Study Design and Setting

We performed a multicenter (25 centers across India) case–control study during January 1–June 30, 2021 (Appendix Figure). We included CAM patients (cases) and at least 2 COVID-19 patients without mucormycosis (controls) for each case. Only those centers willing to provide data on >15 cases of mucormycosis during the outbreak were included (Appendix Table 1).

The Institute Ethics Committees approved the study protocol at the individual study sites. A consent waiver was granted because we used anonymized patient data for analysis. The study is reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (Appendix) (12).

Cases and Controls

We confirmed COVID-19 diagnosis by SARS-CoV-2 RNA positivity in respiratory specimens by reverse transcription PCR or a positive rapid antigen test. We included only confirmed (proven and probable) cases of mucormycosis diagnosed and managed at the individual centers (13). Participating centers were tertiary care referral hospitals or research institutes equipped with the facilities required to evaluate and manage mucormycosis. We evaluated patient samples at the study sites by conventional microscopy (KOH-calcofluor method), culture, histopathology, or molecular diagnostic techniques, as appropriate. We identified positive cultures on the basis of macroscopic and microscopic characteristics of the growth or by sequencing of the internal transcribed spacer region of rDNA. We subjected tissue samples for histopathological examination and used hematoxylin and eosin, periodic acid Schiff, or Gomori's methenamine silver stain. We diagnosed invasive mucormycosis on the basis of existing guidelines and as previously described (11,13). In brief, we categorized participants’ illness as proven mucormycosis if Mucorales were isolated or broad ribbon-like aseptate hyphae were demonstrated from sterile sites or deep tissue biopsy. We defined probable mucormycosis in the presence of host risk factors, consistent imaging (e.g., reversed halo sign, thick-walled cavity, and others on computed tomography), and the demonstration of Mucorales by either microscopy or culture from a nonsterile site (sputum, nasal smear, or bronchoalveolar lavage fluid) (13,14). We defined disseminated mucormycosis as >1 noncontiguous site being involved. We excluded cases without microbiological or pathological confirmation of mucormycosis. We considered isolation of Mucorales from respiratory secretions without compatible host factors, clinical features, or radiology as colonization, and these cases were not included. We further classified CAM as concurrent (occurrence of mucormycosis within 7 days before or after diagnosing COVID-19) or nonconcurrent (after 7 days but within 3 months of COVID-19 diagnosis).

We enrolled >2 age-matched (+5 years) persons with confirmed COVID-19 as controls for each CAM case. The controls were selected randomly by the individual centers. We also reviewed patient records or contacted control-patients by telephone 3 months after COVID-19 diagnosis to ensure CAM had not developed.

Patients were managed per the local institutional protocol and the treating physician’s discretion. Mucormycosis was treated according to the standard recommendations, subject to the availability of drugs and other factors (1,15). We recorded the first antifungal medication (amphotericin B, posaconazole, isavuconazole, or a combination) used for managing mucormycosis and whether lack of availability of the intended drug led to the use of an alternate agent. We defined the use of >2 antifungal agents effective against Mucorales within the first 14 days of CAM as primary combination antifungal therapy. We also noted whether >1 doses were missed because of drug unavailability.

Exposures and Confounding Variables

The primary exposure variable was the type of treatment offered for COVID-19. The therapies consisted of glucocorticoids, remdesivir, tocilizumab, baricitinib, zinc supplements, antibacterial agents, and antifungal therapy (before onset of mucormycosis). We defined inappropriate glucocorticoid therapy as use of systemic glucocorticoids for COVID-19 without related hypoxemia. We recorded information on the dose and duration of glucocorticoids used for COVID-19 and calculated the cumulative dose of glucocorticoids by multiplying the number of days of therapy and the dose used (dexamethasone equivalent).

We accounted for potential confounding variables pertaining to CAM by retrieving demographic and clinical information about COVID-19 illness from patient records. That information was sex; place of residence (rural or urban); investigations performed during admission (the first available values) for acute COVID-19 illness, including glycated hemoglobin (HbA1C), plasma glucose, complete blood count, serum ferritin, C-reactive protein, and neutrophil-to-lymphocyte ratio; host factors, including diabetes mellitus (labeled as recent-onset if diagnosed during the current admission and there was no previous history of diabetes), diabetic ketoacidosis during admission for COVID-19, hematological malignancy, stem cell or organ transplantation, and immunosuppressive therapy (for indications unrelated to COVID-19); chronic comorbidities (chronic liver, kidney, lung, and other diseases); hypoxemia during acute COVID-19 (<94% oxygen saturation while breathing ambient air or requiring supplemental oxygen); and the need for mechanical ventilation.

