Open access Original research BMJ Open: first published as 10.1136/bmjopen-2021-048982 on 22 September 2021. Downloaded from http://bmjopen.bmj.com/ on December 21, 2022 by guest. Protected by copyright. Sex as a prognostic factor for mortality in critically ill adults with sepsis: a systematic review and meta-­analysis Alba Antequera ‍ ‍,1 Jesus Lopez-­Alcalde ‍ ‍,2,3,4,5 Elena Stallings,3,5 Alfonso Muriel ‍ ‍,3,5,6 Borja Fernández Félix,3,5 Rosa del Campo ‍ ‍,7 Manuel Ponce-­Alonso ‍ ‍,7 Pilar Fidalgo,2,8 Ana Veronica Halperin,7 Olaya Madrid-­Pascual ‍ ‍,9 Noelia Álvarez-­Díaz ‍ ‍,10 Ivan Solà ‍ ‍,11,12 Federico Gordo,2,13 Gerard Urrutia,11,12 Javier Zamora ‍ ‍3,5,14 To cite: Antequera A, Lopez-­ ABSTRACT Alcalde J, Stallings E, et al. Objective To assess the role of sex as an independent STRENGTHS AND LIMITATIONS OF THIS STUDY Sex as a prognostic factor for prognostic factor for mortality in patients with sepsis ⇒ To our knowledge, this systematic review is the first mortality in critically ill adults addressing the prognostic independent effect of sex admitted to intensive care units (ICUs). with sepsis: a systematic review on mortality for patients with sepsis following the Design Systematic review and meta-­analysis. and meta-­analysis. BMJ Open Data sources MEDLINE, Embase, Web of Science, ​ recommended standards for reviews of prognostic 2021;11:e048982. doi:10.1136/ bmjopen-2021-048982 ClinicalTrials.​gov and the WHO Clinical Trials Registry from factor studies. inception to 17 July 2020. ⇒ The meta-­analysis pooled adjusted estimates for at ► Prepublication history and least one of five core covariate domains prespec- Study selection Studies evaluating independent additional supplemental material ified following a literature search and consensus associations between sex and mortality in critically ill for this paper are available among experts. adults with sepsis controlling for at least one of five core online. To view these files, covariate domains prespecified following a literature ⇒ The certainty of the evidence was evaluated us- please visit the journal online (http://dx.doi.org/10.1136/​ search and consensus among experts. ing the Grades of Recommendations, Assessment, bmjopen-2021-048982). Data extraction and synthesis Two authors Development and Evaluation approach. independently extracted and assessed the risk of bias ⇒ Heterogeneity was substantial between the included Received 13 January 2021 using Quality In Prognosis Studies tool. Meta-­analysis studies. Accepted 30 July 2021 was performed by pooling adjusted estimates. The Grades of Recommendations, Assessment, Development and Evaluation approach was used to rate the certainty of in intensive care units (ICUs) and accounts evidence. for one of five deaths worldwide.2–4 It is a Results From 14 304 records, 13 studies (80 520 heterogeneous illness affecting males more participants) were included. Meta-­analysis did not find often than females.5 Evaluating if outcomes sex-­based differences in all-­cause hospital mortality (OR differ by sex is a recognised health research 1.02, 95% CI 0.79 to 1.32; very low-­certainty evidence) priority.6 It has been hypothesised that sex may and all-­cause ICU mortality (OR 1.19, 95% CI 0.79 to 1.78; have a prognostic effect on sepsis outcomes. very low-­certainty evidence). However, females presented Biological mechanisms concerning the rela- higher 28-­day all-­cause mortality (OR 1.18, 95% CI 1.05 to 1.32; very low-­certainty evidence) and lower 1-­year tion between sex hormone metabolism and all-­cause mortality (OR 0.83, 95% CI 0.68 to 0.98; low-­ immune responses are known to underpin certainty evidence). There was a moderate risk of bias in this hypothesis.7–11 However, individual the domain adjustment for other prognostic factors in six studies evaluating the relationship between studies, and the certainty of evidence was further affected sex and outcome of sepsis report conflicting by inconsistency and imprecision. and imprecise findings.12–14 © Author(s) (or their Conclusion The prognostic independent effect of sex on Prognostic research that identifies patient employer(s)) 2021. Re-­use all-­cause hospital mortality, 28-­day all-­cause mortality characteristics associated with outcomes in permitted under CC BY-­NC. No and all-­cause ICU mortality for critically ill adults with people with a particular condition15 can be commercial re-­use. See rights sepsis was uncertain. Female sex may be associated with and permissions. Published by collated in evidence syntheses to examine decreased 1-­year all-­cause mortality. BMJ. the role of sex in mortality among patients PROSPERO registration number CRD42019145054. For numbered affiliations see with sepsis. It may help in risk stratification end of article. of these patients by combining independent INTRODUCTION prognostic factors within prognostic models, Correspondence to Dr Alba Antequera; Sepsis, a life-­threatening organ dysfunction which contribute to the selection of the most ​alba.