Diabetic ketoacidosis in patients with SARS-CoV-2: a systematic review and meta-analysis

Background One possible reason for increased mortality due to SARS-CoV-2 in patients with diabetes is from the complication of diabetic ketoacidosis (DKA). Objectives To re-evaluate the association of SARS-CoV-2 and development of DKA and analyse the demographic and biochemical parameters and the clinical outcomes in COVID-19 patients with DKA. Design A systematic review and meta-analysis. Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement was followed. Methods Electronic databases (Proquest, Medline, Embase, Pubmed, CINAHL, Wiley online library, Scopus and Nature) were searched from 1 December 2019 to 30 June 2021 in the English language using the following keywords alone or in combination: COVID-19 OR SARS-CoV-2 AND diabetic ketoacidosis OR DKA OR ketosis OR ketonemia OR hyperglycaemic emergency OR hyperglycaemic crisis. We included studies in adults and children of all ages in all healthcare settings. Binary logistic regression model was used to explore the effect of various demographic and biochemical parameters variables on patient’s final treatment outcome (survival or death). Results Of the 484 papers that were identified, 68 articles were included in the systematic review and meta-analysis (54 case report, 10 case series, and 4 cohort studies). Studies involving 639 DKA patients with confirmed SARS-CoV-2 [46 (7.2%) were children and 334 (52.3%) were adults] were analyzed. The median or mean patient age ranged from < 1 years to 66 years across studies. Most of the patients (n = 309, 48.3%) had pre-existing type 2 diabetes mellitus. The majority of the patients were male (n = 373, 58.4%) and belonged to Hispanic (n = 156, 24.4%) and black (n = 98, 15.3%) ethnicity. The median random blood glucose level, HbA1c, pH, bicarbonate, and anion gap in all included patients at presentation were 507 mg/dl [IQR 399–638 mg/dl], 11.4% [IQR 9.9–13.5%], 7.16 [IQR 7.00–7.22], 10 mmol/l [IQR 6.9–13 mmol/l], and 24.5 mEq/l [18–29.2 mEq/l]; respectively. Mortality rate was [63/243, 25.9%], with a majority of death in patients of Hispanic ethnicity (n = 17, 27%; p = 0.001). The odd ratios of death were significantly high in patients with pre-existing diabetes mellitus type 2 [OR 5.24, 95% CI 2.07–15.19; p = 0.001], old age (≥ 60 years) [OR 3.29, 95% CI 1.38–7.91; p = 0.007], and male gender [OR 2.61, 95% CI 1.37–5.17; p = 0.004] compared to those who survived. Conclusion DKA is not uncommon in SARS-CoV-2 patients with diabetes mellitus and results in a mortality rate of 25.9%. Mortality key determinants in DKA patients with SARS-CoV-2 infection are individuals with pre-existing diabetes mellitus type 2, older age [≥ 60 years old], male gender, BMI ≥ 30, blood glucose level > 1000 mg/dl, and anion gap ≥ 30 mEq/l. Supplementary Information The online version contains supplementary material available at 10.1186/s13098-021-00740-6.


Background
Diabetes is a frequent comorbidity in patients with severe acute respiratory syndrome coronavirus 2 [SARS-CoV-2], with a reported prevalence ranging from 9 to 20% [1][2][3][4]. Diabetes is also associated with more than twofold higher risk of having severe or critical corona virus disease 2019  illness and about threefold increased risk of in-hospital mortality compared to SARS-CoV-2 patients without diabetes [1][2][3][4]. A possible reason for increased mortality due to SARS-CoV-2 in patients with diabetes is from the complication of diabetic ketoacidosis (DKA), one of the most serious acute complications of diabetes. DKA is characterized by the presence of hyperglycaemia [usually < 800 mg/ dl and generally between 350 to 500 mg/dl], arterial pH [≤ 7.30], anion gap [> 12 mEq/l], and serum bicarbonate [≤ 15 mmol/l] [5].
In light of newer case reports, case-series and cohort studies that were done to re-evaluate the association of SARS-CoV-2 and development of DKA, we aimed to analyse the demographic and biochemical parameters and the clinical outcomes in COVID-19 patients with DKA with larger and better-quality data.

Design
We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (PRISMA) in conducting this systematic review and meta-analysis [6]. The following electronic databases were searched: PROQUEST, MEDLINE, EMBASE, PUBMED, CINAHL, WILEY ONLINE LIBRARY, SCOPUS and NATURE with Full Text. We used the following keywords: COVID-19 OR SARS-CoV-2 AND diabetic ketoacidosis OR DKA OR ketosis OR ketonemia OR hyperglycaemic emergency OR hyperglycaemic crisis OR euglycemia OR euglycemic. The search was limited to papers published in English between 1 December 2019 and 30 June 2021. Based on the title and abstract of each selected article, we selected those discussing and reporting occurrence of DKA in COVID-19 patients. We also utilized backward snowballing to increase the yield of potentially relevant articles (Additional file 1).

