Observational study of azithromycin in hospitalized patients with COVID-19.

BACKGROUND: The rapid spread of the disease caused by the novel SARS-CoV-2 virus has led to the use of multiple therapeutic agents whose efficacy has not been previously demonstrated. The objective of this study was to analyze whether there is an association between the use of azithromycin and the evolution of the pulmonary disease or the time to discharge, in patients hospitalized with COVID-19.
METHODS: This was an observational study on a cohort of 418 patients admitted to three regional hospitals in Catalonia, Spain. As primary outcomes, we studied the evolution of SAFI ratio (oxygen saturation/fraction of inspired oxygen) in the first 48 hours of treatment and the time to discharge. The results were compared between patients treated and untreated with the study drug through subcohort analyses matched for multiple clinical and prognostic factors, as well as through analysis of non-matched subcohorts, using Cox multivariate models adjusted for prognostic factors.
RESULTS: There were 239 patients treated with azithromycin. Of these, 29 patients treated with azithromycin could be matched with an equivalent number of control patients. In the analysis of these matched subcohorts, SAFI at 48h had no significant changes associated to the use of azithromycin, though azithromycin treatment was associated with a longer time to discharge (10.0 days vs 6.7 days; log rank: p = 0.039). However, in the unmatched cohorts, the increased hospital stay associated to azithromycin use, was no significant after adjustment using Multivariate Cox regression models: hazard ratio 1.45 (IC95%: 0.88-2.41; p = 0.150). This study is limited by its small sample size and its observational nature; despite the strong pairing of the matched subcohorts and the adjustment of the Cox regression for multiple factors, the results may be affected by residual confusion.
CONCLUSIONS: We did not find a clinical benefit associated with the use of azithromycin, in terms of lung function 48 hours after treatment or length of hospital stay.

Introduction

In December 2019, an epidemic outbreak associated with a novel coronavirus (severe acute respiratory syndrome coronavirus 2, SARS-CoV-2) was reported in Wuhan (China) with mainly respiratory clinical manifestations [1]. The extent of the outbreak reached such a magnitude that the WHO declared it a pandemic on March 12, 2020 [2]. Although mortality rates in those affected (approximately 2% among medically treated patients) [3] seem to be overestimated due to underdiagnosis of affected individuals with mild symptoms, the extent of the pandemic has caused the search for effective treatments to become a top priority.

Several pharmacological agents have been proposed as potential treatments based on theoretical considerations, in vitro studies, or clinical trials conducted in conditions caused by related viruses [4-6]. However, current evidence has not confirmed the presence or absence of a benefit of these treatments and even warns of the probable risks or adverse effects associated with their use. Several randomized clinical trials are underway but have not yet been completed or have not been reported [7, 8].

Azithromycin has been considered, usually combined with hydroxychloroquine, based on its in vitro action against other viruses, such as influenza A [9, 10], and its potential immunomodulatory and anti-inflammatory action in other respiratory diseases [11-13]. Randomized clinical trials specifically evaluating the clinical benefit of azithromycin, as an isolated treatment for COVID-19, are in course [14], however, their results have yet not been reported. From observational studies, at the moment this manuscript is being written, only Rosenberg et al. [15] and Arshad et al. [16] have reported results on the specific use of azithromycin in Covid-19 disease. In both studies, among hospitalized patients (n = 1428 and n = 2541, respectively) treated or not with azithromicyn and/or hydroxychloroquine, a benefit on mortality was not found in the group treated with azithromicyn alone. Other studies have evaluated the use of azithromycin, though in combination with hydroxychloroquine [17-21]. In a randomized, open-label and controlled trial by Cavalcanti et al. (n = 667) [20], a benefit on clinical severity was not found in the group of hospitalized patients treated with the combination of hydroxychloroquine plus azithromycin. There are other studies on the combination but with results that are difficult to interpret because they lacked a comparison group [17, 18, 22] or because some of the patients assigned to the treated group received only hydroxychloroquine and not the combination with azithromycin [23, 24].

The absence of an adjusted comparison group in the majority of reported observational studies is relevant given the well-known confounding of the observational design by factors that influence the choice or not of a certain treatment. One of the methods used to balance this has been the propensity score [25], as used in the studies by Magagnoli et al. [19] and Geleris et al. [24], in which no benefit was found from the use of hydroxychloroquine (with or without azithromycin) in patients with COVID-19. However, this method is not the most appropriate when trying to obtain comparison groups that also match at the time, or follow-up time, in which certain factors appear (e.g., time when a third drug is introduced).

Therefore, we set out to analyze the relevant clinical parameters under the use of azithromycin in patients hospitalized in our health centers, through comparison of groups using multivariate analysis and matching techniques based on brute-force algorithms, which can match patients by variables that change over time.

Method

This observational study was carried out on a cohort of 418 patients admitted to the hospitals of the Consorci Sanitari de l'Alt Penedès and Garraf (CSAPG), which includes three regional hospitals with a total of 275 acute-care beds and which serve a reference population of 247,357 inhabitants from the regions of Alt Penedès and Garraf, Catalonia, Spain.

Data were collected from all patients with a clinical picture compatible with COVID-19 (patients diagnosed of COVID-19 pulmonary disease by their doctors upon admission) seen between March 12 and May 2, 2020, from the time of admission to discharge or up to a maximum of 30 days after admission. Real-time reverse transcription polymerase chain reaction (RT -PCR) for SARS-CoV-2 was performed on a sample obtained by nasopharyngeal smear to all patients. Patients with negative RT-PCR test were excluded.

The data were collected from the electronic medical records by the COVID-19 research group of CSAPG. The data collected included sociodemographic data, previous diseases, chronic treatments, symptoms of disease presentation, vital signs, and clinical evolution each day since admission, including the need for oxygen therapy, the inspired fraction of oxygen (FiO2), and the oxygen administration system (nasal prongs, Venturi mask, reservoir mask, or invasive or noninvasive mechanical ventilation). All treatments given during admission were recorded, as well as all analyses and chest radiographs performed. The researchers responsible for data collection collected the data using a structured form created in the OpenClinica, version 3.1. (Copyright © OpenClinica LLC and collaborators, Waltham, MA, USA), following a common procedure on which they were previously trained. Quality controls were established during the data collection process, and the errors detected were corrected; the responsible researchers were retrained when necessary.

