Early Lopinavir/ritonavir does not reduce mortality in COVID-19 patients: Results of a large multicenter study

Abstract

The COVID-19 pandemic, an unprecedented event for current generations of physicians, has stricken hard on society. 1 There is a significant lack of effective drugs for stopping viral replication. Lopinavir/ritonavir (LPV/r) is a well-known combination used in patients with HIV which was included in the arsenal against SARS-CoV-2 early in the pandemic. 2 Its use in COVID-19 was based on inconsistent results from experimental and clinical research that was mostly done while investigating other β-coronaviruses (SARS and MERS). A number of randomized clinical trials have observed no benefit of LPV/r beyond the standard of care.3, 4, 5 However, voices have been raised against interpreting these results as grounds from definitively ruling out LPV/r since some of these studies lacked statistical power, reported encouraging outcomes in secondary endpoints, and included patients with a prolonged period of symptoms before initiation of treatment. 6 , 7 Indeed, there may be a subpopulation of COVID-19 patients – notably those early in the course of the infection – for whom LPV/r may improve their prognosis. In a recent report, Klement-Frutos et al. describe a favorable outcome of patient with COVID-19 after beginning of LPV/r on day 9 of symptoms. 8 Therefore, we aimed to assess the efficacy of LPV/r in a large, multicenter cohort of patients, with special interest in those who received treatment soon after the onset of symptoms. This work belongs to the SEMI-COVID-19 Registry, which is an ongoing, nationwide, retrospective, anonymized cohort of consecutive adult patients hospitalized in Spain for microbiologically confirmed COVID-19. 9 The Registry was approved by the Ethics Committees of the participating centers, and included data on over 300 variables. The primary endpoint was raw-in hospital mortality at 30 days from admission. Patients were considered to have been treated with LPV/r if they had received at least one dose of the drug. Common dosage of LPV/r was 400/100 mg bid. In order to mitigate the effects of possible confounding variables in a non-randomized assessment of treatment with LPV/r, propensity score (PS) was performed. The propensity of receiving LPV/r was estimated using a logistic regression model that included confounding variables which could have affected treatment choice or outcomes as independent variables. The nearest neighbor method with a caliper of 0.1 as used in PS matching and standardized mean differences (SMD) were calculated to evaluate adequacy of propensity matching. Both conditional logit and mixed effects logistic regressions were performed. Furthermore, univariate and multivariate logistic regression models were fitted in order to estimate the treatment effect using all data, as a sensitivity analysis. Multiple imputation was used to handle missing data and model estimates and standard errors were calculated using Rubin's rules. 10 Statistical analyses were performed using R software (v.3.6.2). As of June 1, 2020, the Registry included 9,594 cases, of which 8,553 met the inclusion criteria (Suppl. Fig. 1). Fifty-seven percent were men, median age was 69 years (IQR 56–79), and half of subjects (50.2%) had high blood pressure. Patients were admitted after a median time since symptoms onset of 7 days (IQR 4–9), with median SaO2/FiO2 ratio of 376 (IQR 300–452), C-reactive protein of 58 mg/L (IQR 19–123), and lymphocytes of 940 cells/µL. LPV/r was administered to 5,396 patients (63%) after a median of 0 days since admission (IQR 0–1). Table 1 shows that LPV/r was more likely to be prescribed to patients who presented with more severe clinical condition, including presence of fever, cough, radiological infiltrates (91.7% vs 80.4% p<0.001) and a lower SaO2/FiO2 ratio. On the other hand, LPV/r was less frequent among at-risk subjects in whom toxicity may be more likely: elderly patients