We also evaluated factors associated with 12-week mortality in CAM patients. We collected the following data: time to diagnosis of mucormycosis after confirmed COVID-19, diagnosis of mucormycosis (microscopy, culture, histopathology), anatomic site of involvement (rhino-orbital mucormycosis, pulmonary, or others), treatment details (antifungal therapy, surgery, and others), and outcome at 6 and 12 weeks.

Study Size

We assumed the primary exposure of interest (treatment practices, primarily glucocorticoids) would be present in 60% of controls and 70% of cases. The estimated sample size was 374 cases and 747 controls for a case: control enrollment ratio of 1:2, at a power of 90% for detecting 1.25 odds (in cases than controls). We planned a convenient sample size by enrolling >15 consecutive CAM cases from each participating center.

Statistical Analysis

We analyzed data by using the commercial statistical package SPSS Statistics 22.0 (IBM, https://www.ibm.com). As appropriate, descriptive data are presented as frequencies, means and SDs, or medians and interquartile ranges. We compared categorical variables by using the χ² test or Fisher exact test and analyzed the differences between continuous data by using the Student t-test or Mann-Whitney U test, as appropriate. We considered a p value <0.05 significant.

We imputed the missing data for performing the subsequent logistic regression analysis. The pattern of missing data (missing at random or not) was ascertained. We then performed multiple imputations (50 iterations) by using the Markov Chain Monte Carlo method (Appendix Table 2). We performed multivariate logistic regression analyses to identify the factors associated with the development of CAM (vs. COVID-19 controls) and ascertain factors associated with 12-week mortality in persons with CAM. The variables included in the logistic regression model were decided on the basis of the univariate analysis and biologic plausibility. The strength of association was reported as an adjusted odds ratio (aOR) with 95% CI. To elucidate the confounding resulting from the difference in severity among the controls and cases, we repeated the multivariate analysis in the following subgroups: nonhypoxemic versus hypoxemic COVID-19 control-patients and COVID-19 control-patients with and without comorbidities. We also report the sensitivity analysis on the available data (complete case analysis) for the logistic regression.

Exposure

Glucocorticoids were used for COVID-19 in 71.6% (3,890/5,431) patients (cumulative dose range 0.76–679.53 mg dexamethasone equivalent), and their use was more frequent in CAM case-patients. The median dose and duration of glucocorticoid use was also higher in case-patients than control-patients (Table 1). The proportion of persons receiving glucocorticoids in the absence of hypoxemia (inappropriate use) was even higher in CAM patients (34.2% vs. 22.3%; p = 0.0001) than control-patients. In addition, the percentage of persons receiving more than the recommended dose (>60 mg of dexamethasone or equivalent) of glucocorticoids was higher among CAM case-patients than COVID-19 control-patients (Figure), both for hypoxemic (269/439 [61.3%] vs. 683/1389 [49.2%]; p = 0.0001) and nonhypoxemic (124/329 [37.7%] vs. 143/633 [22.6%]; p = 0.0001) persons. Zinc supplementation during COVID-19 illness was higher among case-patients than controls (47.9% vs. 41.3%; p = 0.0001). Remdesivir (49.2% vs. 20.6%; p = 0.0001) and antibacterial therapy (64.2% vs. 61.1%; p = 0.03) were more commonly used in controls. No difference was noted in the proportion of participants receiving tocilizumab, baricitinib, or antifungal therapy before the development of CAM.

