Introduction

The outbreak originated by Severe Acute Respiratory Syndrome due to Coronavirus-2 (SARS-CoV-2) started at the end of 2019 in Wuhan, China. The World Health Organization declared it a global pandemic in March 20201. At the time this article was written, a new SARS-CoV-2 variant concerns the scientific and health authorities2. As a result, it has been observed an increase in infections and hospitalizations due to COVID-19 across worldwide3,4,5 and some World Health Organization (WHO) regions such as Western Pacific Regions have experienced an increase in the number of deaths6. The new COVID-19 waves as well as the potential mutations of the virus represent a new risk for the general population and particularly for those at high risk7.

It has been suggested that the psychosocial effects of SARS-CoV-2 on the population will lead to an increase in the use of medication in treating mental health conditions, thus the potential effects of these medications should be considered in the light of COVID-19 treatment8. A risen in the prescription of antipsychotics in at-risk populations after the pandemic caused by COVID-19 has been recently described9.

Controversial results have been published on the potential role of antipsychotics in COVID-19. While some studies pointed out possible adverse effects10,11, others have reported protective effects against infection and prognosis of the disease caused by SARS-CoV-212,13. During SARS-CoV-2 infection multiple inflammatory pathways are activated that promotes secretion of proinflammatory cytokines14. Antipsychotics exert anti-inflammatory effects via reducing proinflammatory cytokines production, modulating monocytes response through Toll-Like Receptors (TLR) and inhibiting microglial activation15,16,17. In humans, the immunomodulatory effect of risperidone and aripiprazole has been demonstrated18, with aripiprazole demonstrating a greater anti-inflammatory effect on TNF-α, IL-13, IL-17α and fractalkine. Nevertheless, important limitations have been highlighted in these studies such as the need to explore specific types of antipsychotics, doses, or the inclusion of relevant covariables in the analyses19. In fact, it has been emphasized the necessity of combining experimental and clinical studies to elucidate the relevance and repercussions that could have the use of antipsychotics during COVID-19 era20.

The aim of this study was to investigate if antipsychotic treatments have a protective or adverse impact on COVID-19 related deaths. We leveraged the Andalusian Public Health System's Health Population Base (BPS), a vast electronic health record (EHR) resource established in 2001. Encompassing over 13 million users, the BPS represents one of the world's most extensive clinical data repositories21. The BPS's size and comprehensiveness create an exceptional environment for conducting large-scale real-world evidence (RWE) studies. Based on our previous studies22,23 we hypothesized that persons under aripiprazole treatment will show reductions in the risk of death by COVID-19.

Materials

Study design

The study was approved by the Ethics Committee for the Coordination of Biomedical Research in Andalusia (29 September 2020, Acta 09/20) and by the Local Ethics Committee of Virgen del Rocío University Hospital (PI-2578-N-20). Informed consent was not required for the secondary use of clinical data for research. For secure real-world data analysis, the study utilized the Infrastructure for secure generation of Real-World Evidence from Real World Data from the Andalusian Health Population Database (iRWD) managed by the Foundation for Progress and Health of the Andalusian Public Health System. The data was securely transferred from the BPS to the iRWD, thereby ensuring the safety and integrity of the information throughout the process.

The study consists of a retrospective cohort of Andalusian patients hospitalized with a COVID-19 diagnosis from January 2020 to November 2020, as registered in the BPS. The sample (n = 2.536) consisted of inpatients who met the following criteria: age of 18 years or older, having a laboratory-confirmed COVID-19 infection (either PCR or antigen test) and who were either not prescribed any medication or were prescribed an antipsychotic during the 15 days prior to the hospitalization event. We identified 15,968 inpatients who met the inclusion criteria before selecting the antipsychotic-treated and untreated subcohort.

Covariate, event, and endpoint definition

To reduce the high dimensionality induced by the ICD codes we grouped the conditions as: obesity and other associated conditions (E66), diabetes mellitus (E11), circulatory (I00–I99), respiratory (J00–J99), neoplasms (C00–D49), dementia (F00–F03), anxiety or mood disorders (F30–F48), and other mental diseases (F04–F29 and F50–F99). In addition, we obtained the following data for each patient: sex, age categorized as [18, 41), [41, 68) and [68, 99), flu vaccination status, and pneumococcal vaccination status. This categorization follows24 and25. Finally, we used death from COVID-19 as the primary outcome, defined as the endpoint by a certified death event during the first 30 days of a COVID-19 hospitalization, based on26.

Treatment definition

The seven antipsychotics examined in this analysis are Quetiapine, Sulpiride, Paliperidone, Aripiprazole, Haloperidol, Olanzapine, and Risperidone (all the antipsychotics found in the population under study). A patient was considered to be under a specific treatment if they received a dispensation of that treatment between 15 days prior to the index date and the index date itself. In addition to antipsychotic medications, we investigated whether each patient had received any anticoagulant medications during the same timeframe (between 15 days before and the index date). To reduce the dimensionality introduced by the variety of anti-inflammatory treatments used by the cohort, we categorized patients into two groups: those receiving any anti-inflammatory medication and those not receiving any (See Supplementary Material).