Inclusion-exclusion criteria
We included case reports, case series and cohort studies, but excluded editorials, commentaries, case and animal studies, discussion papers, preprints, news analyses, and reviews. We considered studies to be eligible regardless of experimental or observational design, and irrespective of their primary objective. However, we excluded studies that did not report data on DKA and SARS-CoV-2; studies that never reported details on SARS-CoV-2 identified cases with DKA; or studies that reported DKA in patients with negative PCR COIVD-19 tests. We evaluated studies that included all children and adults as our population of interest who experienced DKA and SARS-CoV-2 infection during the period from December 1, 2019 through June 30, 2021.

Data extraction
Four authors (S.A., A.A., A.R. and Z.A.) critically reviewed all of the studies retrieved and selected those judged to be the most relevant. The abstracts of all citations were examined thoroughly. Data were extracted from the relevant research studies using key headings, which are noted in Table 1, simplifying analysis, and review of the literature. Articles were categorized as case report, case series or cohort studies.
The following data were extracted from selected studies: authors; publication year; study location; study design and setting; age; proportion of male patients; patient body mass index [BMI] and ethnicity; type of diabetes [newly diagnosed or pre-existing]; use of sodiumglucose transport protein 2 [SGLT2] inhibitors; patient biochemical parameters at hospital presentation [blood glucose level, HbA1c, pH, bicarbonate, and anion gap]; assessment of study risk of bias; and treatment outcome [survived or died].

Quality assessment
The quality assessment of the studies was undertaken based on the Newcastle-Ottawa Scale (NOS) to assess the quality of the selected studies [7]. This assessment scale has two different tools for evaluating case-control and cohort studies. Each tool measures quality in the three parameters of selection, comparability, and

Data analysis
Descriptive statistics were used to describe the data. For continuous variables, mean and standard deviation were used to summarize the data; and for categorical variables, frequencies and percentages were reported. Differences between the COVID-19 and DKA survival group and COVID-19 and DKA death group were analyzed using the Chi-square (χ 2 ) tests (or Fisher's exact tests for expected cell count < 5 in more than 20% of the cells).
To explore the effect of various demographic and biochemical parameters variables on patient's final treatment outcome [survival or death] in COVID-19 cases who presented with DKA and included in our review, binary logistic regression model with the univariate and multivariate logistic regression of the complete model; and their odd ratios [ORs], confidence intervals (CIs) and p-values were produced; and forest plots were generated for visualization purposes. All p-values were based on two-sided tests and significance was set at a p-value less than 0.05. R version 4.1.0 with the packages finalfit and forestplot was used for all statistical analyses.

Patient clinical outcome and predictors of mortality
Potential determining variables associated in survival and death groups were analyzed through binary logistic regression analysis and shown in Fig. 2, Fig. 3 and Fig. 4 Fig. 3 and Fig. 4.
These variables were considered needing further evaluation and, thus, were included in multivariate regression analysis. Nevertheless, multivariate analysis confirmed old age [≥ 60 years], male gender, and a preexisting diabetes mellitus type 2 were significantly associated with increased death. Although univariate analysis showed black ethnicity was significantly associated with increased mortality (p = 0.04), however, this finding was not reciprocated by multivariate analysis; Fig. 2.

