Montané Eva (Orcid ID: 0000-0002-1691-7652) Epidemiology of drug-related deaths in European hospitals: a systematic review and meta-analysis of observational studies Authors: Montané Eva, Castells Xavier. Eva Montané, MD, PhD. Department of Clinical Pharmacology, Hospital Universitari Germans Trias i Pujol, Badalona, Barcelona, Spain. Department of Pharmacology, Therapeutics and Toxicology, Universitat Autònoma de Barcelona, Barcelona, Spain. ORCID number: 0000-0002-1691-7652. Xavier Castells, MD, PhD. Department of Medical Sciences, University of Girona, Girona, Spain. ORCID number: 0000-0002-2619-7273. Corresponding author: Dr. Eva Montané. Carretera de Canyet s/n, 08916 Badalona, Barcelona, Spain. Telephone: (+34) 934978865. Fax: (+34) 934978864. E-mail address: emontane.germanstrias@gencat.cat Short title: Meta-analysis of drug-related deaths Keywords: adverse drug reaction, hospital mortality, incidence/prevalence, epidemiology, meta- analysis, observational studies This article has been accepted for publication and undergone full peer review but has not been through the copyediting, typesetting, pagination and proofreading process which may lead to differences between this version and the Version of Record. Please cite this article as doi: 10.1111/bcp.14799 This article is protected by copyright. All rights reserved. Contributions: Dr. Montané: conceived the idea, contributed to the study design, conducted the literature search, reviewed the selected cases, undertook data acquisition of included cases, performed analysis and interpretation of the data, and wrote the first draft of manuscript. Dr. Castells: contributed to study design, review of the selected cases, data acquisition of included cases, analysis and interpretation of the data, and writing the manuscript. Word count: 5,721 This article is protected by copyright. All rights reserved. ABSTRACT Objectives: To perform a systematic review of observational studies on the epidemiology of drug- related death (DRD) in patients requiring hospitalisation or while hospitalised (hospital-acquired DRD). Methods: We conducted a systematic review of observational studies investigating the occurrence rate of DRD episodes among deceased inpatients. Two independent researchers assessed eligibility criteria, extracted data, and evaluated the risk of bias. Both quality assessment and meta-analysis were performed. Results: From 1,351 identified potential studies, six retrospective studies were included. DRD occurrences rates were 7.3% (CI 95% 4.1 – 12.5) among deceased inpatients and 0.13% (CI 95% 0.04 – 0.40) among hospitalised patients. During hospitalisation, acquired-DRD represented 2.7% (CI 95% 1.0 – 6.9) of inpatient deaths and occurred in 0.05% (CI 95% 0.01 – 0.23) of hospitalised patients. However, these estimates have to be viewed with caution because there was significant heterogeneity (I2 >97%). None of the studies were considered to be at ‘high risk of bias’ according to the criteria of the NIH Quality Assessment Tool. The most common ADRs related to death were haemorrhages due to antithrombotic drugs (39%, CI 95% 26.5 – 53.2) and infections in drug- immunosuppressed patients (27.5%, CI 95% 16.7 – 41.7). Conclusions: We found that the DRD occurrence rate of deceased hospital inpatients has been infrequently studied in Europe. Our findings suggest that drugs are an important cause of death in hospitals. The limited number of studies in European countries stresses the need for more research in this area. This article is protected by copyright. All rights reserved. 1. INTRODUCTION Adverse drug reactions (ADRs) are a known problem in terms of morbidity, costs and mortality [1,2]. In Europe, an estimated 5% of all hospital admissions are due to ADRs, and ADRs are the fifth most common cause of hospital death [3]. Extrapolated data from a meta-analysis revealed that in Europe about 197,000 deaths occur annually due to ADRs [4]. Several studies and some meta-analyses have been published in different settings or populations, mainly focusing on assessing the incidence of ADRs leading to hospitalisation [5-10]. Regarding the epidemiology of fatal ADRs among hospitalised patients, a meta-analysis conducted by Lazarou et al. reported an incidence of 0.32% in the USA [4] and a review conducted by Bouvy et al. reported an incidence of 0.5% in Europe [11]. Recently, Patel et al. published a meta-analysis including 48 international studies showing an incidence of fatal ADRs of 0.2% [12]. Differences in the prevalence of diseases that impact drug consumption could result in large differences in drug-related death (DRD) incidences between low- and high- income countries (0.3% vs 3%) [4,13]. The occurrence rate of DRD in hospitalised patients is an important epidemiological indicator of mortality caused by drugs. Recently Montané et al. [14] found that DRDs among deceased inpatients was 7%. This is, less than half of Pardo Cabello’s finding [15], despite both studies being from Spain. Meta-analysis in the context of a systematic review is the most reliable method for understanding the findings from studies with different and sometimes conflicting results. We performed a systematic review and meta-analysis aiming to assess the incidences and occurrence rates of DRD in hospital settings from studies performed in European countries and studying the sources of between-study variability of results. We limited these to this specific geographic area in order to obtain more homogenous data taking into account the fact that European countries have quite similar health systems, with comparable disease prevalence and patterns of drug consumption [16,17]. To our knowledge, no meta-analysis has studied the incidence of DRD among deceased inpatients so far. The approach of this meta-analysis differs from others which included studies evaluating ADRs This article is protected by copyright. All rights reserved. that led to hospital admission or occurred during hospitalisation, regardless of whether the outcome was fatal or not. 2. MATERIALS AND METHODS This systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement [Prisma statement] [18]. The study was registered in PROSPERO International prospective register of systematic reviews in December 2018. Registration number: CRD42019105618 [19]. 