Between March 4 and June 15, 2020, 2041 patients were flagged in the electronic health record with a diagnosis of COVID-19 and pneumonia. A total of 539 patients were excluded based on eligibility criteria (< 18 years of age, pregnant, received other formulations of corticosteroids, or hospitalized for less than 2 days) Thus, 1072 patients had their data abstracted. A propensity score matched sample was constructed out of 759 patients (381 in NMP and 378 in MP). (Table 1) An examination of the proportional hazard assumption, MP and Fractional inspired oxygen (FiO2) significantly violated it (both with P<0.0001). Data on P/F ratio was lacking; and FiO2 was used since 95% of patients had this data. The supremum test also indicated that non-proportionality was observed in other variables such as nursing home, lack of taste or smell, WBC<11,000 cells/ml, creatinine>1.5 ng/mL, respiratory rate >22 bpm, hydroxychloroquine (HCQ), MP, HD or LD MP, calcium, and initial diastolic blood pressure. All variables with non-proportional hazard were adjusted using FiO2, as indicated above. The Youden Index method yielded a MP dose cut- off 1.36 mg/kg/day. Low dose methylprednisolone (LDMP) was defined as < 1.36 mg/kg/day and high dose methylprednisolone (HDMP) was defined as > 1.36 mg/kg/day. 215 received LDMP and 164 received HDMP. There were 102 Blacks, 338 Whites, 61 Asian/Indians, 251 Non-white / Non-black Hispanics and 7 unknown racial/ethnic group. For each racial/ethnic groups, there was no significant difference between no methylprednisolone and methylprednisolone. (Table 2)
There was a statistically significant difference in age between all 4 groups (Kruskal-Wallis P<.0001). This difference was driven by the difference between Black and White patients (median age: 64 vs 68 years; P=0.0060), between White and Hispanic (median age: 68 vs 60 years; P=0.0005) and between Asian/Indian and Hispanic (median age: 69 vs 60 years; P= 0.0010), after adjusting for multiple tests using Hochberg method. The difference in age between Black and Hispanic patient was trending towards significance (64 vs 60; P=0.0846). Out of 102 Black COVID patients, 25 expired and the difference in age between survivors and non-survivors was significant (median: 60 vs 70 years, P=0.0002). 117 of the 338 White patients expired. The age in the patients that survived and those who did not survive was significantly different (median: 66 vs 74 years, P<0.0001). 23 of 61 Asian/Indian patients expired in the hospital. The age in patients that survived and patients that did not survive was significantly different (median: 64 vs 79 years, P=0.0001). 76 of 251 Hispanic patients expired in the hospital. The age of patients who survivor and patients who did not survive was significantly different (median: 55 vs 74 years, P<0.0001). [i.e. the median age non-survivors were clinically higher than the age in survivors in all other race/ethnicity groups].
For ferritin, the figure shows slightly elevated levels in the Asian/Indians (median=1,048 ug/L) compared to Blacks (median=743.0 ug/L), Whites (median=777.5 ug/L) and Hispanics (median=831.0ug/L), however, this did not translate into statistically significant difference between the groups (Kruskal-Wallis P=0.1689). (Figure 1) In Blacks with ferritin values (n=72), the difference in ferritin between survivors (n=21) and non-survivors (n=51) was not significant (median: 745 vs 575 ug/L, P=0.9280). In Whites with ferritin values (n=338), the difference between survivors (n=243) and non-survivors (n=95) was significant (median: 641 vs 987 ug/L, P=0.0139). In Asian/Indians with ferritin values (n=56), the difference between survivors (n=38) and non-survivors (n=18) was significant (median: 1,265 vs 418 ug/L, P=0.0211). In Hispanics with ferritin values (n=226), the difference between survivors (n=158) and non- survivors (n=68) was not significant (median: 826.5 vs 897.5 years, P=0.0854).
For d-dimer, there were slightly elevated levels in Blacks (median=1.64 mcg/mL) compared to Whites (median=1.33 mcg/mL), Asian/Indians(median=1.21mcg/mL) and Hispanics (median=1.11 mcg/mL), however this did not translate into statistically significant difference between the groups (Kruskal-Wallis P=0.1240). In Blacks with d-dimer values (n=62), the difference between survivors (n=44) and non-survivors (n=18) was not significant (median: 1.91 mcg/mL vs 1.53 mcg/mL, P=0.9119). In Whites with d-dimer values (n=193), the difference between survivors (n=133) and non-survivors (n=60) was significant (median: 1.19 vs 1.76 mcgg/mL, P=0.0034). In Asian/Indians with d-dimer values (n=41), the difference between survivors (n=26) and patients that non-survivors (n=15) was trending towards significance (median: 1.10 vs. 1.51 mcg/mL, P=0.0784). In Hispanics with d-dimer values (n=172), the difference between survivors (n=116) and non-survivors (n=56) was significant (median: 0.89 vs 2.24 mcg/mL, P<0.0001). The d-dimer in non-survivors was significantly or trending higher than that in survivors in White, Asian/Indian and Hispanic groups while the order was reversed in the Black group; the median d-dimer in survivors was higher than the level in the non-survivors, albeit, not achieving statistical significance.