Discussion
This is the largest meta-analysis on the development of DKA in patients with SARS-CoV-2. This study involving 639 patients from 68 observational studies found majority of the DKA patients diagnosed with SARS-CoV-2 were adults (52.3%), men (58.4%) and had pre-existing type 2 diabetes mellitus (48.3%).
DKA is one of the most common and serious hyperglycaemic emergency; and is considered a precipitating event that frequently occurs due to infection [often pneumonia or urinary tract infection], and discontinuation of or inadequate insulin therapy [76,77]. Adults of any age may develop severe SARS-CoV-2 and experience adverse outcomes, especially those with comorbidities [78,79]. Most children with SARS-CoV-2 have mild symptoms or have no symptoms at all [80], however, adults are at higher risk to experience more severe COVID-19  infection than children [81]. Factors proposed to explain the difference in severity of COVID-19 in children and adults include: 1-age-related increase in endothelial damage and changes in clotting function; 2-higher density, increased affinity and different distribution of angiotensin converting enzyme 2 receptors and transmembrane serine protease 2; 3-pre-existing coronavirus antibodies (including antibody-dependent enhancement) and T cells; 4-immunosenescence and inflammaging, including the effects of chronic cytomegalovirus infection; 5-a higher prevalence of comorbidities associated with severe COVID-19 and 6-lower levels of vitamin D [82]. Hence, lower rate of children SARS-CoV-2 patients with DKA in our review can be justified by the fact that the high severity of COVID-19 tends to be much lower in children compared to adults. DKA is thought to happen most often in patients with diabetes mellitus type 1 [49,83], however, this conceptualization is not true and we report fourfold higher rate of DKA in the diabetes mellitus type 2 patients compared to diabetes mellitus type 1. Type 2 diabetes mellitus patients have high susceptibility to DKA under stressful conditions such as trauma, surgery or infections [83]; and majority of the DKA cases worldwide occur in patients with type 2 diabetes due to its higher prevalence [84,85]. DKA occurs more commonly in adult COVID-19 patients with type 2 diabetes mellitus mainly because the worldwide prevalence of diabetes mellitus type 2 is estimated at 9.3 percent in adults, equivalent to 463 million people [86]. Type 2 diabetes accounts for over 90 percent of patients with diabetes [86,87].
In our review, males gender predominated development of DKA in SARS-CoV-2 patients, a finding suggested in most of the reports [11-13, 15-17, 19, 20, 23, 25, 26, 31, 33, 36, 37, 39, 47, 50, 52, 55-58, 60-64, 69, 74, 75] and in contradiction with data from other reports suggesting an equal proportion of DKA cases in COVID-19 patients for both genders [30,34,38,49]. Lifestyle, body fat distribution, hormonal factors, susceptibility to glucotoxicity and lipotoxicity, and changes in insulin sensitivity have been described as potential factors of DKA and possible mechanisms of male predominance [88]. However, male excess in DKA in our review might be attributed mainly to the differences in the inclusion criteria and the population age groups included in the studies; or can be explained by social factors as women are often the primary caregivers for their families, assuming the responsibility of family members' disease management, at the expense of their own treatment [89].
A comparison of the current results with findings from previous studies can offer some validation of the findings of this present meta-analysis and identify methodological differences in their approaches. Regarding the mortality rate in patients who developed DKA during SARS-CoV-2 infection, we report an overall similar and slightly lower death rate [25.9%] compared to the previous two systematic meta-analyses [28.9 and 29%, respectively] [90,91]. The current meta-analysis is more comprehensive and included a total of 68 studies  including a total of 639 patients; whose details on final treatment outcome were available; in comparison to smaller sample size in previous meta-analyses [sample size: n = 45 and n = 21, respectively] [90,91]. The inclusion of 48 recently published studies [8-14, 16, 21, 23-26, 28, 30, 32-35, 40, 42-46, 48, 50, 51, 53, 54, 56-59, 61-64, 66-75] contributed to the refinement on evidence of the demographic, biochemical, and clinical characteristics; in addition to final therapy outcome in COVID-19 patients with DKA.
Consistent with previous meta-analysis, we found development of DKA in SARS-CoV-2 patients was highest in the Hispanics and blacks (24.4% and 15.3%, respectively) [91]. Moreover, we found mortality rate in DKA patients infected with COVID-19 was significantly very high in patients with Hispanic ethnicity [27%, p = 0.001] in whom risk of acquiring SARS-CoV-2 and clinical prognosis of this viral infection was previously described as high and poor [92,93]. Because most of the studies included in our review that reported the ethnicity of DKA cases infected with COVID-19 were either from the United States of America, India or United Kingdom; representation of other ethnicities at risk to develop DKA during COVID-19 can be misleading. For instance, we report a low prevalence of DKA in Asian population, yet, a systematic review and meta-analysis reported the highest DKA incidence rates in Chinese people [94].