As exposure variable, treatment with azithromycin was considered. Azithromycin, according to the hospital protocol, was prescribed at a dose of 500 mg on the first day (oral or intravenous), followed by 250 mg daily, until completing 5 days of treatment. A patient was considered exposed to azithromycin if they received at least three doses of the drug. The main outcome variables for the efficacy analyses were time to discharge and oxygen saturation(%)/FiO2(%) ratio (SAFI) at 48 hours after the start of treatment [26, 27]. As secondary variables, SAFI in the first 96 hours after treatment, and mortality were analyzed.

In the statistical analysis, a double strategy was used: 1) analysis of subcohorts paired by confounding factors and 2) analysis of unpaired subcohorts, adjusted for confounding factors.

As part of the first strategy, a subcohort of patients treated with the study drug was formed, and a control subcohort was matched with the treatment group (1:1 match ratio). The patients were matched by the following prognostic markers, which were collected dichotomously (Yes/No) after detailed reading of all available patient reports: sex, age, obesity, heart failure, chronic renal failure, and sleep apnea–hypopnea syndrome (SAHS).

The above listed prognostic markers, which were used as matching criteria, were identified in multivariate binary logistic regression models, in which severe disease (defined as need for oxygen therapy with a non-rebreathing masks or mechanical ventilation) and death were taken as dependent variables. The variables introduced in these models were pre-selected from those pathological antecedents with significant association to the outcomes (bivariate analyses; p<0.05) by using the Lasso technique [28]. Virtually all the available pathological history of the patients was tested: cardiovascular, digestive, osteoarticular, pulmonary, endocrine, neurological, psychiatric diseases, kidney failure, neoplasms, autoimmune diseases and several immunodeficiencies (all the diseases within these categories were treated in a dichotomic way: presence vs absence of the disease).

Follow-up of each patient started the day the patient took the first dose of a study drug. Follow-up of each control started the day after admission on which SAFI, vital signs (blood pressure and heart rate), radiological involvement, and C-reactive protein (CRP), were similar to those of the patient with whom they were matched. For this purpose, the CRP on day 1 of follow-up of the patient or, failing that, the day before the start of treatment, was taken as reference. Likewise, the radiological involvement on the treatment started, or any previous time up to a maximum of two days before the start of treatment, was considered. Missing data on radiological involvement were imputed in the following way: It was assumed that the radiological involvement on the days between two equal radiographs was the same as on the days of said radiographs (e.g., if a patient had an X-ray with three affected quadrants on day 1 and another with three affected quadrants on day 6, it was assumed that on all intervening days they had three affected quadrants). This interpolation was allowed up to a maximum interval of 6 days between radiographs. No missing data were imputed for other variables. Patients who received the study treatment and their controls were matched only if they had received the same other treatments for COVID-19, including hydroxychloroquine, lopinavir/ritonavir, interferon, corticosteroids, or tocilizumab. A margin of 3 days of lag at the start of the other treatments was tolerated between the patients under study and the matched controls. In preliminary analyzes, we found no effect of heparin treatment on the results of this study, which is why this drug was not included among the matching criteria.

For pairing, a first step was performed using brute-force computing algorithms, which identified all possible controls in the database for each of the patients who received the study treatment. In this first step, controls were chosen who had the same sex and state of obesity ("yes" vs "no", according to the clinical history), the same radiological involvement (number of affected quadrants on anteroposterior radiography: 0–4) and an age difference not exceeding 15 years. The control was allowed to have a SAFI from 1.1 points lower to 2 points higher than the treated patient and a CRP from 6 mg/dL lower to 4 mg/dL higher than the treated patient. The matching was then refined, choosing from among the previously identified potential controls the most similar in terms of SAFI, blood pressure, heart rate, and CRP by the propensity score. The complete process of selecting patient pairs, including the procedures performed by the brute force algorithms, are summarized in Fig 1.

Fig 1

Process of selection of matched controls.

The success of the matching was verified by comparing means or percentages between groups. A different trend in the evolution of patients (improvement in one group and worsening in another) was discarded, verifying that the difference between the SAFI was similar between day 1 of analysis and the day before entering the analysis. In the matched subcohorts, the SAFI was studied at 48, 72, and 96 hours using Student’s t-test for independent samples and the time to discharge using the log-rank test. In the SAFI analyses, patients with palliative sedation were excluded because in these patients SAFI is not related to the severity of the disease. In the analysis of time to discharge, deceased patients were excluded.

In the analysis of unmatched subcohorts (second analysis strategy), the effect of azithromycin was analyzed in a subcohort in which all patients had been treated with hydroxychloroquine and lopinavir/ritonavir, from which patients treated with corticosteroids or other drugs that were distributed significantly asymmetrically between groups were excluded. In the analysis of these subcohorts, the total length of hospital stay was counted from day 1 of admission, and patients were considered exposed to the study drug if they had taken it at any time since admission (at least three doses). The time to discharge (excluding deceased patients) and mortality were studied by fitting Cox regression models adjusted for the following covariates: sex, age, obesity, heart failure, chronic renal failure, SAHS, baseline saturation in the emergency room, CRP in the emergency room, and quadrants affected in the emergency radiography.

For the statistical analysis, R software version 3.6.1 (R Project for Statistical Computing) and IBM SPSS statistics version 26 were used.

The research ethics committee Bellvitge Hospital reviewed the study and accepted the waiver of the patient's informed consent, as it was an observational and ambispective review of clinical data, and the patient's personal data were anonymized for its publication. Approval from the Ethical Committee was granted before starting data collection.