presenting with altered mental status, dementia, or other debilitating baseline conditions, as well as patients on immunosuppressive drugs and pregnant women. Table 1 – Baseline characteristics and clinical presentation of all patients according to the administration of LPV/r. Table 1 Unmatched data (n = 8,553) Matched data (n = 5,068) No LPV/r (n = 3,157) LPV/r (n = 5,396) p No LPV/r (n = 2,534) LPV/r (n = 2,534) p Baseline features  Age (years)† 74.9 [60.4;85.2] 66.2 [54.3;75.8] <0.001 70.8 [57.2;81.5] 69.5 [56.6;78.7] <0.001  Sex (female) 1511 (47.9%) 2155 (39.9%) <0.001 1124 (44.4%) 1083 (42.7%) 0.257  Pregnancy 22 (0.70%) 15 (0.28%) 0.008 18 (0.71%) 13 (0.51%) 0.471  Race (Caucasian) 2850 (91.2%) 4647 (87.9%) <0.001 2268 (89.5%) 2279 (89.9%) 0.644  Charlson Comorbidity Index 1.00 [0.00;2.00] 0.00 [0.00;1.00] <0.001 1.00 [0.00;2.00] 1 [0.00;2.00] 0.626  High blood pressure 1817 (57.7%) 2480 (46.1%) <0.001 1357 (53.6%) 1288 (50.8%) 0.056  Immunosuppressive therapy 257 (8.14%) 291 (5.39%) <0.001 205 (8.09%) 186 (7.34%) 0.040  Dementia 628 (20.0%) 199 (3.70%) <0.001 171 (6.75%) 179 (7.06%) 0.009 Clinical presentation  Duration of symptoms (days) † 6.00 [3.00;9.00] 7.00 [4.00;9.00] <0.001 6.00 [3.00;9.00] 7.00 [4.00;9.00] 0.337  Cough 2170 (69.2%) 4281 (79.6%) <0.001 1892 (74.7%) 1918 (75.7%) 0.678  Dyspnea 1787 (56.9%) 3108 (57.9%) 0.370 1439 (56.8%) 1440 (56.8%) 1.000  Altered mental status 574 (18.4%) 351 (6.58%) <0.001 262 (10.3%) 249 (9.83%) 0.576  Temperature 36.9 [36.3;37.6] 37.1 [36.4;37.9] <0.001 36.9 [36.3;37.7] 37.0 [36.4;37.8] 0.013  Heart rate (beats/min) † 86.0 [75.0;98.0] 88.0 [77.0;100] <0.001 86.5 [76.0;99.0] 87.0 [76.0;99.0] 0.816  Respiratory rate > 20 breaths/min 958 (31.2%) 1558 (29.7%) 0.155 728 (28.7%) 735 (29.0%) 0.852 Laboratory  SaO2/FiO2 † 387 [304;452] 372 [297;452] <0.001 392 [307;457] 381 [304;452] 0.008  Lymphocytes (cells/µL) † 990 [700;1330] 910 [700;1260] 0.080 1000 [700;1320] 940 [700;1300] 0.042  C-reactive protein (mg/L) † 55.0 [17.0;120] 59.9 [19.8;125] 0.002 55.7 [17.1;120] 55.9 [18.0;119] 0.391  Creatinine (mg/dL) † 0.93 [0.75;1.23] 0.89 [0.73;1.10] <0.001 0.90 [0.74;1.17] 0.90 [0.74;1.14] 0.718 Other treatments  Interferon-β 52 (1.65%) 1073 (20.1%) <0.001 52 (2.05%) 78 (3.08%) 0.049  Hydroxychloroquine 2299 (72.8%) 4893 (90.7%) <0.001 2084 (82.2%) 2133 (84.2%) 0.071  Remdesivir 17 (0.54%) 26 (0.48%) 0.842 12 (0.47%) 12 (0.47%) 1.000 Mortality 688 (22.8%) 821 (15.7%) <0.001 431 (17.0%) 406 (16.0%) 0.364 † Continuous variables expressed as median and [interquartile range]. LPV/r: lopinavir/ritonavir. Overall, 1,509 patients died (17.6%). The univariate parameters predicting mortality is shown in Table 2 . A PS allowed for comparing two cohorts with similar values on the parameters associated with the prescription of LPV/r (Table 1). Most parameters were adequately matched according to SMD, although some variables had SMD values >0.02 (Suppl. Table 1): In this matched cohort, the adjusted odds ratio (aOR) for mortality for the use of LPV/r was 0.932 (95CI 0.799–1.087; p>0.05) according to both conditional and mixed effects logistic models. Table 2 – Univariate analysis of mortality in all patients and those with duration of symptoms of less than 8 days. Table 2 – All (n = 8,553) Early cohort (n = 6,099) OR (95CI) P OR (95CI) p Sex (female) 0.783 (0.698–0.879) <0.001 0.775 (0.681–0.882) <0.001 Age (per year) 1.090 (1.084–1.096) <0.001 1.082 (1.076–1.089) <0.001 Race (Caucasian) 3.267 (2.499–4.272) <0.001 3.715 (2.734–5.047) <0.001 High blood pressure 2.833 (2.509–3.200) <0.001 2.700 (2.354–3.097) <0.001 Dementia 4.323 (3.715–5.031) <0.001 3.704 (3.137–4.374) <0.001 Immunosuppressive treatment 1.747 (1.431–2.134) <0.001 1.656 (1.328–2.066) <0.001 Charlson Comorbidity Index (per point) 1.334 (1.296–1.373) <0.001 1.307 (1.265–1.349) <0.001 Duration of symptoms (per day) 0.930 (0.918–0.943) <0.001 0.907 (0.886–0.928) <0.001 Cough 0.652 (0.576–0738) <0.001 0.703 (0.612–0.806) <0.001 Temperature (per °C) 1.088 (1.026–1.154) 0.005 1.069 (1.000–1.141) 0.049 Dyspnea 1.865 (1.653–2.104) <0.001 1.828 (1.597–2.091) <0.001 Confusion 5.341 (4.620–6.174) <0.001 4.433 (3.774–5.208) <0.001 Respiratory rate > 20 breaths/min 3.396 (3.020–3.819) <0.001 3.239 (2.838–3.696) <0.001 SaO2/FiO2 (per unit) 0.993 (0.992–0.993) <0.001 0.993 (0.992–0.993) <0.001 Lymphocytes (per cells x 103/µL) 1.000 (1.000–1.000) 0.424 1.000 (1.000–1.000) 0.470 C-reactive protein (per mg/L) 1.005 (1.005–1.006) <0.001 1.006 (1.005–1.006) <0.001 Creatinine (per mg/dL) 1.789 (1.651–1.938) <0.001 1.647 (1.515–1.792) <0.001 Treatment with Hydroxychloroquine 0.447 (0.391–0.512) <0.001 0.483 (0.416–0.562) <0.001 Treatment with Interferon-β 2.094 (1.813–2.420) <0.001 2.098 (1.787–2.463) <0.001 Treatment with Remdesivir 1.096 (0.507–2.366) 0.816 0.871 (0.360–2.109) 0.760 Treatment with LPV/r 0.651 (0.581–0.729) <0.001 0.705 (0.620–0.801) <0.001 LPV/r: lopinavir/ritonavir. OR: odds ratio. 95CI: 95% confidence interval. Of the 6,099 patients who were admitted to hospital within 8 days since onset of symptoms (median time to admission since onset of symptoms 5 days [IQR 3–7]), LPV/r was prescribed to 3,377 (55%). Variables associated with the use of LPV/r were similar to those observed in the cohort as a whole (Suppl. Table 2). In a propensity score matching carried out on this subset of patients, early use of LPV/r was not associated with a lower mortality (conditional logistic regression: aOR 1.110 (95CI 0.944–1.300; p = 0.245); mixed effects logistic regression: aOR 1.105 (95CI 0.944–1.300; p = 0.272)). Consistent with previous studies, our analysis found no overall benefit to the use of LPV/r.3, 4, 5 We have focused on patients who received the antiviral at an earlier stage in the hope of finding greater activity. Indeed, in other viral diseases, the administration of antiviral drugs must be done as soon as possible in order to have a clinically significant activity. 11 Of note, patients included in Rao's clinical trial had a median duration of symptoms of 13 days (IQR 11–16) 4 , and those recruited in the RECOVERY trial presented after 8 days of disease (IQR 4–12). 3 In our sub-analysis, the median duration was 5 days (IQR 3–7), thus allowing us to perform a evaluation on patients who were indeed at a very early stage of disease. However, results were again disappointing, and add another nail in the coffin of LPV/r when considering its use for COVID-19. Our study has some limitations. First, it has the biases inherent to retrospective observational studies. Also, despite the fact that the number of patients included allowed us to perform PS matching, which may have reasonably controlled for many of these biases, the balance of some parameters was not perfect according to SMD values. Still, as a multicenter study involving a large number of hospitals, it has the strength of being rooted in real-life practice, far from strict inclusion and exclusion criteria of clinical trials. Second, we have used mortality as a primary endpoint, as others have done, but we cannot rule out any benefits of LPV/r that would have emerged had we analyzed softer outcomes, such as time to improvement or disease duration, as suggested by the report of Klement-Frutos et al. 8 Finally, our analysis has used data from COVID-19 first wave in Spain, when efficacy of corticosteroids or other drugs was not yet proved. Thus, our analysis is not adjusted for these treatments. However, these therapies, at least during the first wave of the pandemic, have usually been reserved for severe patients, and thus may be surrogate predictors of unfavorable progress. In conclusion, we have analyzed a large, multicenter cohort of patients with COVID-19 and have not found any benefits to administering LPV/r, even when it was administered within the first 8 days of symptoms. Our results discourage its use in SARS-CoV-2 infection. Funding This work was supported by the Spanish Society of Internal Medicine (SEMI).