2.1 Data sources and search strategy Three databases: PubMed, Web of Science, and SCOPUS were searched from inception to July 2019. Search strings were constructed using Boolean operators (AND, OR) and combining keywords. The following syntax was used in each database: (((adverse drug effect) OR "adverse drug reaction") AND "Hospital Mortality" AND (incidence OR prevalence)), without language restriction (Electronic Supplementary Material [ESM] 1). Furthermore, the citations and references of the included studies and review articles were screened for relevant articles. 2.2 Study selection criteria The inclusion criteria were observational studies that assessed the DRD occurrence rate among all deceased inpatients as a main study objective, providing epidemiological data (such as the number of admitted patients and the number of deceased patients in the hospital) or data allowing their calculation, and performed in European countries. This article is protected by copyright. All rights reserved. The exclusion criteria were clinical trials, review articles, systematic reviews, meta-analyses, commentaries, and editorials, as well as studies with other designs (e.g. spontaneous reporting or case-reports). No studies assessing mortality rate among patients treated in specific hospital services or with specific drugs or drug-classes were included. Studies performed in Asian, American, Oceanian or African countries were also excluded from the study. Titles and/or abstracts of studies retrieved using the search strategy and those from additional sources were screened by the primary author to identify studies that potentially met the inclusion criteria. The full text of these potentially eligible studies was retrieved and independently assessed for eligibility by the same researcher. Doubts about any particular study were settled after a thorough discussion with the second author. 2.3 Data extraction Both authors retrieved the study data independently. A standardised, pre-piloted form was used to extract data from the included studies for assessment of study quality and evidence synthesis. Extracted information included: year of publication, country where the study was conducted, study setting, study period, study design, study methodology, ADR definition, causality assessment tool, study population with demographics of participants and baseline characteristics, incidences of DRD and hospital-acquired DRD among inpatients and overall deceased patients, characteristics of suspected drugs (using the Anatomical Therapeutic Chemical [ATC] classification [20]), drug-drug interactions, characteristics of ADR, median hospital stay, risk factors and preventability of ADR. Discrepancies were resolved through discussion between the researchers. Missing relevant and necessary data were requested from authors of the study. This article is protected by copyright. All rights reserved. 2.4 Methodological quality To assess the risk of bias in individual studies, two raters (EM and XC) independently assessed the methodological quality using the NIH Quality Assessment Tool for Observational Cohort and Cross- Sectional Studies [21]. This quality tool comprises 14 items or criteria and guidance for rating. Six of the 14 criteria (43%) were not applicable due to the design of the studies (retrospective cohorts), these were: items 4, 5, 7, 10, 12 and 13 (table 1). The quality score was calculated by dividing the number of criteria that were present (“yes”) by the total number of applicable criteria. A quality of 100% was obtained if all eight applicable criteria were present in the study and studies were considered to be at high risk of bias if they achieved a score lower than 50%. 2.5 Data analysis The occurrence rate of DRD among deceased inpatients was calculated using the number of DRDs as the numerator and the total number of deceased inpatients during the study period as the denominator; for the incidence of DRD among hospitalised patients, the number of DRDs was used as the numerator and the total number of hospital admissions during the study period as the denominator. The rate of hospital-acquired DRD was calculated using the number of hospital- acquired DRDs as the numerator and the total number of DRDs as the denominator. For the incidence of hospital-acquired DRD among hospitalised patients, the number of hospital-acquired DRDs was used as the numerator and the total number of hospital admissions during the study period as the denominator. As the most common DRDs have been found to be haemorrhages and infections in drug- immunosuppressed patients [14,22,23], we assessed the occurrence rate of these specific type of DRD which were calculated by dividing the number of patients dying from each specific types of DRD by the total number of patients dying from DRDs. Both internal and external haemorrhages were This article is protected by copyright. All rights reserved. considered. Infections were considered both when disseminated and also when affecting a single organ. 