Results

Of the 464 consecutive patients with a clinical diagnosis of COVID-19 and pulmonary involvement who were admitted between March 12 and May 2, 2020, 46 were excluded for having a negative RT-PCR for SARS-CoV-2. Of the 418 patients included in the analysis, 238 (56.9%) were men and 180 (43.1%) were women, the mean age of the sample was 65.4 years (SD 16.6 years), and the median follow-up was 8 days (IQR 5–12 days). In total, 239 (57.2%) patients were treated with azithromycin. Patients who were treated with both hydroxychloroquine and lopinavir/ritonavir during admission totaled 346 (82.8%). In the first 30 days after admission, 79 patients died (18.9%). Fig 2 shows a flow diagram of the sample of the study.

Fig 2

Flow diagram.

The characteristics of the matched subcohorts are shown in Table 1. Comparing to the source cohort, matched cohorts had a larger proportion of men, and included patients with better respiratory function (higher saturation and less chest-x-ray involvement) as can be seen in Table 2. The characteristics of the unmatched subcohorts are shown in Table 3. Table 4 shows the mean change in saturation, FiO2 and SAFI, with respect to baseline, after 48, 72, and 96 hours of treatment, in the matched subcohorts. In the analysis of matched cohorts, hospital satay was significantly longer in patients treated with azithromycin, compared with their paired controls (Logrank; p = 0.039). However, in the unmatched cohorts, the increased hospital stay associated to azithromycin use, was no significant after adjustment using Multivariate Cox regression models: hazard ratio 1.45 (IC95%: 0.88–2.41; p = 0.150). Fig 3 shows the unadjusted Kaplan-Meier comparison curves and log-rank test for the time to discharge in all studied subcohorts.

Table 1

Baseline characteristics of patients treated with azithromycin and their matched controls.

	Azithromycin(n29)	Control (n29)	p
Age (years)	63.0	63.1	0.987
Men (n,%)	21 (72.4%)	21 (72.4%)	1.000
Obesity (n)	3 (10.3%)	3 (10.3%)	1.000
CHF (n)	1 (3.5%)	1 (3.5%)	1.000
CRF (n)	2 (6.7%)	5 (17.2%)	0.423
SAHS (n)	3 (10.3%)	3 (10.3%)	1.000
Tobacco (n)	1 (3.4%)	3 (10.3%)	0.611
Hypertension (n)	12 (41.4%)	11 (37.9%)	1.000
Diabetes (n)	5 (20.7%)	6 (17.2%)	1.000
COPD (n)	1 (3.4%)	1 (3.4%)	1.000
Other cardiopathy (n)	5 (17.2%)	2 (6.9%)	0.433
Saturation (%)	96.1	96.4	0.521
Systolic BP (mmHg)	121.6	123.4	0.690
Diastolic BP (mmHg)	70.7	68.7	0.484
HR (bpm)	77.8	78.5	0.783
Temperature (°C)	36.7	36.6	0.686
SAFI1	3.6	3.5	0.668
SAFI trend2	0.0	0.0	0.983
Radiographic involvement3	1.5	1.5	0.632
CRP (mg/dL)	7.4	8.0	0.655
Urea (mg/dL)	41.3 (n23)*	37.7 (n19)*	0.649
Neutrophils (10e9/L)	4.3 (n25)*	5.4 (n6)*	0.302
Lymphocytes (10e9/L)	1.1 (n25)*	1.0 (n6)*	0.511
Hydroxychloroquine (n)	27.0	26.0	0.640
Lop/Rit (n)	27.0	26.0	0.640
Interferon (n)	3	4	0.687
Tocilizumab (n)	5	3	0.706
Methylprednisolone (n)	1	0	0.313
Dexamethasone (n)	8	7	0.764
Hospital stay (days)	10.0	6.7	0.025

CHF: congestive heart failure. CRF: chronic renal failure. SAHS: sleep apnea–hypopnea syndrome. BP: blood pressure. HR: heart rate. SAFI: saturation (%)/fraction of inspired O2 (%). CRP: C-reactive protein.

1 Maximum value 4.76, corresponding to 100% saturation with FiO2 of 21%.

2 Change in SAFI with respect to the day before the start of the follow-up period.

3 Number of affected quadrants in an anteroposterior chest radiograph. Range: 0–4 (0: no involvement; 4: involvement of the upper and lower lobes of both lungs).

* Information not available for all the patients.

Table 2

Comparison between matched subcohorts and source cohort.

	Total cohort (n 418)	Matched subcohorts (n 58)	p
Age	65,4	63.1	0.320
Men	238 (57,1%)	42 (72.4%)	0.032
Obesity	74 (17.7%)	6 (10.3%)	0.192
CHF	26 (6,2%)	2 (3,4%)	0.558
CRF	61 (14,6%)	7 (12,1%)	0.693
SAHS	34 (8,1%)	6 (10,3%)	0.611
Tobacco	36 (8,6%)	4 (6,9%)	0,804
Hypertension	217 (51,9%)	23 (39.7%)	0.093
Diabetes	99 (23,7%)	11 (19.0%)	0,268
COPD	41 (9.8%)	2 (3.4%)	0.143
Other cardiopathy	62 (14.8%)	7 (12.1%)	0.693
Saturation	91,6	93.9	0.031
Radiographic involvement 1	2,07	1,59	0,003
CRP (mg/dL)	12.4 (n156)*	9.8	0,231
Urea (mg/dL)	48.0 (n337)*	38.3 (n42)*	0.088
Neutrophils (10e9/L)	6.0	4.7	0.015
Lymphocytes (10e9/L)	1.1	1.1	0.654
Hospital stay	9.3	9.2	0.775

* Information is not available for all the patients.

1 Number of affected quadrants in an anteroposterior chest radiograph. Range: 0–4 (0: no involvement; 4: involvement of the upper and lower lobes of both lungs).

Table 3

Baseline characteristics of the subcohorts of patients treated with hydroxychloroquine/lopinavir-ritonavir and patients with additional treatment with azithromycin (un-matched subcohorts).

