-pump CABG patients and even mild hypothermia (no less than 35.5℃) was found an unsatisfied outcome during the follow-up of 47 months. However, another study that enrolled isolated off-pump CABG patients showed no statistically difference for in-hospital mortality between hypothermia group and normothermia group neither before nor after propensity score matching, while a distinction in postoperative transfusion of red cell concentrates, duration of intubation and ICU stay [15]. Unexpectedly, there was a higher rate of in-hospital mortality in the normothermia group than in the hypothermia group after pairing, though this difference was not statistically significant (P = 0.975 vs. P = 0.244). Extending to all sorts of cardiac surgeries, the results of multivariable regression analysis showed higher mortality in hypothermia group (body temperature < 36 ℃) during the 1-year follow-up [16]. To figure out the reason for unstable results of in-hospital mortality, Karalapillai et al. [29] differentiated hypothermia into transient type and persistent type in the multi-center observational study, finding that in-hospital mortality was statistically associated with persistent hypothermia but not transient hypothermia. This may shed light on variations of results through different proportions of persistent hypothermia in those studies. Moreover, low body temperature was recognized as an independent marker of poor cardiovascular mortality and rehospitalization in patients admitted with worsening heart failure and reduced ejection fraction [30].

Therapeutic hypothermia triggered by all sorts of cooling strategies, lowering the temperature between 32 ℃ and 35 ℃ for at least 24 h, has been shown to be increasingly used in the post-resuscitation care for post-cardiac arrest patients [31]. Nevertheless, studies showed that no improvement in mortality or neurologic outcome was discovered in the therapeutic hypothermia group over normothermia group [32] and even higher mortality was noticed in patients without cardiac arrest [31]. This reflects distinct causes of low body temperature (no matter for anesthesia-induced or cooling strategy-induced) may have the same tendency on worse outcomes. Similarly, lower ambient temperature, including during cold spells, could elevate mortality and morbidity of cardiovascular disease [33].

Current data about associations between body temperature and patients with coronary heart disease, especially critically ill ones, are limited. In this specific cohort of ICU patients with CHD, we found that low body temperature was an efficient and convenient independent predictor of greater mortality in these patients.

However, this study has several limitations. First, selection bias cannot be evitable due to its retrospective study nature. Since this is a single-center study for confined regions and populations, external validation should be examined by more prospective cohort studies in the future. Second, important indicators such as LVEF and cholesterol levels were not sufficient (less than 15%) in the database. Third, the possible implementation of therapeutic hypothermia in this study may decrease the hazard ratios of mortality for patients in lower temperatures and conceal the probable damage in patients with abnormally high temperature.

The study showcased critically ill CHD patients with hypothermia after ICU admission had a higher risk of mortality. Measuring body temperature may provide practical evidence for risk stratification and further research is required to testify this.