​antequera.​martin@​gmail.​ produced by a dysregulated host response appropriate therapeutic options.15 Using a com to inflammation,1 is a leading cause of death systematic review search filter in PubMed, we Antequera A, et al. BMJ Open 2021;11:e048982. doi:10.1136/bmjopen-2021-048982 1 Open access BMJ Open: first published as 10.1136/bmjopen-2021-048982 on 22 September 2021. Downloaded from http://bmjopen.bmj.com/ on December 21, 2022 by guest. Protected by copyright. found two potentially relevant citations.16 17 Their detailed aged 16 years and older with a sepsis diagnosis, as defined assessment showed several weaknesses. For example, there by the study authors, treated in an ICU. Studies including was no definition of eligibility criteria concerning studies both adult and paediatric patients were eligible if adults that capture independent associations, a feature that is represented more than 80% of the study sample. Sex critical for focussing the review on prognostic evidence.18 and gender are distinct concepts, though often errone- In addition, specific tools19 for the assessment of risk of ously interchanged in the medical research reports.22 We bias in prognostic studies were not applied. Therefore, accepted any assessment of sex as a biological character- an evidence synthesis tailored to the specific methodolog- istic. We also appraised operational concepts of sex and ical requirements of prognostic research is required to gender provided by the study authors using the classifi- help delineate the significance of sex in sepsis outcomes cation detailed in online supplemental table 2.23 After a in critically ill patients. literature search and consensus among experts (online We conducted a systematic review and meta-­analysis supplemental table 3), we prespecified the following core to summarise the available evidence to assess the role of set of adjustment factors: age, severity score (Sequential sex as an independent prognostic factor for mortality in Organ Failure Assessment score, Simplified Acute Phys- patients with sepsis admitted to the ICU. iology Score II or Acute Physiologic Assessment and Chronic Health Evaluation II), comorbidities (immuno- suppression, pulmonary diseases, cancer, liver diseases METHODS or alcohol dependence), non-­ urinary source of infec- We registered the protocol with PROSPERO tion, and inappropriate or late antibiotic coverage. The (CRD42019145054) and published it in full.20 Online coprimary outcomes were all-­cause hospital mortality and supplemental table 1 details the differences between the 28-­day all-­cause mortality. Secondary outcomes were 7-­day protocol and the review. We adhered to the Preferred all-­cause hospital mortality, 1-­year all-­cause mortality and Reporting Items for Systematic Reviews and Meta-­Analyses all-­cause ICU mortality. Table 1 describes the review ques- statement.21 tion according to the population, index, comparator, Eligibility criteria outcome(s), timing, setting. We included studies (experimental or any observational design) that sought to confirm the independent prog- Search strategy and selection process nostic effect of sex on mortality in critically ill adults We searched MEDLINE Ovid, Embase Elsevier and Web with sepsis controlling for covariates (called phase of Science for studies published from inception to 17 2-­confirmatory studies, which means the objective state- July 2020, and ​ClinicalTrials.​gov and the WHO Interna- ment outlined sex as a prognostic factor of interest and tional Clinical Trials Registry Platform for unpublished analyses adjusted for covariates).18 We included patients and ongoing studies, regardless of language. The search Table 1 PICOTS system Index prognostic Population factor Comparator Outcome(s) Timing Setting Adults with sepsis Sex Non-­applicable to this Primary outcomes   ICUs review*       All-­cause hospital The longest follow-­up   mortality provided by the study authors (until death of hospital discharge)       28-­day all-­cause mortality 28 days from sepsis   diagnosis       Secondary outcomes           7-­day all-­cause hospital 7 days from sepsis   mortality diagnosis       1-­year all-­cause mortality 1 year from sepsis diagnosis         All-­cause ICU mortality The longest follow-­up   provided by the study authors (until death of ICU discharge) *Core set of adjustment factors: age, severity score (Sequential Organ Failure Assessment score, Simplified Acute Physiology Score II or Acute Physiologic Assessment and Chronic Health Evaluation II), comorbidities (immunosuppression, pulmonary diseases, cancer, liver diseases or alcohol dependence), non-­urinary source of infection and inappropriate or late antibiotic coverage. ICUs, intensive care units; PICOTS, population, index, comparator, outcome(s), timing, setting. 