	HCL/LOP (n 63)	HCL/LOP/AZT (n 120)	p
Age	57.2	61.6	0.066
Men	35 (52.2%)	57 (47.5%)	0.534
Obesity	12 (17.6%)	17 (14.2%)	0.526
CHF	3 (4.4%)	7 (5.8%)	1.000
CRF	4 (5.9%)	14 (11.7%)	0.195
SAHS	5 (7.4%)	10 (8.3%)	0.812
Tobacco	3 (4.4%)	17 (14.2%)	0,028
Hypertension	24 (35.3%)	57 (47.5%)	0.126
Diabetes	11 (16.2%)	25 (20.8%)	0,563
COPD	1 (1.5%)	7 (5.8%)	0,262
Other cardiopathy	9 (13,2%)	16 (13.3%)	1.000
Saturation	93.7	94.4	0.301
Radiographic involvement 1	2.13	1.68	0.004
CRP (mg/dL)	9.9 (n26)*	8.2 (n33)*	0.295
Urea (mg/dL)	33.2 (n 38)*	37.6 (n 111)*	0.440
Neutrophils (10e9/L)	5.1 (n38)*	5.0 (n115)*	0.942
Lymphocytes (10e9/L)	1.2 (n38)*	1.2 (n115)*	0.732
Hydroxychloroquine	63 (100%)	120 (100%)	-
Lop/Rit	63 (100%)	120 (100%)	-
Interferon	12 (17.6%)	12 (10.0%)	0.131
Tocilizumab	7 (10.3%)	11 (9.2%)	0.801
Methylprednisolone	0	0	-
Dexamethasone	0	0	-
Hospital stay (days)	5.7	8.5	<0.001

AZT: Azithromycin. CHF: congestive heart failure. CRF: chronic renal failure. CRP: C-reactive protein. HCQ: hydroxychloroquine. L/R: lopinavir/ritonavir. SAHS: sleep apnea-hypopnea syndrome.

1 Number of affected quadrants in an anteroposterior chest radiograph. Range: 0–4 (0: no involvement; 4: involvement of the upper and lower lobes of both lungs).

* Information not available for all the patients.

Table 4

Change in respiratory function parameters with respect to the first day of follow-up in patients treated with azithromycin.

	Azithromycin	Control	Mean difference (IC95%)	p
Saturation increment
48 hours	-0.82 (n29)	-0.81 (n29)	0.02 (-1.35; 1.39)	0.980
72 hours	-0.58 (n29)	-0.43 (n25)	0.15 (-1.18; 1.48)	0.821
96 hours	-0.91 (n27)	-0.40 (n24)	0.51 (-0.72; 1.74)	0.411
FiO2 increment
48 hours	4.93 (n29)	3.33 (n29)	-1.60 (-11.36;8.16)	0.744
72 hours	9.07 (n29)	0.56 (n27)	-8.51 (-21.77; 4.75)	0.203
96 hours	6.65 (n27)	-5.45 (n23)	-12.10 (-23.70; -0.50)	0.041
SAFI increment
48 hours	-0.19 (n29)	-0.01 (n29)	0.19 (-0.26; 0.64)	0.408
72 hours	-0.23 (n29)	0.34 (n25)	0.57 (0.01; 1.14)	0.046
96 hours	-0.08 (n27)	0.40 (n23)	0.49 (-0.08; 1.05)	0.074

FiO2: fraction of inspired oxygen.

Fig 3

Kaplan-Meier comparison curves and log-rank test outcomes of the different subcohorts (unadjusted).

Six deaths (8.1%) were recorded in the unmatched control subcohort (treated with hydroxychloroquine and lopinavir/ritonavir) and 7 people (5.3%) died in the subcohort receiving additional treatment with azithromycin (p = 0.501). In the matched subcohorts 3 people died, 1 of them (3.4%) in the azithromycin group, and 2 (6.9%) in the control group. We considered this numbers of events insufficient to draw conclusions.

Discussion

Our study did not find a benefit associated with the use of azithromycin in terms of respiratory function (SAFI), or time to discharge. In fact, hospital stay was longer in the azithromycin treated group, compared with matched controls.

Most of the patients in our study received hydroxychloroquine and lopinavir/ritonavir, therefore our results are mainly related to the potential benefit of adding azithromycin to this drug regimen. Our results are in line with the work by Cavalcanty et al. [20] who found no benefit on clinical severity in the group of patients receiving hydroxychloroquine plus azithromycin compared to the group who received only hydroxychloroquine. Rosenberg et al. [15] and Arshad et al. [16] investigated similar patients in terms of setting (hospitalized patients), severity (moderate or severe disease) and oxygenation parameters and found no benefit on hospital mortality in the group of patients treated with azithromycin alone.

Our sample included only hospitalized patients, so our results should be considered in the realm of hospital management of COVID-19 and cannot be extrapolated to patients with mild symptoms in whom outpatient treatment and monitoring is usually recommended. In addition, matching criteria cause patient selection, so that only those patients for whom a pair is found, enter the analysis (possibly selecting the most frequent type of patient). This means that the selected subcohorts do not represent well the total hospital population with COVID-19 and the results of this study are only applicable to patients with features similar to ours.

Given the observational nature of this study, the existence of residual confounders cannot be ruled out, thus, a possibility exists that patients assigned to azithromycin treatment would have a higher-risk factors or disease severity. In the case of azithromycin, we believe that this problem is not likely since its use was widespread and not related to the severity of the disease. In any event, the exhaustive matching method used and the verification of the comparability of the groups lead us to assume that this confounding effect was unlikely and, if there, was small. Some factors could affect hospital discharge beyond the resolution of the infection, such as the presence of complications or factors related to social circumstances, especially in elderly patients (which could have prevented them from returning home). Since all the patients in the sample were admitted for COVID-19, the complications derived from hospitalization for this disease, seem to us to be part of the clinical picture we are studying and, therefore, their influence in the outcome seems appropriate. Regarding social factors affecting discharge, we think that the age matching of the paired subcohorts, should have mitigated its possible influence in the results. In any event, the most important confounder that could have affected the time to discharge is death, which was appropriately controlled, by excluding the deceased patients from this analysis.

The worse respiratory function at 72 hours of treatment, observed in azithromycin matched subcohort, may be affected by a selection bias, as there was a loss of data in four matched controls at the time of this comparison, which was a secondary endpoint (loss of patients may have unbalance the previously matched groups).

