Debate continues to rage regarding how ARDS caused by SARS-CoV-2 differs, if at all, from previous viral outbreaks and other causes of ARDS. Such debate includes discussions regarding disease manifestations, which patients are most likely to require intensive care, outcomes, and predictors of mortality. In one meta-analysis, patients with SARS from the 2003–2004 outbreak, with MERS from the 2017–2018 outbreak, and with COVID-19 from the current pandemic were compared and largely perceived to be similar(2). However, this meta-analysis of 28 studies focussed entirely on outcomes. Moreover, whereas all COVID patients were drawn from 10 studies conducted in China, South Korea, Italy, and the USA (10 studies), all SARS patients were extracted from 12 entirely different studies performed in mainland China, Taiwan, Hong Kong, and Canada, and all MERS patients from yet another six studies conducted in Saudi Arabia and South Korea. In other words, no study compared any two of these conditions directly; all the SARS, MERS, and COVID samples were drawn from widely disparate locations; and the studies were conducted in three different decades (SARS: 2003–2004; MERS: 2017–2018; COVID: 2020). In addition, no raw data were collected for multivariable analysis.
In yet another meta-analysis comparing COVID and non-COVID ARDS patients(17), the focus again was only on outcomes, and only one of the ten analysed studies directly compared these two patient groups in an ICU setting(18). Of the remaining nine, five involved retrospective data collection for both COVID and non-COVID cases, six compared COVID patients against non-COVID patients extracted from historical cohorts, two were conducted in emergency departments rather than an ICU, and two studied highly-selective patient populations: one only studying mechanically-ventilated patients, and one only patients undergoing ECMO. A further study only compared 14 COVID patients against seven non-COVID patients with ARDS, and only bronchiolar-lavage specimens were compared. The one study that directly compared COVID and non-COVID ICU ARDS patients only compared 24 COVID against 39 non-COVID patients(18). Another moderately-large French study that was not included in the previous meta-analysis compared 150 COVID and 233 non-COVID ARDS patients, but again only used historical non-COVID cases(19).
Conversely, we compared 160 patients with COVID against 530 patients in whom COVID tests were negative, all prospectively-enrolled over the same ten-month period in the same ICU, and all treated for ARDS. In many regards, our sample of 160 COVID patients were highly similar to COVID samples reported by other investigators(20). Such similarities include the predominance of males, seniors, and obesity, and the frequent complaints of fevers, cough, and dyspnoea. For example, in one study of 24 ICU patients drawn from nine Seattle-area hospitals admitted with confirmed SARS-CoV-2 infection and ARDS, 63% of the patients were male (versus 74% in our sample of COVID patients), their mean age was 64.0 years (vs. 59.4), 58% were diabetic (vs. 56%) and, as in our sample, by far the most common symptoms were cough, dyspnoea, and fever(20). In the above-mentioned French study comparing 24 and 39 COVID versus non-COVID ARDS patients, respectively, the median age of COVID patients was 67, 67% were male, mean BMI was 31.0 (versus 28.4 in our sample), the median duration of symptoms prior to ICU admission was 7 days (vs. 6 days), and 38% (vs. 56%) were diabetic(18). In a Chinese study of 476 hospitalized COVID-confirmed patients, patients were categorized by disease severity — mild, moderate, severe, and critical — and, whereas the median age among patients with mild symptoms was just 51 years, it increased to 61 and 68 in those with severe and critical disease, respectively(13). Similarly, comparing patients with mild versus critical disease, the percentage who were male increased from 54.0 to 68.6%; the percentages complaining of cough and dyspnoea increased from 35.5 and 14.9% to 92.5 and 70.3%; and the percentage who were febrile increased from 82.2 to 97.0%, all numbers consistent with what we observed in our critically-ill COVID patients.
Comparing our 160 COVID and 530 non-COVID ARDS patients revealed numerous differences. Demographically, our COVID patients were roughly five years older, less than half as likely to be under 45-years-old, and 15% more likely to be male (73.8 vs. 58.9%). Age over 45 was not only identified as significantly different on bivariate analysis, but also on multivariable analysis, during which the odds of being under 45 was 2.75 times as great in COVID as non-COVID patients. COVID patients also weighed more (77.4 vs. 72.2kg), and weight itself was retained in our multivariable model (the likelihood of COVID increased by 2% for each 1kg increase in weight). Nonetheless, at p < 0,.007, BMI failed to satisfy our strict, Bonferroni-adjusted p < 0.001 criterion for statistical significance on bivariate analysis and the percentage of patients considered at least obese was no different in COVID and non-COVID patients (30.4 vs. 25.1, p = 0.18).
For pre-existing co-morbidities, the two patient cohorts were similar, with almost half of the patients in both groups having a history of diabetes or hypertension, and a third pre-existing cardiovascular or lung disease. This latter percentage is, admittedly, much lower than the 58% of ICU COVID-ARDS patients reported as having cardiovascular disease alone by Brault et al(18). However, that study was small, with only 24 COVID patients and 39 non-COVID patients; and, as in our study, no statistically-significant difference was detected between COVID and non-COVID patients.