2 Antequera A, et al. BMJ Open 2021;11:e048982. doi:10.1136/bmjopen-2021-048982 Open access BMJ Open: first published as 10.1136/bmjopen-2021-048982 on 22 September 2021. Downloaded from http://bmjopen.bmj.com/ on December 21, 2022 by guest. Protected by copyright. strings included terms related to the population (sepsis), performing subgroup analyses. We undertook a single the prognostic factor (sex), prognostic study methods and prespecified subgroup analysis for prospective vs retro- the outcome (mortality). Furthermore, we handsearched spective studies when appropriate. We compared conference proceedings from 2010 to 2019 of the fore- differences between subgroups by performing a test of most critical care and infectious diseases symposia. Online interaction.33 We carried out no subgroup analyses based supplemental table 4 presents the full search strategy. on other study characteristics because there were insuffi- We used the online software EPPI-­ Reviewer V.4 to cient studies. We conducted sensitivity analyses accounting manage the study selection process.24 Pairs of review for the risk of bias excluding studies with either a high authors independently screened the title and abstracts, or moderate risk of bias in one of the following QUIPS and when appropriate, full texts to determine their key domains: study attrition, prognostic factor measure- eligibility. We used a consensus method and consulted a ment, outcome measurement and adjustment for other third author if disagreement remained. prognostic factors. Additionally, we explored potential differences between meta-­analyses based on unadjusted Data extraction and risk of bias assessment (crude) and adjusted estimates, and the impact of the Two authors independently extracted data and reached a unique information reported in abstract conferences.34 consensus using electronic extraction templates in EPPI-­ We could not perform further sensitivity analyses as no Reviewer V.4. We used the checklist for critical appraisal other comparisons met the predefined criteria. Although and data extraction for systematic reviews of prediction we planned to assess publication bias for each meta-­ modelling studies for prognostic factors guidance for data analysis including ≥10 studies by funnel plot representa- collection.25 We contacted all study authors for missing infor- tion and Peter’s test at a 10% level,35 no meta-­analysis met mation. Two authors independently assessed the risk of bias this criterion. of the included studies, agreed on ratings and a third author participated when required. We applied an outcome-­level Assessment of the certainty of evidence approach and amended the Quality In Prognosis Studies We assessed the certainty of evidence using the Grades of (QUIPS) tool using four categories (low, moderate, high or Recommendations, Assessment, Development and Eval- unclear risk).19 25 26 We defined studies controlling for less uation (GRADE) approach and guidance for prognosis than three of the aforementioned covariates as ‘minimally studies (online supplemental table 5).27 36–41 We tabulated adjusted for other prognostic factors or moderate risk’, our findings for each outcome using the GRADEpro GDT and those controlling for at least three of these covariates software.42 We described results for prognostic effect esti- as ‘adequately adjusted or low risk of bias’ for the QUIPS mate considering the certainty of evidence and its clinical adjustment domain.27 We assessed selective reporting bias importance (important effect, slight effect and little or no by: (1) searching for a prospective study protocol or regis- effect). As we found no well-­established clinically important tration, (2) dealing with related conference abstracts and thresholds for prognostic effects, we agreed a priori on (3) carefully examining the study methods section.19 an absolute risk difference of at least ±10‰ as clinically important difference. Data synthesis For each study and prognostic factor estimate, we Patient and public involvement extracted the measures of associations alongside its CIs. No patients or the general public involved. We transformed association