2.6 Statistical analysis Individual results were pooled using a random effects model. Statistical heterogeneity was determined by calculating the I² statistic, which measures the percentage of variance across studies due to heterogeneity rather than chance [24]. We initially planned to study the sources of statistical heterogeneity by means of meta-regression using patient characteristics and the type of causality scale administered as covariates. However, the final number of studies identified was too low for this to be undertaken. Publication bias was analysed by means of funnel plot, and Begg and Mazumdar rank correlation test and Egger’s test of the intercept [25,26]. Statistical analysis was achieved with comprehensive meta- analysis v3 [27]. 3. RESULTS 3.1 Study selection The database searches identified 1,351 potentially relevant citations. After duplicate removal, a total of 1,029 studies were reviewed, 1,016 (98.7%) were excluded after screening of the titles and abstracts. After full text review of the remaining 13 studies, five publications met the inclusion criteria. One article identified was an abstract published at the society's journal [28]; for which detailed data were available in a doctoral thesis published online [29].Thus, a total of six studies were included in the review (Figure 1: PRISMA Diagram) [14,15,29-32]. This article is protected by copyright. All rights reserved. 3.2 Study Characteristics and measures All the included studies were observational retrospective cohorts performed in the last 20 years (from 2000 to 2015) in University Hospitals of three European countries: France (one study), Finland (two studies) and Spain (three studies). The study period was 12 months in all studies except for one which lasted 22 months [15]. All the studies reported the number of DRDs and the number of deceased inpatients. All studies reported the number of hospital admissions except for one which reported an approximate value [31]. The number of hospital-acquired DRDs was reported in all studies but two (in one of which it could be calculated) [15,30]. Table 2 summarises the characteristics of the included studies. 3.3 Assessment of methodological quality On the basis of the NIH Quality Assessment tool for observational studies, no study scored under 50%, which was our pre-specified criterion for low quality. The quality score ranged from 75 to 100%. Three studies did not report details of exposures (item 8), such as doses, routes of administration and time of starting treatment. Descriptions of potential confounding variables measured and adjusted statistically (item 14) were also lacking in three studies (table 1). 3.4 ADR definition All studies but one used the “WHO” definition for an ADR as “a response to a drug that is noxious and unintended and that occurs at doses normally used in humans for prophylaxis, diagnosis or therapy of disease” [33]. The remaining study used the definition proposed in the latest European pharmacovigilance legislation (Directive of the European Parliament and of the Council 2010/84/EU) as “a response to a medicinal product which is noxious and unintended” [34]. The European definition widened the WHO definition of ADR (including medication errors, abuse, misuse as well as doses and indications not authorized for the regulatory agencies or “off label” uses). This article is protected by copyright. All rights reserved. 3.5 Pooled DRD incidences and occurrence rates A total of 676,548 patients were hospitalised over 2000-2015. Of the 7,578 patients who died, 657 died as a result of an ADR. The number of DRD per year ranged from 17 to 256 [15,32], the number of in-hospital deceased per year ranged from 820 to 1,708 [29,32], and the number of hospital admitted patients per year ranged from 21,483 to 400,000 [14,31]. The overall incidence of DRD among hospitalised patients was 0.13% (CI 95% 0.04 – 0.40), ranging from 0.01% to 0.74% (ESM 2). The occurrence rate of DRD among deceased inpatients was 7.3% (CI 95% 4.1 – 12.5), ranging from 3% to 18% (ESM 3). The incidence of hospital-acquired DRD among hospitalised patients was 0.05% (CI 95% 0.01 – 0.23, I2=98.9%) ranging from 0.03% to 0.4% (ESM 4). The occurrence rate of hospital- acquired DRD among deceased inpatients was 2.7%, (CI 95% 1.0 – 6.9, I2=97.6%) ranging from 0.6% to 11.5% (ESM 5) (table 3). As the final number of studies included in the statistical analysis was low, we did not determine the sources of between-study variability by means of meta-regression. 3.6 Causality assessment All the studies reported the tool used for causality assessment; three studies used two different causality tools [14,29,32]. The WHO-UMC criteria and the Naranjo algorithm were the most frequently used tools [35,36]. The French algorithm and the Wulff criteria were the other tools used for causality assessment in one study each [37,38] (table 2). In two studies using the WHO-UMC criteria, cases were included only if DRD attribution causality was ‘certain’ or ‘probable’, excluding all ‘possible’ DRD cases [30,31]. In the other two studies using the Naranjo algorithm, DRD were ‘probable’ or ‘possible’ related to drugs in 100% and 98,7% of patients respectively [14,29]. Finally, the study using the Wulff criteria, DRDs were classified as ‘suspected of causing death’ or ‘suspected of contributing to deaths’. This article is protected by copyright. All rights reserved. 