General description of the population under study

Table 1 shows the characteristics of the cohort studied. The cohort is divided into those who died and those who survived. The table reveals several key differences between the two groups. Notably, the deceased group had a significantly higher average age (79.09 years) compared to survivors (62.89 years). Vaccination rates were also higher among deceased patients, with a greater percentage having received flu and pneumococcal vaccinations. Perhaps most significantly, the table suggests a higher prevalence of underlying medical conditions in the deceased group, including diabetes, circulatory diseases, neoplasms, respiratory diseases, and dementia. Interestingly, the data also shows that Quetiapine (DB01224), an antipsychotic medication, was used more frequently by deceased patients (22.4%) compared to survivors (15.1%), whereas out of the 31 patients that were using Aripiprazole only 1 died, highlighting a potential difference in the impact of these medications on mortality risk. As expected, Table 1 shows a cohort where deceased patients were older, had a higher prevalence of vaccinations and underlying health conditions, and were more likely to have used Quetiapine.

Table 1 Covariate associations with end point: Chi-squared tests, p-values, counts, and proportions.
Full size table

Drug processing

Drug prescription data of Andalusian public health care patients were received in three separate data dumps from the BPS database: the first one on Nov 17, 2020 with 12,009 prescription entries containing any one of the seven antipsychotics, and the second and third ones on Feb 15, 2021 with 13,413 and 19,391 entries, respectively. The data sets were combined into one and parsed as follows: each drug prescription entry was broken down into its active constituent agents, and for each patient, each of their prescription date and each active agent the following fields were extracted:

  • Encrypted NUHSA code (patient identifier)

  • Description of the prescribed medication

  • Active agent

  • Amount of agent per unit (e.g. per capsule, tablet, solution, or injection)

  • Dose/mode of administration

  • The total amount administered by prescription

  • Prescription date

In the case of the seven antipsychotics at hand, there is only one active ingredient per prescription, which is the antipsychotic itself, so the number of entries/rows before and after parsing remains the same. Duplicate entries, which might have resulted from overlaps in the data dumps, were removed. The resulting data set contained 33,348 entries.

Next, for each entry the number of days between the prescription date and that of the subsequent entry (same patient, active agent), if there was any, was calculated. Based on this number of days between two given subscriptions and the total amount of active agents administered in the preceding prescription, which had previously been determined, the average daily dose of the preceding prescription was calculated. This was done for all prescriptions followed by a subsequent prescription. Finally, for each patient and each active ingredient occurring in their prescriptions, an average over all the average daily doses was calculated. So, the following three fields were added to the data set:

  • Days to next prescription

  • The average daily dose of prescription

  • Total average of all average doses (patient-wise, per active agent)

Drug information is summarized in Table 2.

Table 2 Antipsychotic type, mode of administration and dosage.
Full size table

Methods

The study procedures were in accordance with local legislation and the Declaration of Helsinki27. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies28.

To elucidate if any given treatment could potentially reduce the mortality in COVID-19 inpatients we conducted three statistical tests that consider covariates that present an a priori possibility of confounding the association between a treatment and the odds of surviving COVID-19 disease (sex, age, co-morbidities, anti-inflammatory medication use, and influenza and pneumococcal vaccination status):

  • The odds outcome was estimated using a general binomial linear model weighted using the inverse probability of treatment weighting (IPTW) technique, with the weights computed by means of a logistic regression model and adjusted for estimating the average treatment effect on the treated population (ATT), conditioned to the confounders of interest using the whole cohort, using the untreated population as the reference.

  • In addition, we conducted a similar analysis using the untreated group as the reference. In this case, the covariate adjustment was carried out by estimating the average treatment effect on the population (ATE).

  • Finally, we obtained the Kaplan–Meier survival curves for each treatment comparing the untreated versus the treated groups. We provide unadjusted curves along with IPTW covariate-adjusted matched survival curves.

In both cases, we provide FDR-adjusted p values. Significance is achieved at level 0.05, and we provide 95% confidence intervals. To get an accurate measure of the variability of the marginal odds ratio we used heteroskedasticity-consistent standard errors29.

The results have been summarized in Table 3.

Table 3 Covariate-adjusted Odds Ratio Estimates, Confidence Intervals and FDR-adjusted p-values for the antipsychotics tested.
Full size table

Software

To perform the analysis, we used:

  • R version 3.6.3 (2020-02-29)

  • Survival analysis: Survival R package (v 3.2.11)30

  • Weights computation for IPW: WeightIt R package (v 0.12)31

  • General linear models: glm R stats core package32.

  • Survival plots: survminer R package (v 0.4.9)33.

  • Table 1: tableone R package (v 0.13.2)34.