Our COVID patients were much more symptomatic at presentation than our non-COVID patients, in almost all the symptoms recorded, with 86.3% versus 28.5% reporting dyspnoea, 71.9% versus 14.0% fevers and/or chills, and 58.8% versus 11.5% cough. In fact, that triad of fevers and/or chills, dyspnoea, and cough was 95.3% specific for COVID, though only 45.9% sensitive, while the symptom dyad of fevers/chills and dyspnoea was 93.0% sensitive and 63.4% specific. Other symptoms that were appreciably more common in COVID than non-COVID patients were fatigue (42.5 vs. 17.4%), musculoskeletal aches or pain (23.1 vs. 7.7%), loss of taste or smell (35.0 vs. 2.1%), sore throat (29.4 vs. 2.3%), and nasal congestion or coryza (20.0 vs. 2.8%). Headaches, nausea and/or vomiting, and diarrhoea were similarly uncommon in the two patient groups. It must be noted that such differences in symptoms might only pertain to ICU patients, since few differences in the frequency of many of the same symptoms we studied were noted in the study by Shah et al, in which 33 COVID and 283 non-COVID patients presenting to an emergency room were compared(21). Nonetheless, on multivariable analysis of our data, of the ten parameters retained as significant predictors of COVID-19, five — loss of smell/taste (OR = 8.57), dyspnoea (OR = 5.87), fevers and/or chills (OR = 4.90), sore throat (OR = 2.51), and cough (OR = 2.17) — were presenting symptoms.
In addition to being more symptomatic, our COVID patients’ general health status at admission was significantly worse, their mean APACHE-II score 21.1, versus just 18.2 in non-COVID patients. Their admission PaO2 also was markedly lower (53.6 vs. 88.9), as was their PaO2 receiving supplemental oxygen just prior to intubation (92.5 vs. 96.6). However, neither the admission APACHE-II score nor either PaO2 measurement remained as predictors in our final regression model. One difference that must be noted, however, is that every one of the 37 COVID patients whose presenting APACHE-II score exceeded 30 ultimately died, while non-COVID patients survived with admission APACHE-II scores up to 40. On multi-variable analysis, baseline APACHE-II score also remained as one of only two mortality predictors, the other being mechanical ventilation while in the ICU. Combining a baseline APACHE-II score ≥ 17 and mechanical ventilation was 94.4% sensitive and 70.5% specific for mortality. This association, between presenting APACHE-II score and mortality in COVID-19 patients with ARDS, has also been reported by others(22, 23).
With respect to admission-time laboratory results, only a lower lymphocyte count (1.3x109 vs. 2.2x109 cells/litre), lower serum troponin level (0.36 vs. 2.23ng/ml), and higher serum LDH level (568.5 vs. 362.3units/litre) were evident in COVID patients. Serum LDH also was statistically higher on the day of intubation (757.5 vs. 449.9units/litre). What is interesting here is that both lower lymphocyte counts and elevated LDH have been linked to increased mortality in COVID-19 patients in a recently-published systematic review of the literature(24), while elevated troponin levels have been linked to myocarditis(25). However, none of these four inter-group differences detected in laboratory values was retained on multivariable analysis. Serum ferritin, elevated levels of which were found to be predictive of increased mortality in a recently-published meta-analysis of 614 COVID-19 patients(26), was more than twice as high in our COVID than non-COVID patients (2356 vs. 1032µg/l, p = 0.013); however, large variances in levels prevented this difference from satisfying our stringent criterion for statistical significance.
Where our COVID-19 and non-COVID patients were most different was in modes of treatment, with 15.6 versus 3.2% receiving non- invasive intermittent positive pressure ventilation, 48.8 vs. 2.3% a high-flow nasal cannula, 61.9 vs. 1.7% prone-positioning during mechanical ventilation, 4.4 versus 0.4% inhaled nitrous oxide, 53.8 versus 28.1% vasopressors, 9.4 versus 0.8% therapeutic plasma exchange, 9.4 versus 0.6% the anti-viral drug remdesivir, 4.4 versus 0.2% hydroxychloroquine, 86.3 versus 64.5% at least one anti-bacterial agent, 85.0 versus 20.9% corticosteroids, and 33.1 versus 0.8% an IL-6 receptor antagonist. Such is not surprising, however, given the newness of COVID-19 and resulting dearth of data guiding management, its higher mortality rate relative to non-COVID ARDS, and the multiple different mechanisms known to cause ARDS in the absence of COVID-19. Accordingly, treatments commonly used for ARDS in general, like mechanical ventilation and tracheostomy, were no different in the two patient groups.
A markedly-elevated mortality rate — 58.8 versus 25.9% — was observed in our COVID-19 relative to non-COVID patients. Such a high rate of mortality is consistent with rates reported elsewhere, including one study in which 96.6% of mechanically-ventilated COVID-19 patients perished(23). We again point out the limitations of the one recently-published meta-analysis which failed to detect any significant differences in mortality comparing COVID-19 and non-COVID patients with ARDS(17), limitations which included the inclusion of six (out of ten) studies using historical, rather than contemporary non-COVID cases, five using retrospectively-collected data on COVID-19 patients, two analysing data from emergency department rather than ICU patients, and two restricting their analysis to just mechanically-ventilated or ECMO patients.
Another difference we detected in an adverse outcome that is not commonly reported was a more than five-fold rate of required supplemental oxygen among COVID survivors, required by 15/66 COVID-19 patients who survived (22.7%) versus just 16/393 surviving non-COVID patients (4.1%; OR = 6.93, CI = 3.23, 14.86). This highlights the potential long-term adverse health consequences of COVID-19, especially among those who develop critical illness, the frequency and severity of which only time will tell(27).
Every study, including ours, has strengths and limitations. Among its strengths are prospective data collection, extensive number of variables examined, and very low rate of missing data, especially among the numerous baseline patient characteristics, measures of health status, and outcomes, where data capture exceeded 99%. Among its weakness are the lack of discrimination between pre-admission cardiovascular and pulmonary disease, and between pre-admission renal and hepatic disease, and the lack of imaging data, since chest radiographic findings have been identified as predictors of mortality in COVID patients(28, 29) and may provide further insights into the differences between ARDS caused by SARS-CoV-2 versus not.