In line with our findings, severely low pH (i.e. pH of < 7) has been identified as an important predictor of mortality in patients with DKA and COVID-19 compared to those who survived (p = 0.047) [90,91]. Very high uncontrolled random blood glucose level (> 1000 mg/dl) was the other biochemical parameter at presentation that differed significantly between the survival and death groups in DKA patients infected with SARS-CoV-2 (p = 0.048); a finding suggested in previous studies [91,95,96] and in contradiction with data from case reports demonstrating death in DKA cases during COVID-19 infection when their blood glucose levels were kept at < 500 mg/dl [27,59,63]. Moreover, increasing age in combination with male gender and BMI ≥ 30 might denote seriously sick patients who can potentially have more morbidity and propensity to die. The majority of patients hospitalized with SARS-CoV-2 are older and seemed to have underlying medical conditions [97,98], with increased age being associated with clinical severity, including case fatality [97,99]. Fortunately, however, mortality from DKA in elderly people have also declined dramatically during the past 10 years [100]. Therefore, these patients should be identified at the earliest and treated preferably in a special care set up to avoid morbidity and mortality. It is worth mentioning increasing age in patients may result in increased hospital stay and might put SARS-CoV-2 patients at risk to develop medical complications like coagulopathy, pneumonia, acute respiratory distress syndrome, organ failure and nosocomial coinfections [97,101]. The presence of these factors in severely ill patients may have necessitated the use of advanced therapies like renal replacement therapy or ventilator support which would have delayed hospital discharge [102]. Although COVID-19 has a higher survival rate than other chronic diseases, the incidence of complications in the geriatric population are considerably high, with more systemic complications [103]. Of the patients admitted to hospital for management of COVID-19, 49.7% (36,367 of 73,197) had at least one complication [104]. Overall, complications and worse functional outcomes in patients admitted to hospital with SARS-CoV-2 are high in old people, and even in young, previously healthy individuals; and COVID-19 complications could strain health system for years.
In our review, the odd ratio of mortality was the highest in DKA patients with the pre-existing the diabetes type 2 variable [OR 5.24, 95% CI 2.07-15.19; p = 0.001]; and DKA patients with pre-existing type 1 diabetes had very low OR of death [OR 0.83, 95% CI 0.23-2.92; p = 0.774]. In diabetes mellitus type 2 diabetes, underlying severe illness is almost always the direct cause of both the DKA and ensuing death; while in diabetes mellitus type 1 diabetes, DKA is most often caused by missed insulin doses but death is rare with prompt treatment [49].
There is growing evidence to suggest that SARS-CoV-2 might cause diabetes in some people [105,106]. In our study, out of the 639 DKA patients infected with SARS-CoV-2, there was (n = 75, 11.7%) newly diagnosed diabetes mellitus cases at admission. Of these  [8, 9, 11, 12, 20, 22, 33-35, 39, 40, 44, 45, 53, 56, 67, 69] and three of which had a BMI > 30 [44,69], suggesting these patients had undiagnosed diabetes mellitus and improbable was caused by SARS-CoV-2 infection. DKA in COVID-19 patients was the least to occur in newly diagnosed diabetes cases probably as a result of increased diabetes screening and early recognition, DKA now occurs more frequently in persons with established diabetes rather than at the time of the initial diagnosis [100]. COVID-19 likely unmasked existing diabetes mellitus by aggravating its metabolic complications rather than causing the new-onset diabetes in these patients. Out of the 11 (1.7%) DKA cases infected with COVID-19 who were taking SGLT2 inhibitors, two patient [18.2%] were diagnosed with SGLT2-associated euglycemic DKA [blood glucose < 250 mg/dl at presentation] [16,48]; in addition to seven patients who had gestational diabetes mellitus [53,65,70]. Euglycemic DKA is a rare life-threatening complication associated with the use of SGLT2 inhibitors in patients with type 2 diabetes that may be unnoticed, particularly in COVID-19 pandemic, due to the absence of significant hyperglycaemia, delaying its treatment [16]. Given their undisputed cardiovascular and renal benefits, these medications are common in patients with type 2 diabetes [107]. There are recommendations that patients using SGLT2 inhibitors should be monitored for ketosis using available home testing kits in case of infections and should discontinue the medication in case of SARS-CoV-2 while the administration of insulin is considered the safest pharmacotherapy choice [108].

Limitations
First, while most of the evidence discussed were based on few cohorts, some case series and many case reports, many of these are small and not necessarily generalizable to the current COVID-19 clinical environment. Second, to asses factors associated with mortality, larger cohort of patients is needed. Third, almost all studies included in this review were retrospective in design which could have introduced potential reporting bias due to reliance on clinical case records. Fourth, study was not registered in Prospero, an international prospective register of systematic reviews, as this might have added extra work and the merit was mostly limited to the avoidance of duplication. Last, the study population included paediatric patients and hence its results cannot be generalized to adult patients.

Conclusion
Patients with diabetes are at increased risk of severe complications from SARS-CoV-2 which may include DKA. Acute diabetes-related DKA in SARS-CoV-2 patients lead to increased mortality; key determinants are individuals with pre-existing diabetes mellitus type 2, older age [≥ 60 years old], male gender, BMI ≥ 30, blood glucose level > 1000 mg/dl, and anion gap ≥ 30 mEq/l.