Protective effect of high body mass index in elderly critically ill with severe COVID-19 pneumonia

2.1 Setting

A retrospective cohort study was carried out in a combined 25-bed, level 3 ICU at General and Teaching Hospital Celje, Slovenia, spanning a 12-month period from March 2020 to February 2021.

2.2 Patients

This ICU was exclusively designated for the treatment of adult (≥18 years old) SARS-CoV-2 positive patients. COVID-19 diagnosis was established through the identification of a positive result in at least one real-time polymerase chain reaction (RT-PCR) test for SARS-CoV-2, conducted on respiratory specimens such as nasopharyngeal swab, sputum or lower respiratory tract specimens. All consecutive patients aged 70 years or older were included in the study, while those with a history of congestive heart failure were excluded from the final analysis.

2.3 Patient data

Information about patients, encompassing fundamental demographic details, previous medical history, and specifics regarding chronic conditions like malignant disease, arterial hypertension, diabetes, heart failure, chronic obstructive pulmonary disease, and chronic renal disease, was gathered from the hospital electronic database BIRPIS21 (SRC Infonet, Kranj, Slovenia).

2.4 Laboratory

The General Laboratory of our institution conducted the majority of laboratory analyses. Our focus on laboratory data encompassed admission details as well as the most extreme values recorded throughout the entire ICU stay. This included parameters such as the lowest pH, highest partial pressure of carbon dioxide (pCO₂), lowest partial pressure of oxygen (pO₂), highest D-dimer, highest troponin T, highest procalcitonin (PCT), highest C-reactive protein (CRP), highest creatinine, highest leukocyte count, and others.

2.5 Body mass index (BMI)

The weight of each patient was measured at admission using an internal scale system of ICU-certified beds (Progressa ICU Hospital Bed, Hill-Rom Holdings, Illinois, USA). Only admission weight data was used in BMI calculation. The height of each patient was measured at admission using standard tape measure in supine position, from top of the head to the heels. Body Mass Index (BMI) was determined using the standard formula: weight (kg) divided by [height (m)]². Obesity was defined as a BMI equal to or exceeding 30 kg/m² [14].

2.6 Echocardiography

A transthoracic echocardiographic examination was conducted at admission by the intensive care specialist using a cardiac probe on the GE Vivid S60 Ultrasound machine (GE Healthcare, Chicago, IL, USA). The collected data included left ventricular ejection fraction (LV EF), determined by eyeballing, velocity time integral (VTI) in the left ventricular outflow tract (LVOT), tricuspid annular plane systolic excursion (TAPSE), and maximal inferior vena cava diameter (VCI max). The definitions of left ventricular heart failure were established in accordance with the American Heart Association Guidelines [15].

2.7 Treatment

Data regarding treatment was extracted from the intensive care information system (Centricity Critical Care, GE Healthcare, USA). Initially, all patients were administered parenteral methylprednisolone at a dose of 1 mg/kg body weight.

To characterize respiratory support, the following data were collected: instances of self-proning and proning during mechanical ventilation, frequency and duration of high-flow nasal cannula (HFNC) oxygen therapy, non-invasive ventilation (NIV), and invasive mechanical ventilation (IMV). Duration of HFNC and NIV respiratory support was guided by periodically calculated respiratory rate-oxygenation (ROX) index values [16]. For patients on IMV, the following maximal values at any time during treatment were recorded: positive end-expiratory pressure (PEEP), peak inspiratory pressure, and tidal volume.

Documentation also included information on the use of nitric oxide inhalation therapy, norepinephrine, levosimendan, systemic thrombolytic treatment, and renal replacement therapy.

2.8 Complications and mortality

Complications data, including ventilator-associated pneumonias, catheter-related bloodstream infections, urosepsis, ICU mortality, and hospital mortality, were gathered.

2.9 Definitions

Ventilator-associated pneumonia (VAP) was defined as the presence of new or changing infiltrates on chest X-ray, occurring more than 48 hours after the initiation of IMV, in addition to both of the following criteria: (i) new onset of fever (body temperature ≥38 °C) or hypothermia (body temperature ≤35 °C) and/or leukocytosis (total peripheral white blood cell (WBC) count ≥10,000 cells/μL) or leukopenia (total peripheral WBC count ≤4500 cells/μL) with >15% immature neutrophils; (ii) new onset of suction respiratory secretions and/or the need for acute changes in ventilatory support parameters to improve oxygenation.

Catheter-related bloodstream infection (CRBSI) was defined as the identification of bacteraemia originating from an intravenous catheter. Microbiological samples were obtained using BacT/ALERT SA (aerobic) and BacT/ALERT SN (anaerobic) bottles, which were then incubated in the BacT/Alert 3D blood culture instrument (bioMérieux, Ballerup, Denmark).

The diagnosis of fungal infection was made using the established criteria as described previously [17].

2.10 Primary outcome

The primary outcomes assessed were mortality at ICU and hospital discharge.