  • HC standard errors computation: Sandwich R package (v 3.0.1)35

Ethical standards

The study procedures were in accordance with local legislation and the Declaration of Helsinki. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies. The protocol was approved by the Local Ethics Committee of Virgen del Rocío University Hospital (PI-2578-N-20). All participants data were anonymized to ensure confidentiality.

Results

Aripiprazole is the only antipsychotic treatment that shows a protective effect, as can be evidenced by the 95% confidence intervals for the covariate-adjusted odds-ratio analysis (left side figure). Find a survival curve plot for each treatment in Supplementary material. This finding is further reinforced by the covariate-adjusted survival analysis on the matched population (right side figure). See Fig. 1.

Figure 1
figure 1

Antipsychotic's impact on patient survival. Left: adjusted odds ratios (with 95% confidence intervals) for all treatments before and after FDR correction, to assess the effect on death after hospitalization. OR below 1 indicate protective effects. Right: Kaplan–Meier survival curves for patients who received Aripiprazole compared to those who didn't, controlling for covariates.

Full size image

Discussion

The main finding derived from this study is that aripiprazole leads to reduction in the risk of death caused by COVID-19. On the contrary, the rest of the antipsychotic treatments included in the analyses did not significantly reduce the mortality related to COVID-19.

Antipsychotics have been shown to exert anti-inflammatory effects through decreased proinflammatory cytokine production, modulation of monocyte response through TLR and inhibition of microglial activation16,17. In fact, Chlorpromazine protects mice from severe clinical disease and SARS-CoV-236. Clozapine an atypical antipsychotic, has been revealed to be effective in suppressing proinflammatory cytokine expression by limiting NLRP3 inflammasome activation in an in vitro model of schizophrenia37.

Our group showed that aripiprazole (marked Phenylpiperazine) reverts the changes caused by COVID-19 in gene expression which could validate aripiprazole as a treatment for COVID-1922. Interestingly, research investigating approximately 12,000 drugs in clinical-stage or Food and Drug Administration (FDA)-approved small molecules to identify candidate drugs to treat COVID-19, reported that Elopiprazole (a never marketed phenylpiperazine antipsychotic drug) was listed among the 21 most potent compounds to inhibit SARS-CoV infection38. A possible explanation for the superiority of aripiprazole, in convergence with a recent publication, is that aripiprazole increases the expression of anti-inflammatory markers compared to other antipsychotics. In particular, aripiprazole induced mTORC1 inhibition, which is an important mechanism of action in microglial cells that leads to an anti-inflammatory shift39.

Aripiprazole has been included into the called Functional Inhibition of acid sphingomyelinase (FIASMA) medications40. Carpinteiro et al. (2020) suggested that acid sphingomyelinase (ASM)/ceramide system plays an important role in the infection of cells with SARS-CoV-2 since its activation by SARS-CoV-2 facilitates viral entry into cells41. FIASMA medications would inhibit ASM and reduce the formation of ceramide-enriched membrane platforms, decreasing cell infection with SARS-CoV-2 and subsequent inflammation42. Among the antipsychotics included in the present study, aripiprazole was the only considered FIASMA medication. Aripiprazole property to inhibit ASM as well as its anti-inflammatory effects could explain the superiority showed by the molecule against fatal outcome derived from COVID-19.

None of the antipsychotics included in our analysis contribute significantly to the increase in mortality related to COVID-19 in the adjusted analyses. As was noted previously, former publications which have reported adverse effects of antipsychotics against COVID-19 have shown significant limitations such as considering antipsychotics as homogeneous group14. In that sense, previous works that have not differentiated between different types of antipsychotics have reported adverse effects on COVID-1943,44,45. However, those studies that have explored potential adverse effects of specific antipsychotics have not reported a worse prognosis of COVID-1912,46,47. These results underline the need to study the possible effects of antipsychotics on a one-to-one basis, as different antipsychotics may have very different effects on each other.

Some limitations should be considered when interpreting the results. Firstly, we have used preprocessed electronic records that could not include sensitive (uncodified) information. Secondly, our treatment information is based on dispensation records, so unless patients are receiving an injectable medication, which is administered at the time of purchase, we cannot be confident that the patient is using the drug. However, we consider that most patients adhere to their prescribed treatment, as the majority present a mean time span between dispensations below 40 days. Finally, aripiprazole resulted to be the least prescribed antipsychotic, which limits the generalizability of the results as well as the performance of complementary analyses that could elucidate which dosage of the drug is more likely to play a protective role for the risk of death. Nevertheless, a strength of the study is that, to the best of our knowledge, this is the first study to examine the mortality of COVID-19 linked to specific antipsychotic drugs, employing a sizable study population for that purpose.

In conclusion, our results exploring a population-based case–control cohort pose the potential usefulness of aripiprazole, in COVID-19 infected individuals with psychiatric disorders. Aripiprazole could play an important role in minimizing the fatal outcomes related to COVID-19. The anti-inflammatory properties of aripiprazole against cytokine storm as well as its FIASMA properties may explain its genuine effects in reducing the risk of death associated with COVID-19.