measures into an OR with its 95% CIs to allow statistical pooling whenever adequate.28 We estimated no data from Kaplan-­Meier curves because RESULTS of the risk of overestimation of events and censorship Our searches threw a total of 14 304 records. After removing concerns.29 We presented results consistently, so associa- duplicates, we screened 13 115 titles and abstracts and tions above one indicated a higher mortality for female identified 146 full texts for further examination. Finally, participants. We pooled estimates in meta-­analyses when the review included 13 studies43–55 (figure 1). One study valid data were available. For the primary analyses, we included55 was reported as a conference abstract. Thus, used estimates from the model that adjusted for more we examined database information published elsewhere56 covariates from the core of adjustment factors. We to obtain further details on study methods. The included performed random-­ effects meta-­analyses applying the studies involved a total of 80 520 adult participants (45.25% Hartung-­Knapp-­Sidik-­Jonkman (HKSJ) adjustment,30 females). Table 2 and online supplemental table 6 display using RevMan V.5.3 (The Cochrane Collaboration, their characteristics. Online supplemental table 7 and Copenhagen, Denmark) and the template for conversion online supplemental table 8 show the sepsis definition and provided by IntHout.31 We examined statistical heteroge- covariates included in the adjusted models of each study, neity computing prediction intervals when the random-­ respectively. Although four studies47 50 53 54 had phase 2 effects meta-­analysis contained at least three studies.30 32 designs and provided adjusted data on mortality, their time We also calculated I2 and τ2 statistics to provide further frames differed from ours and/or reported unadjusted quantifications of statistical heterogeneity. We planned to estimates for some of the review outcomes. Hence, we only explore possible methodological causes of heterogeneity used those data for sensitivity analyses. Antequera A, et al. BMJ Open 2021;11:e048982. doi:10.1136/bmjopen-2021-048982 3 Open access BMJ Open: first published as 10.1136/bmjopen-2021-048982 on 22 September 2021. Downloaded from http://bmjopen.bmj.com/ on December 21, 2022 by guest. Protected by copyright. hospital mortality (OR 1.02, 95% CI 0.79 to 1.32; I2=64%; very low-­certainty evidence) (figure 2A). The 95% predic- tion interval ranged from 0.5 to 2.08. Sensitivity analyses results remained unaltered either excluding the study55 only reported as a conference abstract (OR 0.95, 95% CI 0.55 to 1.64), or using unadjusted estimates (OR 1.00, 95% CI 0.88 to 1.14) (online supplemental figure 2 and online supplemental figure 3, respectively). We examined sex-­based differences in 28-­day all-­cause mortality. We found six studies44 49 50 52–54 (20 930 recruited participants) addressing this question. Three studies44 49 52 (12 579 analysed participants) provided adjusted results. Meta-­analysis found higher 28-­day all-­cause mortality in the female group (OR 1.18, 95% CI 1.05 to 1.32; I2=0%; very low-­certainty evidence) (figure 2B). Considering a risk of 24% for 28-­day all-­cause mortality in male patients, 31 more female patients per 1000 will die (95% CI from 9 to 54 more), as compared with male patients. The 95% prediction interval ranged from 0.56 to 2.5. Sensitivity analysis results were inconclusive either pooling only studies with low or uncertain risk of bias for all key QUIPS domains (OR 1.17, 95% CI 0.88 to 1.56) or unadjusted estimates (OR 1.05, 95% CI 0.84 to 1.32) (online supple- mental figure 4). Figure 1 Flow diagram. ICU, intensive care unit. Secondary outcomes No study evaluated the prognostic role of sex on 7-­day Online supplemental figure 1 depicts the risk of bias all-­cause hospital mortality. We sought sex-­related differ- assessment at outcome level of each included study using ences in 1-­year all-­cause mortality. Of two studies50 53 inves- QUIPS. Over half of the tigating this question, only one50 (6134 analysed patients) studies43 45 46 48–50 54 were at low risk for study participa- provided adjusted estimates reporting as Cox propor- tion, study attrition, and outcome measurement domains. tional hazard regression with OR (95% CI). We were While three studies51 52 55 described baseline characteris- unable to get further clarification from the study authors; tics inadequately, and another two44 47 provided insuf- therefore, we considered this a misspelling error, and so ficient data on drop-­ outs. All studies were at unclear we transformed their estimate (assumed HR) into OR. risk for the