Our study was limited by its small sample size, which caused problems of statistical power, especially in the case of some of the outcomes of greatest interest, such as mortality, which cannot be sufficiently studied in this sample. In addition, the use of secondary data obtained from the clinical history might have led to information biases. However, given that the main variables were quantitative parameters that were little influenced by the observers or their expertise in measurement, and given that these parameters are routinely collected in clinical practice and hospital management, we consider unlikely the existence of a relevant bias of this type. In any case, the sample size and observational nature of our study make it necessary to wait for the results of randomized clinical trials, which are ongoing, to confirm ours.

In conclusion, in this observational study, we did not find evidence of a clinical benefit from the use of azithromycin in patients hospitalized with COVID-19. The use of azithromycin may be associated with worse clinical results, compared with matched controls.

22 Jul 2020

PONE-D-20-19615

Observational study of azithromycin in hospitalized patients with COVID-19

PLOS ONE

Dear Dr. Rodríguez-Molinero,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

The three reviewers found your study quite interesting but simultaneously raised some concerns which are mainly attributed to methodological issues and the way that results were interpreted.

Please submit your revised manuscript by Sep 05 2020 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Stelios Loukides

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Please ensure you have mentioned the potential limitations of this study also in your Abstract. Moreover, please ensure that no statements of causation, which cannot be supported by the observational nature of this study, are reported. Furthermore, please provide more detail on how matching was performed, and on the definitions used (for example, please define "heart failure"); and provide a citation for the algorithm used, or describe the methodology in more detail if not previously published.

3.We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially identifying or sensitive patient information) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. Please see http://www.bmj.com/content/340/bmj.c181.long for guidelines on how to de-identify and prepare clinical data for publication. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

4. One of the noted authors is a group or consortium [COVID-19 research group of CSAPG]. In addition to naming the author group, please list the individual authors and affiliations within this group in the acknowledgments section of your manuscript. Please also indicate clearly a lead author for this group along with a contact email address.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: No

Reviewer #3: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: I Don't Know

Reviewer #2: No

Reviewer #3: I Don't Know

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #2: No

Reviewer #3: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: This is an interesting paper on an important topic with patient data drawn from a huge database. There is value in the study.

But, the value seems a bit hidden to me. First, even though I am an academic clinician who has cared for lots of sick people, I have never used the SAFI measurement. This leaves me uncertain of the value of the findings related to SAFI, and the validity of this as a measure of progression of pulmonary findings is never discussed. In addition, without knowing weaning parameters, it is not clear if the use of oxygen is based only on pulmonary disease or if oxygen could have been continued because the patient was still sick with other issues. Second, the basis of the matching procedure is not clear. It is surprising that out of 239 azithromycin-treated patients and a similar number of un-treated patients, there would only be 29 pairs that were similar enough to compare. I'd need to know more about the matching procedures to know if the selected 29 patients/pairs were really representative of the larger population of COVID-19-positive patients.

The ethics statement is reasonable, but the actual paper only says that ethical approval was requested, not granted.

The data statement is confusing. It is not clear what "re-identification" of patients means and why de-identified data could not be made available.

Abstract. It would help in the Background to say what was being tested (such as severity of pulmonary disease) instead of the obscure (to me) ratio that is of uncertain clinical relevance.

Abstract. The results are only about a small part of the results presented in the paper. And, by mixing the analyses without fully explaining them in the abstract, it sounds contradictory - azithromycin was associated with a longer time to discharge but was also of "no significant difference." It is also not clear why one would look at a log rank of a simple measure such as length of stay.

The third paragraph of the introduction could be updated with new studies when a revised manuscript is submitted.

The first mention of the matching is incomplete. Only later do we learn that obesity was a yes/no characterization, but we still don't hear how much obesity counted or how obesity was defined. Matching age to within 15 years seems a bit broad - the risks of bad outcomes are reported to be very different between 60 and 75 year olds, for instance.

The SAFI data in the abstract are only for 48 hours, but there were significant differences (p<0.05) in Fi)2 and SAFI increment at 72 and 96 hours. It is not clear why one time point being "not different" makes it into the final conclusion of the study, and the other "significant" findings do not. The "loss of data" explanation does not generate much confidence in the conclusions.

It would help to mention a p value with the 8.1% vs 5.3% mortality figures.

These seem to be valid and important data in this study, but attention to these points would help at least me better understand the meaning and significance of some of the reported details. Thanks!

Reviewer #2: The methodology used in the paperi is interesting. The authors have tried to solve the issue of the absence of a control group and to reduce the impact of confounding factors by using matching techniques based on brute-force algorithms to identify a subcohort of patients paired by confounding factors and a second subcohort of patients where confounding factors were adjusted according to a normal distribution.

However there al many major issues which should be addressend in the manuscript.

Methods

To identify the paired subcohort if patients authors have matched the initial sample according to prognostic markers identified at a bivariate analysis and a multivariate model. Could you please explicit which variables have you initially considered for the bivariate analysis and the multivariate model and which significativity threshold have you used?

Could you please add a flow diagram (according to the STROBE guidelines) summarizing how from the source population the two subcohorts have been identified?

One of the pairing criteria were other concomitant treatments for COVID-19, but among those you have considered there isn’t heparin. Could you please motivate this choice?

There are some concernings regarding the use of lenght of hospital stay as a primary endpoint. Could you please discuss if there are other factors which can affect the leght of hospital stay besides the infection resolution?

Results

In the paired cohort, 72% of patients were male, while in the source population authors report a prevalence of male subjects of 56,9%. The subcohort of paired patients does not represents anymore the population sample. An analysis evaluating differences between the paired subcohort and the source population should be performed and summarized in a table showing significativity.

Table 1: Please add variables percentages consistently with the format of table 2

According to table 1, 27 (or 26) out of 29 patients in the paired subcohort, have been treated with hydroxychloroquine and lopinavir/ritonavir with (or without) aziththromycin. So you are not evaluating the efficacy of azythromycin in COVID-19, but the effect of the addittion of azithromycin to the combination of hydroxychloroquine and lopinavir/ritonavir. Please consider this in your discussion.

