The excess insulin requirement in severe COVID‐19 compared to non‐COVID‐19 viral pneumonitis is related to the severity of respiratory failure and pre‐existing diabetes

Abstract Introduction Severe COVID‐19 has been anecdotally associated with high insulin requirements. It has been proposed that this may be driven by a direct diabetogenic effect of the virus that is unique to SARS‐CoV‐2, but evidence to support this is limited. To explore this, we compared insulin requirements in patients with severe COVID‐19 and non‐COVID‐19 viral pneumonitis. Methods This is a retrospective cohort study of patients with severe COVID‐19 admitted to our intensive care unit between March and June 2020. A historical control cohort of non‐COVID‐19 viral pneumonitis patients was identified from routinely collected audit data. Results Insulin requirements were similar in patients with COVID‐19 and non‐COVID‐19 viral pneumonitis after adjustment for pre‐existing diabetes and severity of respiratory failure. Conclusions In this single‐centre study, we could not find evidence of a unique diabetogenic effect of COVID‐19. We suggest that high insulin requirements in this disease relate to its propensity to cause severe respiratory failure in patients with pre‐existing metabolic disease.


| RE SULTS
During the study period March to June 2020, we identified 92 patients treated for severe COVID-19 within our intensive care unit (Appendix S1: Figure S1). In univariate linear regression analysis, BMI, pre-existing type 1 or type 2 diabetes and severity of respiratory failure assessed by ordinal scale (0-self ventilating, 1-mechanical ventilation, 2-neuromuscular blockade, 3-nebulized epoprostenol, 4-prone ventilation, 5-extracorporeal membrane oxygenation) were all associated with insulin dose when analysed by the average dose across the whole ICU stay or by the highest cumulative insulin dose over a 24-h period (Appendix S1: Table S1).
To determine if the observed insulin requirements were a unique feature of COVID-19, we examined insulin requirements in   46 patients admitted to our ICU with a diagnosis of non-COVID-19   viral pneumonitis between 2014 and 2019. 89% of the non-COVID-19 cohort had a confirmed viral isolate consistent with viral pneumonia. Five patients included in the analysis did not have any relevant positive virology related to their ICU admission but were diagnosed with viral pneumonitis on clinical grounds while two patients had more than 1 virus identified that could have explained their symptoms (Appendix S1: Table S2 for a breakdown of viral isolates). Insulin requirements were higher in patients with COVID-19 versus non-COVID-19 viral pneumonitis while mean glycaemic indices were comparable in both groups ( Figure 1A,B and Appendix S1: Figure S2A,B). However, the groups were poorly balanced with respect to diabetes status and requirement for respiratory salvage therapies (Appendix S1: Table S3), and multiple regression analysis did not identify any association between COVID-19 and insulin dose after adjustment for pre-existing diabetes and severity of respiratory failure (Table 1). While severity of respiratory failure was independently associated with insulin requirement, we did not find a similar relationship for vasopressor use or APACHE II score (data not shown). When COVID-19 and non-COVID-19 viral pneumonitis patients were matched (1:1, N = 36 per group) by the level of respiratory support required and pre-existing diabetes status, the proportion of patients requiring insulin (COVID-19: 41%, non-COVID-19: 49%) and insulin dose was similar in those who were treated with insulin ( Figure 1C,D and Appendix S1: Figure S2C,D).
We acknowledge that diabetes is a heterogenous condition with different intensity of treatment and adequacy of glycaemic control and controlling for it as a single variable is limited. However, the number of patients with type 1 diabetes in COVID-19 and non-COVID-19 was the same and the treatments used in type 2 diabetes COVID-19 and non-COVID-19 viral pneumonitis are not dissimilar (Appendix S1: Table S4). Unfortunately, a large proportion of patients in our study did not have pre-morbid HbA 1c values on records at our institution therefore precluding meaningful comparison of pre-morbid glycaemic control.
Finally, we sought to determine if the insulin requirements observed in patients with COVID-19 resolved in convalescence.

Diabetes therapies used in patients with COVID-19 and non-
COVID-19 viral pneumonitis are summarized in Appendix S1: Table   S4. We had information on discharge medication in 61/62 patients who survived to hospital discharge. Of these patients, 39 required insulin while in ICU but on discharge only five patients were discharged from hospital on subcutaneous insulin, four of whom were on insulin prior to admission. The one patient commenced on insulin had type 2 diabetes prior to admission managed with oral therapy, and insulin was discontinued soon after discharge when they were readmitted with hypoglycaemia. There was no initiation or escalation in oral hypoglycaemic agents in the cohort; however, we cannot comment on the adequacy of glycaemic control. Similar findings from a smaller cohort have recently been reported. 3

| DISCUSS ION
We report here, to the best of our knowledge, the first longitudinal assessment of glycaemia and insulin requirements in patients who are critically ill with COVID-19. Our data suggest that high insulin requirements observed in patients with severe COVID-19 are related to the severity of respiratory failure and pre-existing diabetes mellitus rather than a direct diabetogenic effect of SARS-CoV-2. that COVID-19 patients had poorer glycaemic control and higher insulin requirements. However, the COVID-19 group required much more respiratory support compared to the non-COVID-19 group and the authors did not make any attempt to adjust for this difference.
While preparing our paper for publication, two separate analyses have suggested that insulin therapy in hospital may be associated with increased mortality in patients with type 2 diabetes. 9,10 One of these reports has been widely interpreted as supporting the supposition that insulin therapy is detrimental in COVID-19. 11 We would take the view that these findings represent reverse causation-whereby those with severe illness require insulin to treat stress hyperglycaemia. An alternative explanation is that the optimal glycaemic targets