3.7 Characteristics of patients with DRD Patients with DRDs were predominantly men (mean proportion of 61.2%, ranging from 50% to 73%), with a mean age of 74.2 years (ranging from 65 to 77 years). For studies that provided data (all except two), the mean hospitalisation stays ranged from 8 to 16 days (table 2). 3.8 Adverse drug reactions Haemorrhages were the most common ADR with proportion of 39.0% (CI 95% 26.5 – 53.2), ranging from 19 to 71% of DRDs (ESM 6). Infections in drug-immunosuppressed patients were the second leading cause of DRD, with a proportion of 27.5% (CI 95% 16.7 – 41.7), ranging from 16% to 44% (ESM 7). Another commonly reported cause of DRD in one study was cardiac arrhythmia (14%) (table 2). 3.9 Suspected drugs Cytostatic agents, classified as group L (Antineoplastic and immunomodulating agents), and antithrombotic agents, classified as group B (Blood and blood forming organs), were the most commonly involved drug type. Drug-drug interactions were reported in four studies. The proportion of drug-drug interactions was 37.4% of DRDs (CI 28.3 – 47.6); ranging from 30% to 64% (ESM 8) (table 1). All except one of the DRDs due to drug-drug interactions, were due to pharmacodynamic and synergistic interactions. 3.10 Avoidability assessment Only two studies assessed potential avoidability of DRDs, using two different tools: Olivier score [39] and Schumock Thornton criteria [40]. One study [29] found that the main reasons for causing preventable DRDs were that drugs were not appropriate for the patient’s condition and that the This article is protected by copyright. All rights reserved. duration of treatment was too long. The other study [14] found that the drug-drug interactions involved in the DRDs were the main cause of avoidability. Proportions of DRD avoidability were 34% and 47%, respectively (table 2). 3.11 Risk factors assessment Three (50%) studies reported the assessment of risk factors for DRDs. The number of administered drugs and the effect of comorbidity were the most consistent findings of these studies, with positive associations between these risk factors and the rate of DRD. The number of administered drugs was assessed using mean, median and ≥10 drugs, depending on the study. The comorbidity was assessed using the Charlson score in two studies and having ≥4 diseases in one study. Mixed results were found for gender and age, with two studies finding no association with DRD and another finding that men and younger patients have a higher risk of DRD. 3.12 Publication bias All funnel plots were reasonably symmetrical (ESM 9 to 12) and neither Egger nor Begg tests were statistically significant for any outcome. 4. DISCUSSION In this meta-analysis, we investigated the epidemiology of DRDs in the hospital setting. We found that the DRD occurrence rate of deceased hospital inpatients has been infrequently studied. All the included studies were performed in the last two decades. Overall, DRD occurrence rate of deceased hospital inpatients was 7.3%, and more than a third were hospital-acquired. Both rates are high and denote that is a relevant healthcare problem in European countries. Another important finding of this meta-analysis is the DRD incidence among hospitalised patients, which was 0.1%. Other studies with different methodology, such as a European review and a meta-analysis including worldwide This article is protected by copyright. All rights reserved. studies reported higher incidences, which were 0.5 and 0.2% respectively [11,12]. The incidence of hospital-acquired DRD among hospitalised patients in our study was almost 0.1%. This result is similar to that of the recently published meta-analysis by Patel et al., despite differences in the study design, which included cross-sectional studies [41]. Although all the included studies had a similar study design and length, there was high heterogeneity in the incidence results due to variability in the methodological characteristics, as follows: Selection and assessment of deceased patients were not well reported in all studies. When they were, in one study a preselection was carried out (from predefined diagnoses of death suggesting ADR) leading to an underestimation of DRD incidence [14]. Even though most studies used WHO’s definition for ADR, there was significant variability in causality assessment of DRD cases, using at least four different causality methods, which could lead to differences in ADR attribution. Moreover, in two studies [30,31], the DRD cases were included only if attribution causality was certain or probable, excluding possible DRD cases, thus possibly leading to an underestimation of the number of DRD cases. A clear example of this is a cerebral haemorrhage in an anticoagulated patient with a preceding trauma, which could be classified as a possible DRD or as a probable DRD with a contributing role, depending on the study. There was also great variability in the sample size of studies, with the number of patients admitted and the number of patients assessed, varying from 300 to almost 2,000. Demographic characteristics of patients with DRD showed a predominance of elderly men, which surprisingly diverged from other studies where the proportion of women is higher [23]; although elderly people are also the group with higher DRD incidences [41,42]. This could be due to elderly patients often being polymedicated, as well as having other comorbidities and pharmacokinetic changes that facilitate the appearance of ADRs and fatal outcomes [43-45]. This article is protected by copyright. All