Population characteristics in table 1 and 2 should be better described. Please add smoke hystory, hypertension, diabetes, COPD, cardiovascular disease other than CHF, WBC and lenght of hospital stay.

It is not clear how many patients died in which subcohort and group. Furthermore authors says that numbers of deaths are insuffient to draw conclusions but few lines below they say “Our study found no benefit associated with the use of azithromycin in terms of respiratory function (SAFI), time to discharge, or mortality”. This is a conclusion regarding mortality.

Please express all results with 95%IC

Reviewer #3: Title: Observational study of azithromycin in hospitalized patients with COVID-19

This study aimed to analyze whether there is an association between the use of azithromycin and the evolution of saturation/fraction of inspired oxygen (SAFI) or time to discharge in patients hospitalized with COVID-19 in a Spanish cohort.

Comment 1: T I recommend checking the text "covid-19" and change it to "COVID-19”.

Comment 2: In the methods section, the authors define that they included patients with a clinical picture compatible with COVID-19. What were these criteria (clinical, imaging, nucleic acid amplification tests, serology)? I recommend adding it to the section.

Comment 3: Include in the methods section, it is not clear the type of diagnostic tests used, please detail it.

Comment 4: In the results section, the interquartile range of the median follow-up does not have one of the quartiles. Please verify this result.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

Reviewer #3: Yes: Luis Gabriel Parra-Lara

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

1 Aug 2020

Reviewer #1:

1.This is an interesting paper on an important topic with patient data drawn from a huge database. There is value in the study.

But, the value seems a bit hidden to me. First, even though I am an academic clinician who has cared for lots of sick people, I have never used the SAFI measurement. This leaves me uncertain of the value of the findings related to SAFI, and the validity of this as a measure of progression of pulmonary findings is never discussed. In addition, without knowing weaning parameters, it is not clear if the use of oxygen is based only on pulmonary disease or if oxygen could have been continued because the patient was still sick with other issues.

ANSWER: Oxygen saturation / FiO2 ratio (SAFI) is a parameter we choose, in order to make the oxygen saturation proportional to the amount of oxygen that it is being administered, otherwise, the oxygen saturation is not interpretable. SAFI (also known as S/F) is highly related to PAFI, which is Oxygen partial pressure / FiO2, and has been used, in order to interpret oxygen saturation in patients with oxygen therapy.

Here we show you some examples from bibliography, were the SpO2/FIO2 ratio is used. We have now included in the paper a couple of them.

• Rice TW, et al. Comparison of the SpO2/FIO2 ratio and the PaO2/FIO2 ratio in patients with acute lung injury or ARDS. Chest. 2007;132(2):410-417. doi:10.1378/chest.07-0617

• Lu X, et al. Continuously available ratio of SpO2/FiO2 serves as a noninvasive prognostic marker for intensive care patients with COVID-19. Respir Res. 2020 Jul 22;21(1):194.

• Fernández-Ruiz M et al. for the treatment of adult patients with severe COVID-19 pneumonia: A single-center cohort study. J Med Virol. 2020 Jul 16;

• Serpa Neto A et al. The use of the pulse oximetric saturation/fraction of inspired oxygen ratio for risk stratification of patients with severe sepsis and septic shock. J Crit Care. 2013 Oct;28(5):681-6.

• Schmickl CN et al. Decision support tool for early differential diagnosis of acute lung injury and cardiogenic pulmonary edema in medical critically ill patients. Chest. 2012 Jan;141(1):43-50. doi: 10.1378/chest.11-1496. Epub 2011 Oct 26.

• Lobete Prieto C, et al. [Prediction of PaO₂/FiO₂ ratio from SpO₂/FiO₂ ratio adjusted by transcutaneous CO₂ measurement in critically ill children]. An Pediatr (Barc). 2011 Feb;74(2):91-6.

• Tripathi RS et al. Pulse oximetry saturation to fraction inspired oxygen ratio as a measure of hypoxia under general anesthesia and the influence of positive end-expiratory pressure. J Crit Care. 2010 Sep;25(3):542.e9-13.

Honestly we consider that oxygen has not been continued for any other reason than the pulmonary disease (the whole database includes patients with COVID-19 pulmonary disease). In line with the reviewer’s comment, we took the precaution of excluding patients with palliative sedation, since in them oxygen therapy does not usually correspond to lung function.

Please see new references 26, 27 and 10th paragraph of the method section.

2. Second, the basis of the matching procedure is not clear. It is surprising that out of 239 azithromycin-treated patients and a similar number of un-treated patients, there would only be 29 pairs that were similar enough to compare. I'd need to know more about the matching procedures to know if the selected 29 patients/pairs were really representative of the larger population of COVID-19-positive patients.

ANSWER: We intended to ensure that both groups were comparable. Thus, we matched for a large number of characteristics, including the many different treatments used for the disease and their timing within the admission. It is very difficult to find in a “natural cohort” (no inclusion criteria), pairs of patients so similar in so many respects. As physicians tend to use different treatments in different types of patients, it is hard to find totally equally pairs. That is why the size of the paired cohorts is small (small but comparable)

The reviewer is right when says that the matched cohort is not representative of the hospital population with COIVD-19. Paired cohorts closely resemble a clinical trial sample with strict inclusion criteria. Clinical trials often do not represent all patients with a certain disease, but rather the proportion that meets the trial criteria (which is sometimes a small proportion of the total population). The characteristics of the trial population are usually shown in publications’ table 1, which is equivalent to our table 1. The table helps to understand the type of patient’s to whom the results can be extrapolated. We have added a new table (table 2) which compares the paired cohorts and the source cohort. We have also commented on the differences in the results section. We have also discussed this topic in the discussion section, inviting readers not to generalize the results to any COVID patient, but only to those who resemble those included in our study.

Regarding the control selection process, we have added a new figure, explaining the procedure (new fig.1)

Please see changes in the 2nd paragraph of the results, 3rd paragraph of the discussion, table 2, and figure 1.