2.11 Secondary outcome

The correlation between ICU complications and ICU mortality was analysed as a secondary outcome.

2.12 Sample size estimation

It was determined that the sample size of 96 patients (48 in each group) was required to detect statistically significant differences (type I error (α) 0.05 and power, type II error (β) 0.20) in BMI between ICU survivors and non-survivors. This estimation was based on the proportion of survivors (0.60) with BMI >30 kg/m² and non-survivors (0.40) with BMI 25–30 kg/m² [18]. Sample size estimation was conducted using MedCalc version 12.5 (MedCalc Software Ltd, Ostend, Belgium).

2.13 Statistical analysis

The data were summarized as mean (±standard deviation) for metric variables and as absolute and relative frequencies for categorical variables. The D’Agostino-Pearson test for normal distribution did not reject the null hypothesis stating that variables were normally distributed. Student’s t-test and Mann-Whitney test for independent samples were employed for metric variables, while Chi-Square was used for categorical data. Multivariate logistic regression modelling (Backward method, enter variable if p < 0.05, remove variable if p > 0.1), with odds ratio (OR) and 95% confidence interval (95% CI) calculations, was used to test the relationship between ICU/hospital mortality and dichotomous BMI (≥30 kg/m² as 1, <30 kg/m² as 0) at admission, age, gender, chronic diseases, C-reactive protein level, lactate at admission and mechanical ventilation at any time during ICU stay. Significant correlations between continuous variables were excluded before inclusion in statistical models. The analyses were conducted using SPSS version 25.0 (SPSS Inc., Chicago, IL, USA) and MedCalc version 12.5 (MedCalc Software Ltd, Ostend, Belgium) software packages. A p-value of < 0.05 was deemed to indicate statistical significance.

4.1 Obesity and its relationship to mortality

Obesity and overweight nutrition status have been identified as unfavorable outcome predictors for patients infected with novel coronavirus, namely, the higher the BMI, the worse the outcomes in the population of unselected patients [19].

On the other hand, analysis from the SMAtteo COVID-19 REgistry (74 out of 331 patients had BMI >30 kg/m², average age 67 ± 21 years) supports our results [9]. They asserted that, despite the complex management of obese patients in the ICU, which includes challenges related to ventilation and nursing, obesity should not be considered a definitive predictive factor for an unfavorable outcome in intensive care.

Obesity (mortality hazard ratio (HR): 0.69, 95% CI: 0.09 to 5.29, p = 0.72) was also not predictive of COVID-19 associated mortality in elderly patients from a long‑term care facility [20].

In 3019 elderly (>65 years) and metabolically unhealthy (defined as metabolic abnormalities including diabetes, hypertension and dyslipidemia) hospitalised patients from Wuhan, China, underweight nutritional status (HR: 4.58, 95% CI: 1.56 to 13.48, p = 0.03) and obesity (HR: 6.49, 95% CI: 2.35 to 17.95, p = 0.001) have been associated with COVID-19 mortality [21]. Notably, overweight nutritional status was not associated with mortality in either metabolically healthy or metabolically unhealthy elderly patients.

Our study observed a U-shaped curve association between BMI and COVID-19 mortality, which is consistent with findings from a study involving 10,861 adult patients with SARS-CoV-2 infection in New York, along with other smaller studies [22, 23]. The findings in the report by Kompaniyets L. et al. [24] suggest that overweight and obesity are risk factors for invasive mechanical ventilation. Additionally, obesity or severe obesity are identified as risk factors for hospitalization, ICU admission, and death among patients aged ≥65 years. However, the U-curve (or J-shaped curve) for predicting ICU mortality in patients aged >75 years has been almost flat, indicating no influence of BMI on the mortality of elderly population [24].

In a recent study involving 222 patients (mean age of 62.4 ± 12 years), individuals with moderate obesity (BMI 30–39.9 kg/m²) exhibited a lower mortality rate compared to patients with normal weight, overweight, or severe obesity [4]. Consistent with these findings are results of Japanese study that confirmed protective effect of higher BMI [25].

Obesity paradox presents a paradox within the paradox [26]. This phenomenon could be elucidated by a potentially protective role of excess adipose tissue, creating a more favorable environment to endure a substantial reduction in the patient’s caloric intake, which is frequently encountered in intensive care settings, such as sepsis or ventilator-induced lung injury [27]. Nevertheless, it is important to acknowledge the limitations of a blind BMI classification, as it is unable to distinguish between muscle mass and adipose tissue.

Another plausible explanation for the paradox is that clinicians might perceive patients with obesity as being at a higher risk for adverse outcomes, leading to their earlier admission, more assertive and timely medical interventions [28].

