key: cord-0899246-2os5afe7
authors: Jiang, Nan; Chen, Zhenyuan; Yin, Xiaoxv; Liu, Li; Yang, Heping; Tan, Xiangping; Wang, Jing; Li, Hui; Tian, Mengge; Lu, Zuxun; Xiong, Nian; Gong, Yanhong
title: Association of metformin with mortality or ARDS in patients with COVID-19 and type 2 diabetes: a retrospective cohort study
date: 2020-12-10
journal: Diabetes Res Clin Pract
DOI: 10.1016/j.diabres.2020.108619
sha: 2759d0b19a932c091d2d4af2d978b1be7be1a4c7
doc_id: 899246
cord_uid: 2os5afe7

Aims To determine the association between metformin use and mortality and ARDS incidence in patients with COVID-19 and type 2 diabetes. Methods This study was a multi-center retrospective analysis of COVID-19 patients with type 2 diabetes and admitted to four hospitals in Hubei province, China from December 31st, 2019 to March 31st, 2020. Patients were divided into two groups according to their exposure to metformin during hospitalization. The outcomes of interest were 30-day all-cause mortality and incidence of ARDS. We used mixed-effect Cox model and random effect logistic regression to evaluate the associations of metformin use with outcomes, adjusted for baseline characteristics. Results Of 328 patients with COVID-19 and type 2 diabetes included in the study cohort, 30.5% (100/328) were in the metformin group. In the mixed-effected model, metformin use was associated with the lower incidence of ARDS. There was no significant association between metformin use and 30-day all-cause mortality. Propensity score-matched analysis confirmed the results. In the subgroup analysis, metformin use was associated with the lower incidence of ARDS in females. Conclusions Metformin may have potential benefits in reducing the incidence of ARDS in patients with COVID-19 and type 2 diabetes. However, this benefit differs significantly by gender.

As an emerging infectious disease, the Coronavirus disease 2019 has 64 affected 216 countries and regions worldwide [1] , posing a significant threat to human 7 86 Therefore, it is essential to effective inhibition of the cytokine storm in the treatment of 87 patients with COVID-19. Considering the anti-inflammatory and immunomodulatory 88 of metformin, it has been proposed as a candidate for host-directed therapy of COVID-89 19 [13] . Metformin may have potential benefits in reducing mortality in COVID-19 90 patients with diabetes [14] , but high-level clinical evidence to prove the efficacy was 91 lacking. 92 We conducted a multi-center retrospective cohort study to analyze the effect of Huazhong University of Science and Technology. The requirement for informed 100 consent was waived by the Ethics Committee. Only pseudonymized data with no risk 101 of identification were used for our analyses. 8 severe COVID-19 if they met any of the following criteria: respiratory rate (RR) ≥30 111 breaths/min; pulse oxygen saturation (SpO2) ≤93%; more than 50% lesion 112 progression in lung imaging within 24 to 48h; shock; acute organ failure; or death. 113 The severity of COVID-19 was assessed on the basis of the condition of patients 114 within 48h of admission to hospital. Diabetes was diagnosed according the criteria of 115 WHO that fasting plasma glucose≥ 7.0mmol/L or two hour postprandial glucose≥ 116 11.1mmol/L [17] . ARDS was diagnosed according to the Berlin Definition [18] . 117 We used the following inclusion and exclusion criteria to determine the study 118 cohort. The inclusion criteria contained: 1) patients were diagnosed as admitted to the four hospitals from December 31 st , 2019 to March 31 st , 2020; 2) 120 diagnosed with diabetes and/or had medical history of diabetes; 3) aged more than 18 121 years old;. The exclusion criteria included type 1 diabetes, pregnancy, greater than stage 122 4 renal insufficency, acute heart failure, acute liver failure, and patients without using 123 antdiabetic drugs during hospitalization. The demographic information (age and gender), clinical symptoms (fever, dyspnea, and 126 asthma), comorbidities (Charlson comorbidity index and coronary heart disease The exposure of this study was defined as receiving metformin therapy during 140 hospitalization, and the initial treatment dose is at least 500 mg daily (Table S5) ; 141 patients in this group were classified as metformin group. Patients who receive other 142 antidiabetic drugs other than metformin were classified as the non-metformin group.

The primary outcome of the study was 30-day all-cause mortality. For patients with the 144 length of stay less than or equal to 30 days, we determined the patients as survivor or 145 non-survivor based on discharge diagnosis. The secondary outcome was the incidence 146 of ARDS during hospitalization. In the subgroup analysis, we further explored the 147 impact of metformin use on the outcomes disaggregated by gender. substantially between participants with observed data and those with imputed data (Table S1 ).