3. The ethics statement is reasonable, but the actual paper only says that ethical approval was requested, not granted.

ANSWER: The ethics approval was granted; we have changed the sentence to better express this fact.

Please, see changes in the last paragraph of the methods section.

4. The data statement is confusing. It is not clear what "re-identification" of patients means and why de-identified data could not be made available.

ANSWER: As we gathered so many data of each single patient, some of them are still identifiable, even without de identification data (example: there are no so many people admitted to our hospital, who are male, 96 years old, had sleep apnea, and survived COVID-19, between March and May 2019). For this reason, we could not make data available, specially, when the patients did not consent for this study, as stated in the ethical approval.

We could only publish a limited set of variables, aggregated (for example, in ranges of age of 5-10 years), previously revised and explicitly approved by our Ethical Committee.

The editorial team has also raised the same question; we are answering in parallel to them. They will modify the statement, according to the information provided by us, and the final decision.

5. Abstract. It would help in the Background to say what was being tested (such as severity of pulmonary disease) instead of the obscure (to me) ratio that is of uncertain clinical relevance.

ANSWER: We have added the suggestion of the reviewer to the abstract.

Please, see changes in the first paragraph of the abstract.

6. Abstract. The results are only about a small part of the results presented in the paper. And, by mixing the analyses without fully explaining them in the abstract, it sounds contradictory - azithromycin was associated with a longer time to discharge but was also of "no significant difference." It is also not clear why one would look at a log rank of a simple measure such as length of stay.

ANSWER: We have modified the abstract to make it more understandable, and also have added means of hospital stay for each groups.

Please, see changes in the third paragraph of the abstract.

7. The third paragraph of the introduction could be updated with new studies when a revised manuscript is submitted.

ANSWER: We have updated references to existing bibliography in this paragraph.

Please, see changes in the third paragraph of the introduction, and the second paragraph of the discussion.

8. The first mention of the matching is incomplete. Only later do we learn that obesity was a yes/no characterization, but we still don't hear how much obesity counted or how obesity was defined. Matching age to within 15 years seems a bit broad - the risks of bad outcomes are reported to be very different between 60 and 75 year olds, for instance.

ANSWER: This is a study based on secondary data (clinical records), which limits the use of strict definitions for patient characteristics. In the case of obesity, a patient was considered obese if, in the pathological history of his emergency report or admission note, a doctor had mentioned obesity as part of the chronic conditions description.

Regarding age, it must be considered that fifteen years is the maximum deviation allowed by the matching algorithm, the difference is not so large in most pairs. Given the high number of matched variables, a reduction in the age-matching margin would have limited the number of patients matched. In any event, existing age differences in patient-pairs, seems to be balanced (equally in favor and against the treatment group), as the mean age of groups are almost identical, which nullifies any possible effect of age.

We have tried to explain better the matching process (a new figure is now added), the selection of matching criteria, and how variables were collected. Pease see 6th and 7th paragraph of the methods, and the new figure 1.

9. The SAFI data in the abstract are only for 48 hours, but there were significant differences (p<0.05) in Fi)2 and SAFI increment at 72 and 96 hours. It is not clear why one time point being "not different" makes it into the final conclusion of the study, and the other "significant" findings do not. The "loss of data" explanation does not generate much confidence in the conclusions.

ANSWER: As it is shown in table 3 (now table 4), loss of data (reduction in n) affects only to variables after 72 hours. As the loss of patients’ data is not homogeneous between groups, and occurs after the moment of pairing, it is possible that it could lead to residual confusion (due to "un-pairing"). That it is why we were cautious, and we did not highlight this finding in the conclusions.

In any case, we have tested now, how the pairing of the patients remains after the losses regarding SAFI at 72 hours, and the groups continue to be well matched except for a greater presence of chronic renal failure before admission, in the azithromycin group (5 patients vs 1 patient). As this is not very likely to affect respiratory function on the third day of treatment, we have dared to add to the conclusions that “azithromycin may be associated with worse results.”

Please, see changes in the last paragraph of the discussion.

10. It would help to mention a p value with the 8.1% vs 5.3% mortality figures.

ANSWER: We have included a p value for this difference.

Please see the last paragraph of the results.

11. These seem to be valid and important data in this study, but attention to these points would help at least me better understand the meaning and significance of some of the reported details. Thanks!

ANSWER: We have tried to address each point raised by the reviewer. Many thanks for the careful review.

---------------------------------------------------------------

Reviewer #2:

1. The methodology used in the paperi is interesting. The authors have tried to solve the issue of the absence of a control group and to reduce the impact of confounding factors by using matching techniques based on brute-force algorithms to identify a subcohort of patients paired by confounding factors and a second subcohort of patients where confounding factors were adjusted according to a normal distribution.

However there al many major issues which should be addressend in the manuscript.

Methods

To identify the paired subcohort if patients authors have matched the initial sample according to prognostic markers identified at a bivariate analysis and a multivariate model. Could you please explicit which variables have you initially considered for the bivariate analysis and the multivariate model and which significativity threshold have you used?

ANSWER: We have explained now in the paper the bivariate analysis and multivariable models. Variables included in the models were pre-selected using the Lasso technique [reference 28], from those pathological antecedents, with statistical significance (p<0.05) in a bivariate analysis performed with all the known pathologies of the patients. Although p<0.05 may seem to be a too strict entry criterion for multivariable models, we did not consider it to be so, in the light of the very high number of comparisons performed in the bivariate analysis (p<0.05 was not corrected for the large number of comparisons in this step)

In response to the reviewer’s comment, we have now introduced a short explanation in the manuscript. Please see the 7th paragraph of the methods section.

2. Could you please add a flow diagram (according to the STROBE guidelines) summarizing how from the source population the two subcohorts have been identified?

ANSWER: According to the reviewer’s comment, we have included a flow diagram.

Please see new figure 2.

3. One of the pairing criteria were other concomitant treatments for COVID-19, but among those you have considered there isn’t heparin. Could you please motivate this choice?