rights reserved. Regarding the characteristics of ADR, there was also a high variability in the reporting of the types of ADRs identified, making a summarized report problematic. The two most common ADRs related to death were haemorrhages and infections in drug-immunosuppressed patients, which was similar to other meta-analyses or population-based studies [12,23]. Both these ADRs are type A, that is, dose dependent, augmented pharmacological effect, predictable and potentially avoidable reactions [46]. For cerebral haemorrhages in patients with head trauma or high blood pressure, antithrombotic drugs could be a contributing factor. Accordingly, two pharmacologic groups were responsible for most ADR, chemotherapy (group L) and antithrombotic agents (group B), which was similar to the findings of other meta-analyses [12,41,42]. Chemotherapy-induced neutropenia and subsequent infections were an important cause of mortality, where medulla aplasia is a pharmacologic effect often necessary for chemotherapy effectiveness, but which can lead to patient death. Antithrombotic agents were associated with a significant risk of haemorrhage. Contrary to other studies, non-steroidal anti-inflammatory agents (NSAIDs) and cardio-vascular drugs (such as digoxin, diuretics or renin-angiotensin system inhibitors) were not frequently involved in DRD in this study [12,41]. Moreover, nervous system agents and antiinfectives were frequently involved in DRDs in studies assessing data from voluntary reporting registers of ADR but not in our study.[47,48]. This difference could be explained by a bias in voluntary registers, where type A and well-known ADRs are underreported, making voluntary registers not useful for calculating ADR prevalences. Drug-drug interactions were present in more than one third of deceased patients, mainly due to pharmacodynamic interactions. In chemotherapy, combining drugs that have similar pharmacologic effect but different mechanisms of action is common and necessary to achieve the desired immunosuppressant effect. A common synergic drug-drug interaction is a cytotoxic agent with a corticosteroid [49]. Regarding antithrombotic agents, they are sometimes prescribed by different medical specialists for different diseases. A typical example of synergic drug-drug interaction is when This article is protected by copyright. All rights reserved. patients receive AAS for ischemic cardiopathy, cilostazol for peripheral arteriopathy and anticoagulation for atrial fibrillation. Avoidability of DRD was not widely assessed in the included studies, but when it was, rates were higher than in other studies which also assessed fatal drug poisoning [50]. This could be explained as the tools used differed, as well as the population studied. The risk factors related to DRD were not systematically assessed. The number of drugs and comorbidities were the most commonly reported risk factors related to DRD in this and other meta- analyses [51]. Conversely, older age and female gender, despite having been associated with an increased risk for ADR-related hospitalisation [52], were not associated with DRD in two studies included in this meta-analysis and were associated with a reduced risk of DRD in another. Differences in how the effect of confounding covariates were analysed may explain the discrepancies between these findings. We limited our meta-analysis to studies conducted in European countries in order to have greater homogeneity of data. Surprisingly, all the included studies were single-centre and were performed in only three European countries and four university hospitals leading to low representativeness of the obtained results. In Spain and Finland, researchers compared the studies’ results at two different periods of time. This review clearly shows that few studies have been conducted using this epidemiological methodology which assess DRD cases among deceased inpatients. Therefore, multicenter studies conducted in several European countries assessing the epidemiology of DRD are needed to gain better knowledge of DRD incidences among deceased inpatients in this geographic area. In the same way, in low or middle-income countries, epidemiological DRD data are also scarce. The results of a study conducted in South Africa with a methodology similar to the studies included in this review, showed that epidemiological incidence data for DRD of inpatients were extremely This article is protected by copyright. All rights reserved. high; and antiinfectives for the management of HIV and tuberculosis were the drugs most commonly implicated in DRD [13]. Strengths and Limitations The main limitation of this meta-analysis is that it was limited to studies assessing incidence of DRD among deceased inpatients. Therefore, studies which assessed fatal ADR as part of a broader study investigating all ADRs in hospitalised patients were excluded. Other study limitations are related to the quality of the studies included, which, in general, were deemed to have low risk of bias. Large statistical heterogeneity was found, which reduces confidence in the incidences and rates calculated. Reporting bias can threaten the validity of any meta-analysis. However, no evidence of small study effects was suggested by the funnel plots, nor by the Egger and Begg tests, indicating that publication bias is unlikely. Nevertheless, it must be stressed that the validity of these plots and tests is low when the number of studies included in the meta-analysis is, like in our review, also low [53]. The restriction of including studies conducted