The risk of outcomes was calculated by the Cox proportional hazard model if the Kaplan-Meier method. A two-side α less than 0.05 was considered statistically different.

Analyses were performed in SAS 9.4 (by SAS Institute Inc., Cary, NC, USA) and R- (Table 1) .

After propensity score matching, the two cohorts were balanced with no 200 significant difference existed (Table 1 ). The characteristics on admission for male and 201 female were shown in Table S2 . The most common in-hospital complication was ARDS 202 (15.55%), followed by heart failure (3.35%) (Tables S3). 

The incidence of ARDS in the metformin group was significantly lower than that in 205 the non-metformin group (8.0% [8/100] vs 19.1% [43/228]; P=0.0175). In the mixed-effect model, patients received metformin during hospitalization was associated with 207 a lower risk of ARDS incidence compared to the non-metformin users (adjusted OR, 208 0.18; 95%CI, 0.05-0.62; P=0.0070) ( Table 2) . 209 We further performed a propensity score-matched analysis to avoid confounding Table 2) .

In the subgroup analysis, we explored the effect of metformin use on the 216 incidence of ARDS by gender. The results showed that metformin use was 217 significantly associated with lower incidence of ARDS in females (adjusted OR, 0.13; 218 95%CI, 0.02-0.80; P=0.0276), but not in males (adjusted OR, 0.21; 95%CI, 0.03-1.47; 219 P=0.1150) ( Table 3) . The 30-day all-cause mortality was 3.0% (3/100) and 11.0% (25/228) in the 222 metformin and non-metformin groups, respectively (P=0.0175). The main cause of 223 death was ARDS, which accounted for 64.29% of all deaths (Table S4) . Meier analyses showed that metformin users had lower mortality than those in the 225 non-metformin group, but there were no significant differences in mortality between 226 metformin and non-metformin groups before and after propensity score matching 227 ( Figure 2 ).

In the mix-effect Cox model, metformin users was not significantly associated 229 with a lower risk of 30-day all-cause mortalitycompared with the non-metformin group (adjusted HR, 0.48; 95%CI, 0.13-1.74; P=0.2635) . The result was consistent in 231 the propensity score-matched analysis and in the subgroup analysis (Table 2; Table   232 3).

The results of this study showed that the use of metformin while hospitalized 235 significantly reduced the risk of ARDS incidence in patients with COVID-19 and type 236 2 diabetes, especially in the females. However, there was no significant difference in 237 30-day all-cause mortality between the metformin and non-metformin groups.

ARDS is one of the most common complications in patients with .

It is of great significance to prevent the incidence of ARDS for improving the outcome 240 of patients [19] . We found that metformin treatment was significantly associated with We also found that there were differences in the effect of metformin on the disease 262 outcome of COVID-19 patients with type 2 diabetes between male and female.

Metformin use was associated with a lower incidence of ARDS in female, but this 264 association has not been observed in male. Studies showed that metformin could inhibit . It may be one of the reasons that metformin use was associated with a 270 lower incidence of ARDS in the female of this study.

Our study was based on a multicenter design to explore the association between 272 metformin use and unfavorable outcomes of COVID-19 patients with type 2 diabetes. 273 We further explored the gender difference in the effects of metformin on clinical 274 outcomes. This study also has some limitations. First of all, though data regarding 275 HbA1c was missing in 56.2% of this study, FBG was adjusted in the multivariate 276 analysis. Second, since the nature of the retrospective study, the missing data on 277 metformin treatment prior to hospitalization and duration of diabetes may cause some 278 biases of the research results. But we used multiple imputation to adjust for these missing data and made the best use of the existing information in our analysis. Third, 280 some previous studies used mechanical ventilation [31-33] and ICU admission as 281 outcomes [31, 34, 35] , but due to lack of data on the two variables, we failed to explore 282 the effect of metformin on these outcomes. Finally, the secondary outcome variable 283 ARDS in this study was obtained by discharge diagnosis, it is hard to classify according 284 to the Berlin classification. 

Patients with COVID-19 and type 2 diabetes in the metformin group had a lower 287 incidence of ARDS than those in the non-metformin group, especially in females.

However, the association between metformin use and lower mortality was not 289 significant. Given that patients can benefit from metformin therapy, it is recommended 290 that patients with COVID-19 and type 2 diabetes continue to use metformin in the 291 absence of obvious contraindications. 

Data are n (%) or median (IQR)

ALT, 438 alanine aminotransferase; FBG, fasting blood glucose; HbA1c

ARDS, acute respiratory distress syndrome