Confounders

Chronic comorbidities (chronic liver, kidney, lung, and other diseases) were present in 37% of study participants and were seen more often in controls than in CAM patients. Significantly higher plasma glucose, neutrophil-to-lymphocyte ratio, platelet count, C-reactive protein, and serum ferritin were seen in CAM patients during their hospitalizations for COVID-19. We identified >1 host factors for mucormycosis in 46.9% of controls versus 81.8% of case-patients (p = 0.0001). The proportion of diabetes mellitus was significantly higher in case-patients (80.9% vs. 45.1%; p = 0.0001) than in controls. Recent-onset diabetes mellitus was similar in both the study groups, whereas diabetic ketoacidosis was significantly more common among CAM than in the controls (5.2% vs. 2.7%; p = 0.0003). We found hypoxemia during acute COVID-19 (54.5% vs. 46.2%), admission to an intensive care unit (40.7% vs. 20.5%), and mechanical ventilation (29.9% vs. 9.5%) significantly higher among control-patients.

Association of COVID-19 Treatment Practices with CAM

We found the cumulative dose of glucocorticoid used (OR 1.006, 95% CI 1.004–1.007) and zinc supplementation (OR 2.76, 95% CI 2.24–3.40) independently associated with CAM, in addition to elevated C-reactive protein, host factors (renal transplantation, diabetes mellitus, diabetic ketoacidosis during COVID-19), and rural residence (Table 2). Hypoxemia during COVID-19 and comorbidities were associated with lower odds of CAM. The results were similar on a complete case analysis (Appendix Table 3). Because the proportion of persons with hypoxemia and comorbidities was higher in COVID-19 controls than in CAM patients, we also performed a subgroup analysis (hypoxemic vs. nonhypoxemic controls and any comorbidity vs. no comorbidity; Appendix Tables 4, 5). We found the primary exposure of interest (COVID-19 treatment factors) significantly associated with CAM in both the subgroups (Appendix Tables 4, 5).

We also assessed the association of inappropriate glucocorticoid therapy (i.e., glucocorticoids administered in the absence of hypoxemia) with CAM. In an alternative logistic regression model, we replaced glucocorticoid doses with inappropriate glucocorticoid therapy (Appendix Table 6) and found inappropriate therapy also significantly associated with CAM.

Additional Analyses of Clinical Features, Diagnosis, Treatment, and Outcome of CAM

CAM occurred nonconcurrently (>7 days after COVID-19 diagnosis) in 1,405 (81.7%) of 1,720 persons; the remaining 315 (18.3%) cases were concurrent with the COVID-19 illness (within 7 days of diagnosis). The duration between COVID-19 and the diagnosis of CAM was shorter for those hospitalized for COVID-19 (mean 20, 95% CI 19–21 days) versus those isolated at home (mean 25, 95% CI 24–27 days). The median duration between symptoms of CAM and confirmation of the diagnosis was 6 (interquartile range 3–10) days (n = 1,024 persons). Rhino-orbital mucormycosis accounted for 92.4% of the cases, followed by pulmonary (7%) and other sites (0.6%). The proportion of pulmonary mucormycosis was higher in persons who had undergone organ transplants (17.2%) and persons with hematological malignancies (33.3%) than in those with diabetes mellitus (6.4%). Of the 1,602 patients with rhino-orbital mucormycosis, we noted intracranial extension in 261 (16.3%) cases. Nearly two thirds (1,143/1,733; 66.0%) of the CAM patients had evidence of mucormycosis from >1 sample (smear, culture, or histopathology) (Table 3). The most common etiologic agent causing mucormycosis was Rhizopus spp. (mainly R. arrhizus). The other reported organisms included Mucor and Rhizomucor spp., and rarely Cunninghamella, Syncephalastrum, Apophysomyces, Lichtheimia spp., and others.

The treatment of CAM varied widely and was influenced by antifungal drug availability. The intended antifungal agent could not be administered in 21.0% (321/1,526) of cases, and >1 dose was missed because the antifungal drugs were unavailable in 17% (248/1,547) of CAM cases. Amphotericin B was the most prescribed antifungal agent. A combination of antifungal agents was used for primary therapy in 41.6% of patients. Surgery was performed in most case-patients (1,449/1,733; 83.6%). The mortality rate at 6 weeks (data available for 89.2% of cases) was 28.6% and at 12 weeks (data available for 84.9% cases) was 32.2%.

We found surgical resection and primary antifungal combination therapy independently associated with better odds of survival at 12 weeks (Table 4). We also found increasing age and intracranial extension associated with worse odds of survival after adjusting for sex, comorbidities, and COVID-19–related hypoxemia. The results were similar on a complete case analysis (Appendix Table 7).