ANSWER: The vast majority of patients in the sample had heparin, some of them at prophylactic doses, others at higher doses. We previously analyzed the effect of heparin on the outcomes of our study and found no effect, so we did not match the sample for this drug.

We have now added a sentence explaining it to the manuscript. Please see the 8th paragraph of the method section.

4. There are some concernings regarding the use of lenght of hospital stay as a primary endpoint. Could you please discuss if there are other factors which can affect the leght of hospital stay besides the infection resolution?

ANSWER: Other than the infection resolution, the main factor that can affect hospital stay is death; however, we excluded deceased patients from the analyses of time to discharge.

Following the reviewer’s comment, we have included a discussion on other factors in the manuscript. Please see the 4th paragraph of the discussion section.

5. In the paired cohort, 72% of patients were male, while in the source population authors report a prevalence of male subjects of 56,9%. The subcohort of paired patients does not represents anymore the population sample. An analysis evaluating differences between the paired subcohort and the source population should be performed and summarized in a table showing significativity.

ANSWER: We have added table comparing our matched cohorts with the source cohort, and commented the differences in the text.

Nevertheless, our paired subcohorts are not intendent to represent the population, they are intendent to be comparable, in order to test azithromycin efficacy. Just like in clinical trials, where the sample selected by the inclusion criteria typically does not represent all the population with a given disease. As a consequence, the readers of a clinical trial paper, should refer to the “table 1”, in order to verify the type of patients involved in the trial (ie. disease severity, age range…), because the results will only be applicable to patients with similar characteristics to those included (it is a well-known drawback of most clinical trials; for instance, excluding elderly people) Our matched study is a clinical trial-like research. Matching criteria are thought to make groups comparable, not to select a representative sample. Our Table 1 shows the characteristic of the matched patients, and readers should pay attention to this table, before generalizing conclusions to their COVID-19 patients.

We have added a comment regarding the “representativeness” of the matched cohorts in the discussion.

Please see the new table 2, changes in the second paragraph of the results and changes in the 3rd paragraph of the discussion.

6. Table 1: Please add variables percentages consistently with the format of table 2

ANSWER: Done. Please see changes in the table 3 (old table 2).

7. According to table 1, 27 (or 26) out of 29 patients in the paired subcohort, have been treated with hydroxychloroquine and lopinavir/ritonavir with (or without) aziththromycin. So you are not evaluating the efficacy of azythromycin in COVID-19, but the effect of the addittion of azithromycin to the combination of hydroxychloroquine and lopinavir/ritonavir. Please consider this in your discussion.

ANSWER: You are wright. We have tried to clarify this point. Please see changes in the second paragraph of the discussion.

8. Population characteristics in table 1 and 2 should be better described. Please add smoke hystory, hypertension, diabetes, COPD, cardiovascular disease other than CHF, WBC and lenght of hospital stay.

ANSWER: We have added to the tables, the characteristics requested by the reviewer.

Please see the new tables of the paper.

9. It is not clear how many patients died in which subcohort and group. Furthermore authors says that numbers of deaths are insuffient to draw conclusions but few lines below they say “Our study found no benefit associated with the use of azithromycin in terms of respiratory function (SAFI), time to discharge, or mortality”. This is a conclusion regarding mortality.

ANSWER: Thanks for the last comment, it is a mistake we did not detected in the first version. We have corrected accordingly the first paragraph of the discussion. In addition, we have changed the last paragraph of the results, to clarify the other issue raised by the reviewer and reports deaths in every subcohort.

Please see corrections in the las paragraph of the results section and the first paragraph of the discussion section.

10. Please express all results with 95%IC

ANSWER: We have added 95% IC to the results in the text and table 4.

We appreciate very much the careful review, which in our view have contributed to improve the manuscript.

---------------------------------------------------------------

Reviewer #3: Title: Observational study of azithromycin in hospitalized patients with COVID-19

This study aimed to analyze whether there is an association between the use of azithromycin and the evolution of saturation/fraction of inspired oxygen (SAFI) or time to discharge in patients hospitalized with COVID-19 in a Spanish cohort.

Comment 1: T I recommend checking the text "covid-19" and change it to "COVID-19”.

ANSWER: We have updated the text following the reviewer suggestion.

Comment 2: In the methods section, the authors define that they included patients with a clinical picture compatible with COVID-19. What were these criteria (clinical, imaging, nucleic acid amplification tests, serology)? I recommend adding it to the section.

ANSWER: There were no defined set of clinical criteria. We included all patients whose doctors considered the COVID-19 as the clinical diagnosis upon admission, and who were treated with SARS-CoV-2 directed treatments (some of this patients were afterwards excluded when PCR results were available)

In line with the reviewer comment we have made a slight modification in the text, to clarify this point. Please see changes in the second paragraph of the methodology.

Comment 3: Include in the methods section, it is not clear the type of diagnostic tests used, please detail it.

ANSWER: We have now clarified the diagnostic test in the methods. Please see changes in the second paragraph of the methods section.

Comment 4: In the results section, the interquartile range of the median follow-up does not have one of the quartiles. Please verify this result.

We intended to show the “range” of the interquartile range rather than its limits. We have now display the interquartile range limits, for clarity (thanks, it is now more understandable). We also have detected an erratum in the median previously reported, which we have now corrected.

Please see changes in the first paragraph of the results section.

We would like to thank reviewer 3 for his comments.

Submitted filename: Response to reviewers.docx

Click here for additional data file.

24 Aug 2020

Observational study of azithromycin in hospitalized patients with COVID-19

PONE-D-20-19615R1

Dear Dr. Rodríguez-Molinero,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Stelios Loukides

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #3: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #3: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #3: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #3: No

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #3: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: (No Response)

Reviewer #3: (No Response)

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #3: Yes: Luis Gabriel Parra-Lara

27 Aug 2020

PONE-D-20-19615R1

Observational study of azithromycin in hospitalized patients with COVID-19

Dear Dr. Rodríguez-Molinero:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr Stelios Loukides

Academic Editor

PLOS ONE