The obesity paradox in patients with ARDS has been the subject of exploration in numerous studies and two recent meta-analyses [13, 29]. In the study conducted by Ni and colleagues [13], which involved the analysis of evidence pertaining to the association between BMI and clinical outcomes in 6268 ARDS patients, the authors concluded that both obesity and morbid obesity were associated with a lower mortality rate in individuals with ARDS, thus supporting the concept of the obesity paradox. In a separate analysis that included four additional studies, Zhi et al. [29] reported that obesity heightened ARDS-associated morbidity in the ICU population; nonetheless, mortality due to ARDS was lower in obese patients compared to non-obese individuals.

The prevalence of obesity and its impact on outcomes in patients with COVID-19 were evaluated in a recent systematic review and meta-analysis, which included 19 studies [30]. Obesity was not found to be associated with a higher risk of mortality, potentially attributable to the obesity survival paradox and/or unidentified factors.

As age increases, the association between the obesity paradox and mortality appears to strengthen [31]. Some recent studies have reported similar findings [12, 32]. For instance, Fukuoka et al. [12] identified a significant obesity paradox among older patients (aged ≥70 years) but did not observe a similar phenomenon among younger patients with acute myocardial infarction after percutaneous coronary intervention. They proposed two potential explanations for this finding. Firstly, elderly patients with higher BMI are expected to be more resilient to the deleterious effects of obesity than younger patients, potentially due to survival bias [12]. Secondly, older patients may have a higher composition of adipose tissue than younger patients with the same BMI [33], reinforcing the association between BMI and mortality.

The obesity paradox extends beyond a mere association between BMI and mortality, underscoring the significance of illness severity in understanding this complex phenomenon. Lactate is a well-known indicator of the severity of acute circulatory failure [34]. In our study, non-survivors exhibited more severe circulatory/respiratory failure (as evidenced by higher lactate levels at admission), and a greater need for invasive mechanical ventilation at any time during ICU stay. In the multivariate logistic regression model, lactate and need for invasive mechanical ventilation were positively associated, while BMI was negatively related to hospital mortality. In a previous study, the mortality benefit that was adjusted for in patients with higher BMI was found to be diminished in those with elevated lactate levels. No mortality benefit was observed in higher BMI categories when lactate levels exceeded 5 mmol/L [35].

Other authors maintain a skeptical stance and offer methodological explanations to elucidate the obesity paradox, such as different types of bias, namely confounding or selection bias [36]. The first potential type of bias is confounding bias, which can arise, particularly as a consequence of reverse causation. In this scenario, pre-existing disease results in unintended weight loss and higher mortality, creating an illusion of obesity being protective [37].

Other potential types of bias involve BMI group misallocation, especially when estimating rather than actually measuring height and weight in critically ill patients, (both were measured at admission in our study) [38]. Additionally, treatment biases may occur due to closer monitoring and different treatment approaches for obese patients [39].

4.2 Mortality rates in elderly patients with severe or critical COVID-19 pneumonia

Both age and frailty exhibit independent associations with adverse outcomes in individuals with COVID-19 [40]. The mortality rate among elderly patients following admission to the intensive care unit for COVID-19, as reported in a recent study [41], was comparable to the mortality observed in our cohort. Notably, our study was conducted during the second surge of COVID-19 in Europe, a period characterized by higher mortality rates among critically ill elderly patients compared to the initial surge [42].

We have included only elderly critically ill, analyzed based on BMI. The important point that our research did not directly explore is influence of age. Researches of large retrospective study in Spain [43] that included 5746 patients, found that obesity was negatively correlated with negative outcomes in patients aged 65 years or younger, but did not identify the same in older patients. The study was not limited to ICU setting. Furthermore, Graziano et al. [44] did not find abdominal adiposity to be associated with ICU admissions or mortality in 195 non-critically ill patients with mean age of 71 years. Similarly, BMI was only associated with increased in-hospital mortality in patients aged 50 years or younger in study conducted by Hendren et al. [23] in association with American Heart Association registry data. Reported data further indicate that obesity paradox could be limited to elderly patients with COVID-19 pneumonia and is not applicable to younger patients.

Consistent with our study, nosocomial infections were already identified as a significant risk factor for mortality (OR: 7.86, 95% CI: 2.16 to 28.57, p = 0.002) of elderly (mean age 72 years) patients hospitalized with COVID-19 in Korea [45] as is also true in younger patients [46].

4.3 Limitations of our study

This study was designed as a cohort study, utilizing a complete sample from a single center, and it involved the utilization of data from ICU admission to the hard endpoint of mortality or discharge. Nevertheless, this study had the following limitations. Firstly, our study only analyses elderly critically ill patients with COVID-19 pneumonia. Results thus might not be generalizable to all age groups. Secondly, retrospective design of our study did not allow assessment of body weight changes during ICU stay. Thirdly, our study was not designed to determine basal level of sarcopenia and frailty status of our cohort, which is important in elderly [47]. Fourthly, we gathered all data during the hospitalization period and did not perform a long-term follow-up after discharge. Fifthly, the reversal causality could not be completely ruled out. Lastly, the limited number of included patients has hindered the feasibility of conducting complex statistical modeling.