prognostic factor domain, given that none This study showed lower 1-­year all-­cause mortality in the defined sex. The risk of bias for the adjustment for female group (OR 0.83, 95% CI 0.68 to 0.98; low-­certainty other prognosis factors domain was low for half of the of evidence). Considering a risk of 50.5% for 1-­year all-­ studies43 44 47 52 54 55 and moderate for the others45 46 48–51 cause mortality in male patients, 46 fewer female patients because of an acceptable or minimal adjustment, respec- per 1000 will die (95% CI from 95 to 5 fewer), as compared tively. Three studies45 50 55 were at unclear risk for the statis- with male patients. Sensitivity analysis results using unad- tical analysis and reporting domain, while the remaining justed estimates were inconclusive (OR 0.86, 95% CI 0.54 studies were at low risk of bias. to 1.37) (online supplemental figure 5). Evidence synthesis We evaluated sex-­related all-­cause ICU mortality. We Online supplemental table 9 presents the summary found seven studies43 46–48 51 53 54 (51 936 recruited partic- outcome estimates for each study. Table 3 displays ipants) addressing this question. Five studies43 46 48 51 54 ‘Summary of findings’ for each review outcome. (31 562 analysed participants) provided adjusted estimates. One of them48 reported adjusted OR stratified by age, Primary outcomes and after failing to get an overall adjusted estimate from We investigated the independent prognostic effect the study author, we considered it as two substudies. of sex on all-­ cause hospital mortality. We found Pooled adjusted estimates found inconclusive results on seven studies43–45 47 50 53 55 (38 016 recruited partic- sex-­based differences in all-­cause ICU mortality (OR 1.19, ipants) addressing this question. Among the five 95% CI 0.79 to 1.78; I2=69%; very low-­certainty evidence) studies43–45 47 55 (30 349 analysed participants) that (online supplemental figure 6). The 95% prediction provided adjusted results, four of them43 44 47 55 (28 915 interval ranged from 0.49 to 2.89. Results of analyses analysed participants) presented sufficiently similar data comparing subgroups by longitudinal designs showed no allowing quantitative synthesis. Meta-­ analysis showed differences (p=0.83). Sensitivity analysis results including inconclusive results on sex-­based differences in all-­cause only studies with low or uncertain risk of bias for all key 4 Antequera A, et al. BMJ Open 2021;11:e048982. doi:10.1136/bmjopen-2021-048982 Table 2 Characteristics of included studies Sample size N of study participants Study Study dates Study design Sites Population Primary outcome (N with outcome) Inclusion criteria Exclusion criteria 43 Adrie et al 2007 1997–2005 Prospective 12 Adults admitted to the ICU ICU mortality 1692 (1608) >16 years old; ICU NS nested case– for severe community-­ Post-­ICU mortality stays >24 hours; control acquired sepsis community-­acquired severe sepsis Caceres et al 201344 2006–2007 Retrospective 4 Adults admitted to the All-­cause mortality 416 (319) ≥18 years old; ICU None cohort ICU for hospital-­acquired admission; clinical pneumonia suspicion of pneumonia Dara et al 201255 1998–2007 Retrospective 28 Adults admitted to the Hospital mortality 8670 (8670) Consecutive adults with NS cohort ICU for septic shock septic shock patients Luethi et al 201048 2008–2014 Post hoc analysis 51 Adults presented to the 90-­day all-­cause 1387 (1387) ≥18 years old; septic NS of an RCT ED with septic shock. illness severity-­ shock Data were available for adjusted mortality ICU setting Madsen et al 201445 2005–2012 Retrospective 1 Adults admitted to the SSC resuscitation 814 (814) >18 years old presenting Only comfort cohort ICU for severe sepsis or bundle completion to the ED with criteria measures within the septic shock for severe sepsis/septic first 24 hours; non-­ Antequera A, et al. BMJ Open 2021;11:e048982. doi:10.1136/bmjopen-2021-048982 shock ICU admission Mahmood et al 201251 2004–2008 Retrospective NS* Adults admitted to the ICU mortality 27 935 (27 935) Consecutive adults in the Readmission to the cohort ICU (sepsis subgroup) APACHE IV database; ICU sepsis subgroup Nachtigall et al 201146 January/March Prospective cohort 1 Adults admitted to mixed ICU mortality 327 (327) Consecutive adults NS 2006; February/ ICUs with a special focus (≥18 years); ICU May 2007 on sepsis patients (sepsis stays >36 hours; sepsis subgroup) criteria for at least 1 day during the ICU stay Pietropaoli et al 2003–2006 Retrospective 98 Adults admitted to the Hospital mortality 18 757 (18 318) ≥16 years old; severe If gender, age, or 201047 cohort ICU for severe sepsis or sepsis/septic