in Europe limits generalisability. A strength of this study is that we performed a systematic review using the most important data bases, searching for a long period of time and with no language restriction, which allowed us to include as many studies as possible. Another relevant strength of the present study was the fact that we contacted the authors of the studies and hence were able to include one more study and calculate the maximum of the DRD and hospital-acquired DRD occurrence rates, as well as DRD and hospital-acquired DRD incidences to meta-analyse. No studies were excluded for not providing enough data to calculate incidence or occurrence rates. CONCLUSION Our findings suggest that in European hospitals drugs are an important cause of death, occurring in 1 out of 1,000 patients admitted to hospital, and in at least 1 in 14 deceased patients, denoting that is This article is protected by copyright. All rights reserved. a relevant healthcare issue. Despite the similarity of the design of the studies included, high heterogeneity was found. The number of drugs and comorbidities were the most common risk factors related to DRD. The limited number of studies in European countries highlights the need for more research in this area. Acknowledgements We would like to thank the authors of the included studies: Professor Pertti Neuvonen and Dr. Outi Lapatto-Reiniluoto from Finland, Dr. Puche from Spain, and Dr. Grévy from France for their help providing the lacking study data. The authors sincerely appreciate the help of Alethea Charlton for reviewing the language of this manuscript. Compliance with Ethical Standards Funding: This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Conflicts of Interest: Montané E and Castells X declare that they have no conflict of interest. Ethical approval: For this type of study ethical approval is not required. Patient consent: For this type of study formal consent is not required. This article is protected by copyright. All rights reserved. REFERENCES 1. Sultana J, Cutroneo P, Trifirò G. Clinical and economic burden of adverse drug reactions. J Pharmacol Pharmacother. 2013;4(Suppl 1):S73–S77. doi:10.4103/0976-500X.120957. 2. Watanabe JH, McInnis T, Hirsch JD. Cost of Prescription Drug-Related Morbidity and Mortality. Ann Pharmacother. 2018;52(9):829–837. doi:10.1177/1060028018765159. 3. Strengthening pharmacovigilance to reduce adverse effects of medicines. MEMO/08/782. Brussels, December – November 2008. Available at http://europa.eu/rapid/press-release_MEMO- 08-782_en.htm (last accessed 23 Sep 2020). 4. Lazarou J, Pomeranz BH, Corey PN. Incidence of adverse drug reactions in hospitalized patients. A meta-analysis of prospective studies. JAMA 1998;279:1200–5. 5. Taché SV, Sönnichsen A, Ashcroft DM. Prevalence of adverse drug events in ambulatory care: a systematic review. Ann Pharmacother. 2011;45(7-8):977-989. doi:10.1345/aph.1P627. 6. Lisha J, Annalakshmi V, Maria J, Padmini D. Adverse Drug Reactions in Critical Care Settings: A Systematic Review. Curr Drug Saf. 2017;12(3):147-161. doi:10.2174/1574886312666170710192409 7. Oscanoa TJ, Lizaraso F, Carvajal A. Hospital admissions due to adverse drug reactions in the elderly. A meta-analysis. Eur J Clin Pharmacol. 2017;73(6):759-770. doi:10.1007/s00228-017-2225-3. 8. Smyth RM, Gargon E, Kirkham J, et al. Adverse drug reactions in children--a systematic review. PLoS One. 2012;7(3):e24061. doi:10.1371/journal.pone.0024061 9. Impicciatore P, Choonara I, Clarkson A, Provasi D, Pandolfini C, Bonati M. Incidence of adverse drug reactions in paediatric in/out-patients: a systematic review and meta-analysis of prospective studies. Br J Clin Pharmacol. 2001;52(1):77-83. doi:10.1046/j.0306-5251.2001.01407.x 10. Davies EC, Green CF, Taylor S, Williamson PR, Mottram DR, Pirmohamed M. Adverse drug reactions in hospital in-patients: a prospective analysis of 3695 patient-episodes. PLoS One. 2009;4(2):e4439. doi:10.1371/journal.pone.0004439 11. Bouvy JC, De Bruin ML, Koopmanschap MA. Epidemiology of Adverse Drug Reactions in Europe: A Review of Recent Observational Studies. Drug Saf 2015;38:437–53. 12. Patel TK, Patel PB. Mortality among patients due to adverse drug reactions that lead to hospitalisation: a meta-analysis. Eur J Clin Pharmacol. 2018;74(6):819-832. doi: 10.1007/s00228-018- 2441-5. 13. Mouton JP, Mehta U, Parrish AG, Wilson DP, Stewart A, Njuguna CW, et al. Mortality from adverse drug reactions in adult medical inpatients at four hospitals in South Africa: a cross-sectional survey. Br J Clin Pharmacol. 2015;80(4):818-26. 14. Montané E, Arellano A L, Sanz Y, Roca J, Farré M. Drug-related deaths in hospital inpatients: A retrospective cohort study. Br J Clin Pharmacol. 2018;84(3):542-552. This article is protected by copyright. All rights reserved. 15. Pardo Cabello AJ, Del Pozo Gavilán E, Gómez Jiménez FJ, Mota Rodríguez C, Luna Del Castillo J de D, Puche Cañas E. Drug-related mortality among inpatients: a retrospective observational study. Eur J Clin Pharmacol. 2016;72(6):731-6. 16. Gaeta M, Campanella F, Capasso L, et al. An overview of different health indicators used in the European Health Systems. J Prev Med Hyg. 2017;58(2):E114–E120. 17. Simó Miñana J. Utilización de medicamentos en España y en Europa [Use of prescription drugs in Spain and Europe]. Aten Primaria. 2012;44(6):335–347. doi:10.1016/j.aprim.2011.06.009 18. Moher D, Liberati A, Tetzlaff J, Altman DG; PRISMA Group. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med. 2009;6(7):e1000097. doi: 10.1371/journal.pmed.1000097 19. PROSPERO: International prospective register of systematic reviews: http://www.crd.york.ac.uk/PROSPERO/display_record.php?ID=CRD42019105618 (last accessed 23 Sep 2020). 20. WHO Collaborating Centre for Drug Statistics Methodology, Guidelines for ATC classification and DDD assignment 2019. Oslo, Norway, 2018. Available at: http://www.whocc.no (last accessed 23 Sep 2020). 21. National Heart Lung and Blood Institute. Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies - NHLBI, NIH. National Institutes of Health. Available at https://www.nhlbi.nih.gov/health-topics/study-quality-assessment-tools. (last accessed 23 Sep 2020). 22. Buajordet I, Ebbesen J, Erikssen J, Brørs O, Hilberg T. Fatal adverse drug events: the paradox of drug treatment. J Intern Med. 2001;250(4):327-41. 23. Wester K, Jönsson AK, Spigset O, Druid H, Hägg S. Incidence of fatal adverse drug reactions: a population based study. Br J Clin Pharmacol. 2008;65(4):573–579. doi:10.1111/j.1365- 2125.2007.03064.x 24. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ. 2003;327, 557-60. 25. Begg CB, Mazumdar M. Operating characteristics of a rank correlation test for publication bias. Biometrics. 1994;50(4):1088–1101. 26. Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ. 1997;315(7109):629–634. doi:10.1136/bmj.315.7109.629 27. Borenstein M, Hedges L, Higgins J, Rothstein H. (2014). Comprehensive meta-analysis version 3.3.070. Englewood, NJ: Biostat. 28. Grévy A, Lepelley M, Khouri C, Mallaret M, Roustit M. Drug-related deaths: retrospective analysis in French Universitary Hospital. Clin Ther 2017;39(8):Suppl e93-94. https://doi.org/10.1016/j.clinthera.2017.05.293 This article is protected by copyright. All rights reserved. 29. Grévy A. Décès imputables aux médicaments: analyse rétrospective au CHU de Grenoble-Alpes en 2014. Sciences pharmaceutiques. 2017. ⟨dumas-01719243⟩. Available at https://pdfs.semanticscholar.org/8339/ae2e3f19875bef471805976e10e82eb73efd.pdf?_ga=2.5091 3321.1629777483.1564047146-1795757843.1562769786 (last accessed 23 Sep 2020). 30. Juntti-Patinen L, Neuvonen PJ. Drug-related deaths in a university central hospital. Eur J Clin Pharmacol. 2002;58(7):479-82. 31. Lapatto-Reinilouto O, Patinen L, Niemi M, Backman JT, Neuvonen PJ. Drug-Related Inadvertent Deaths in a University Hospital--A Declining Trend. Basic Clin Pharmacol Toxicol. 2015;117(6):421-6. 32. Pardo Cabello AJ, González Contreras LG, Manzano Gamero MV, Gómez Jiménez FJ, Puche Cañas E. Prevalence of fatal adverse drug reactions in hospitalized patients. Int J Clin Pharmacol Ther. 2009;47(10):596-602. 33. World health organization technical report series (1972, no 498). International Drug Monitoring. The Role of the National Centres. Report of a WHO Meeting. Geneva, Switzerland. Available at http://apps.who.int/iris/bitstream/10665/40968/1/WHO_TRS_498.pdf (last accessed 23 Sep 2020). 34. Commission Directive 2010/84/EU of the European Parliament and the Council of 15 December 2010 amending as regards pharmacovigilance, Directive 2001/83/EC on the Community code relating to medicinal products for human use. Official Journal of the European Union 31.12.2010: L348/74-L348/99. http://ec.europa.eu/health/files/eudralex/vol- 1/dir_2010_84/dir_2010_84_en.pdf (last accessed 23 Sep 2020). 35. World Health Organization. The use of the WHO-UMC system for standardized case causality assessment. Uppsala: The Uppsala Monitoring Centre. https://www.who- umc.org/media/164200/who-umc-causality-assessment_new-logo.pdf (last accessed 23 Sep 2020). 36. Naranjo CA, Busto U, Sellers EM, Sandor P, Ruiz I, Roberts EA, et al. A method for estimating the probability of adverse drug reactions. Clin Pharmacol Ther.1981;30(2):239-45. 37. Bégaud B, Evreux JC, Jouglard J, Lagier G. Imputation of the unexpected or toxic effects of drugs. Actualization of the method used in France. Therapie. 1985;40(2):111–8. 38. Wulff HR. Rational diagnosis and treatment. Oxford, England: Blackwell; 1981. 39. Olivier P, Caron J, Haramburu F, Imbs J-L, Jonville-Béra A-P, Lagier G, et al. Validation d’une échelle de mesure : exemple de l’échelle française d’évitabilité des effets indésirables médicamenteux. Thérapie. 2005;60(1):39–45. 40. Schumock GT, Thornton JP. Focusing on the preventability of adverse drug reactions. Hosp Pharm 1992; 27: 538. 41. Patel PB, Patel TK. Mortality among patients due to adverse drug reactions that occur following hospitalisation: a meta-analysis. Eur J Clin Pharmacol. 2019;75(9):1293–1307. doi:10.1007/s00228- 019-02702-4.42. Oscanoa TJ, Lizaraso F, Carvajal A. Hospital admissions due to adverse drug reactions in the elderly. A meta-analysis. Eur J Clin Pharmacol. 2017;73(6):759–770. doi:10.1007/s00228-017-2225-3. This article is protected by copyright. All rights reserved. 43. Mangoni AA, Jackson SH. Age-related changes in pharmacokinetics and pharmacodynamics: basic principles and practical applications. Br J Clin Pharmacol. 2004;57(1):6-14. doi:10.1046/j.1365- 2125.2003.02007.x 44. Routledge PA, O'Mahony MS, Woodhouse KW. Adverse drug reactions in elderly patients. Br J Clin Pharmacol. 2004;57(2):121-126. doi:10.1046/j.1365-2125.2003.01875.x 45. Shi S, Mörike K, Klotz U. The clinical implications of ageing for rational drug therapy. Eur J Clin Pharmacol. 2008;64(2):183-199. doi:10.1007/s00228-007-0422-1 46. Rawlins MD, Thompson JW. Mechanisms of adverse drug reactions. In: Davies DM, ed. Textbook of adverse drug reactions. Oxford: Oxford University Press, 1991:18-45. 