For CRP, the figure shows slightly elevated levels in Blacks (median=11.40 mg/mL), moderate to extremely elevated Asian/Indians (median=12.85 mg/mL) and Hispanics (median=14.32 mg/mL), compared to Whites (median=9.47 mg/mL), and this yielded a statistically significant difference between the groups (Kruskal-Wallis P<0.0001). This result was driven by the significant difference between Whites and Hispanics adjusting for multiple testing using Hochberg method. In Blacks with CRP values (n=85), the difference between survivors (n=63) and non-survivors (n=22) was not significant (median: 10.2 mg/mL vs 12.07 mg/mL, P=0.3586). In Whites with CRP values (n=276), the difference between survivors (n=183) and non-survivors (n=93) was significantly different (median: 7.71 vs 12.30 mg/mL, P=0.0004). In Asian/Indians (n=60), the difference between survivors (n=38) and non-survivors (n=22) was significantly different (median: 9.05 vs. 20.75 mg/mL, P<0.0001). In Hispanics with CRP values (n=228), the difference between survivors (n=158) and non-survivors (n=70) was trended towards being significantly different (median: 13.70 vs 16.70 mg/mL, P=0.0514).
For IL-6, there was no significant difference between the race/ethnicity groups (Kruskal- Wallis P=0.9843). Except for Asian/Indians with median IL-6 of 11 pg/ml, all of the other groups had a median IL-6 of 14.0 pg/ml. Blacks with IL-6 values (n=51), the difference between survivors (n=34) and non-survivors (n=17) was not significant(P=0.7868). Whites with IL-6 values (n=157), the difference between survivors (n=94) and non-survivors (n=63) was significant (median=11.0 pg/ml vs 27.0 pg/ml P=0.0005). In Asian/Indians with IL-6 values (n=65), the difference between survivors (n=42) and non-survivors (n=23) was trending towards significance(median 30.0 pg/ml vs 8.0 pg/ml P=0.0924). In Hispanics with IL-6 values (n=110) the difference between survivors (n=73) and non-survivors (n=37) was significant (median=19.0 pg/ml vs 11.0 pg/ml P=0.0168).
Overall, in hospital survival was significantly difference between the race/ethnicity groups (Wilcoxon P=0.0320). This result was driven by the significant difference between Whites and Blacks (P=0.0249) and Whites and Asian/Indian (P=0.0463), after adjusting for multiple testing. The 30-day in hospital survival rates amongst Asia/Indian, Blacks, Hispanics, and Whites were 54.4% (95% CI 35.6 to 72.5%), 37.9% (95% CI 19.8% to 57.8%), 33.4% (95%CI 22.9 to 44.7%), and 41.0% (95% CI 32.6 to 49.6%), respectively. These differences become less significant after 30 days. (Figure 2).
LDMP was associated with prolonged IhS compared to NMP in Whites (0.350 95% CI 0.217- 0.566 (P<0.0001), Asian/Indians (HR 0.083 95% CI 0.011-0.653)(P=0.0180), and Hispanics (HR 0.294 95% CI 0.149-0.581)(P0.0004). (Table 3) HDMP was associated with prolonged survival compared to NMP in Asian/Indian, and White but there was no significant difference between LDMP and HDMP (P<0.05). HDMP was not associated with prolonged IhS in Blacks [HR 1.336 (95% CI 0.455-3.926)(P 0.5979)] and Hispanics [HR 0.594 (95% CI 0.329- 1.075)(P=0.0851)]. HDMP was associated with prolonged IhS compared to NMP in Whites (HR 0.557 95% CI 0.353-0.881)(P=0.0122) and Asian/Indians (HR 0.294(95% CI 0.149- 0.581)(P=0.0004). There was no association with prolonged IhS in Blacks for LDMP (HR 0.612 95% CI 9.264-1.416)(P=0.2514) or HDMP (HR 1.336 95% CI 0.455-3.926)(P=.597). HDMP was associated with shorter survival than LDMP in Hispanics (HR 1.975 95% CI 1.123-3.473) (P=0.0181).
Due to the varying frequencies, we were not able to compare the association of NMP, LDMP, and HDMP between all racial/ethnic groups. We did compare NMP, LDMP, and HDMP and their association with survival between Whites and Hispanics and they were not significant. (P> 0.05). (Figures S1-S3)