shock hospital mortality septic shock patients; data from the was missing first ICU admission Sakr et al 201354 April/Sep Post hoc analysis 24 Adults admitted to the ICU mortality 305 (305) >18 years old; severe NS 200614 of a prospective medical and/or surgical sepsis; data from the first cohort ICU for severe sepsis ICU admission Samuelsson et al 2008–2012 Retrospective 65 Adults admitted to the 30-­day mortality 9830 (9830) Consecutive SAPS III– Reasons for not 201552 cohort ICU (sepsis subgroup) scored adults ICU (>15 being able to obtain years old); validated mortality data: non-­ mortality data in the Swedish residency registry; sespsis subgroup and patients with concealed identity Continued Open access 5 BMJ Open: first published as 10.1136/bmjopen-2021-048982 on 22 September 2021. Downloaded from http://bmjopen.bmj.com/ on December 21, 2022 by guest. Protected by copyright. 6 Open access Table 2 Continued Sample size N of study participants Study Study dates Study design Sites Population Primary outcome (N with outcome) Inclusion criteria Exclusion criteria Sunden-­Cullberg et al 2008–2015 Retrospective 42 Adults admitted to the Sepsis bundle 2720 (2430) ≥18 years old; ICU Data non-­registered 202049 cohort ICU for sepsis or shock completion; 30-­day admission within 24 hours simultaneously septic via the ED within mortality of arrival to an ED; in two selected 24 hours community-­acquired registries, severe sepsis or septic alongside SAPS3 shock data. Multiple registrations. van Vught et al 201753 2011–2014 Prospective cohort 2 Adults admitted to the 90-­day mortality 1533 (1815 Consecutive patients Transfer from other ICU for sepsis admissions†) >18 years old; sepsis; ICUs expected ICUs stay >24 hours; data from multiple ICU admission‡ Xu et al 201950 2001–2012 Retrospective 1 Adults admitted to the 1 year mortality 6134 (6134) All adults diagnosed with <18 years old cohort ICU for sepsis sepsis, severe sepsis, or septic shock in the database *Information reported as ‘large number of ICUs’. †van Vught analysed 1815 admissions for its primary outcome. Data were available at the patient level for the review outcomes. ‡ICU demographic and long-­term follow-­up data from the first ICU admission, host response data from overall admissions. APACHE, Acute Physiologic Assessment and Chronic Health Evaluation; ED, emergency department; ICU, intensive care unit; NS, not stated; RCT, randomised controlled trial; SAPS, Simplified Acute Physiology Score; SSC, surviving sepsis campaign. Antequera A, et al. BMJ Open 2021;11:e048982. doi:10.1136/bmjopen-2021-048982 BMJ Open: first published as 10.1136/bmjopen-2021-048982 on 22 September 2021. Downloaded from http://bmjopen.bmj.com/ on December 21, 2022 by guest. Protected by copyright. Open access BMJ Open: first published as 10.1136/bmjopen-2021-048982 on 22 September 2021. Downloaded from http://bmjopen.bmj.com/ on December 21, 2022 by guest. Protected by copyright. Table 3 Summary of findings Anticipated absolute prognostic effects* Effect estimate (95% CI) Certainty of the Assumed risk Risk in females ARD in females (95% prediction No of participants evidence Outcomes in males (95% CI) (95% CI)† interval) (studies) (GRADE) All-­cause hospital 303 per 1 000‡ 307 per 1 000 4 more per 1000 OR 1.02 28 915 ⨁◯◯◯ mortality (median (255 to 364) (47 fewer to 62 (0.79 to 1.32) (4 observational phase VERY LOW§¶** observed length of more) (0.5 to 2.08) 2 studies) stay ranged from 6 to 26 days) 28-­day all-­cause 240 per 1 000‡ 271 per 1 000 31 more per OR 1.18 12 579 ⨁◯◯◯ mortality (249 to 294) 1000 (1.05 to 1.32) (3 observational phase VERY (9 more to 54 (0.56 to 2.50) 2 studies) LOW§**††‡‡ more) 1-­year all-­cause 505 per 1 000‡ 459 per 1 000 46 fewer per OR 0.83 6134 ⨁⨁◯◯ mortality (410 to 500) 1000 (0.68 to 0.98) (1 observational phase LOW**††§§¶¶ (95 fewer to 5 N/M 2 study) fewer) All-­cause ICU 200 per 1 000‡ 229 per 1 000 29 more per OR 1.19 31 562 ⨁◯◯◯ mortality (167 to 308) 1000 (0.80 to 1.78) (5 observational phase VERY LOW§¶** (median observed (33 fewer to 108 (0.49 to 2.89) 2 studies) length of stay more) ranged from 2.7 to 13 days) Not meaningful: <3 studies for computing of the 95% prediction interval a meaningful estimate. *The risk in the female group (and its 95% CI) is based on the assumed risk in the male participants group and the estimated effect of sex (OR and its 95% CI). †We considered an ARD of at least ±10‰ as large enough to be clinically meaningful. Thus, we defined the clinical importance of the absolute prognostic effect for all the review outcomes as follows: important improvement (ARR of at least 10‰), slight improvement (10‰