47. Mittmann N, Liu BA, Iskedjian M, et al. Drug-related mortality in Canada (1984-1994). Pharmacoepidemiol Drug Saf. 1997;6(3):157–168. doi:10.1002/(SICI)1099- 1557(199705)6:3<157::AID-PDS260>3.0.CO;2-J 48. Leone R, Sottosanti L, Luisa Iorio M, Santuccio C, Conforti A, Sabatini V, et al. Drug-related deaths: an analysis of the Italian spontaneous reporting database. Drug Saf. 2008;31(8):703–713. doi:10.2165/00002018-200831080-0000749. van Leeuwen RW, Brundel DH, Neef C, van Gelder T, Mathijssen RH, Burger DM, et al. Prevalence of potential drug-drug interactions in cancer patients treated with oral anticancer drugs. Br J Cancer. 2013;108(5):1071-8. 50. Jönsson AK, Hakkarainen KM, Spigset O, Druid H, Hiselius A, Hägg S. Preventable drug related mortality in a Swedish population. Pharmacoepidemiol Drug Saf. 2010;19(2):211–215. doi:10.1002/pds.1890 51. Leelakanok N, Holcombe AL, Lund BC, Gu X, Schweizer ML. Association between polypharmacy and death: A systematic review and meta-analysis. J Am Pharm Assoc (2003). 2017;57(6):729– 738.e10. doi:10.1016/j.japh.2017.06.002 52. Angamo MT, Chalmers L, Curtain CM, Bereznicki LR. Adverse-Drug-Reaction-Related Hospitalisations in Developed and Developing Countries: A Review of Prevalence and Contributing Factors. Drug Saf. 2016;39(9):847–857. doi:10.1007/s40264-016-0444-7. 53. Higgins JPT, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, Welch VA (editors). Chapter 13: Assessing risk of bias due to missing results in a synthesis. In: Cochrane Handbook for Systematic Reviews of Interventions version 6.1 (updated September 2020). Cochrane, 2020. Available from www.training.cochrane.org/handbook (last accessed 4 Dec 2020). This article is protected by copyright. All rights reserved. This article is protected by copyright. All rights reserved. Table 1: Quality rating of included studies Criteria / Studies Juntti- Pardo Lapatto- Pardo Grévy Montané Patinen Cabello Reiniluoto Cabello [29] [14] [30] [32] [31] [15] 1. Was the research question or objective in this paper clearly stated? Yes Yes Yes Yes Yes Yes 2. Was the study population clearly specified and defined? Yes Yes Yes Yes Yes Yes 3. Was the participation rate of eligible persons at least 50%? Yes No Yes Yes Yes Yes 4. Were all the subjects selected or recruited from the same or similar populations (including the same time period)? NA NA NA NA NA NA Were inclusion and exclusion criteria for being in the study prespecified and applied uniformly to all participants? 5. Was a sample size justification, power description, or variance and effect estimates provided? NA NA NA NA NA NA 6. For the analyses in this paper, were the exposure(s) of interest measured prior to the outcome(s) being measured? Yes Yes Yes Yes Yes Yes 7. Was the timeframe sufficient so that one could reasonably expect to see an association between exposure and outcome if it existed? NA NA NA NA NA NA 8. For exposures that can vary in amount or level, did the study examine different levels of the exposure as related to the outcome (e.g., Yes No No No Yes Yes categories of exposure, or exposure measured as continuous variable)? 9. Were the exposure measures (independent variables) clearly defined, valid, reliable, and implemented consistently across all study Yes Yes Yes Yes Yes Yes participants? 10. Was the exposure(s) assessed more than once over time? NA NA NA NA NA NA 11. Were the outcome measures (dependent variables) clearly defined, valid, reliable, and implemented consistently across all study Yes Yes Yes Yes Yes Yes participants? 12. Were the outcome assessors blinded to the exposure status of participants? NA NA NA NA NA NA 13. Was loss to follow-up after baseline 20% or less? NA NA NA NA NA NA 14. Were key potential confounding variables measured and adjusted statistically for their impact on the relationship between exposure(s) and outcome(s)? No Yes No Yes No Yes Quality score (n, %) 7/8 6/8 6/8 7/8 7/8 8/8 87.5% 75% 75% 87.5% 87.5% 100% NA: not applicable This article is protected by copyright. All rights reserved. Table 2: Characteristics of included studies DRD DRD Main DRD Study Drug-drug Main adverse drug patients hospitalization DRD potentially ADR Causality suspected patients Study design interactions reactions age stay preventable Risk factors Author definition assessment drug class gender (%) (mean, Mean, range % (tool) (ATC) (% women) range) (days) - Haemorrhages (35%, 26/75) Juntti-Patinen - Infections in 65 45.3 NR [30] Retrospective WHO - WHO - UMC L and B NR immunosuppressed patients 8.2 (1-81) NR (3-93) (34/75) (29%, 22/75) - WHO - UMC - Number of drugs Pardo Cabello 65% - Haemorrhages (71%, 12/17) 73 29.4 NR - Charlson comorbidity Retrospective WHO - Naranjo algorithm M and B 7.9 (2-39) index [32] (11/17) (39-91) (5/17) modified - Length stay - Haemorrhages (33%, 17/52) Lapatto- - Infections in 74 36.5 NR Reiniluoto Retrospective WHO - WHO - UMC L and B NR immunosuppressed patients NR NR [31] (25-99) (19/52) (25%, 13/52) - Haemorrhages (19%, Pardo Cabello 30% 48/256) 77 50.4 - Number of drugs Retrospective WHO - Wulff B and N NR NR [15] (77*/256) - Cardiac arrhythmia (14%, (69-86) (129/256) - Comorbidity 35/256) - Haemorrhages (46%, 85/184) - French method 30.4% 76 43.5 34% (63/184) Grévy [29] Retrospective WHO L and B - Infections in 15.5 (1-158) NR - Naranjo algorithm (56/184)) immunosuppressed patients (31-100) (80/184) (Olivier) (16%, 30/184) - Age - Haemorrhages (47%, 34/73) 47% (34/73) - Gender European - WHO - UMC 44% - Infections in 70 27.4 Montané [14] Retrospective L and B immunosuppressed patients 8.8 (0-57) (Schumok- - Number of drugs Directive - Naranjo algorithm (32/73) (19-94) (20/73) - Charlson comorbidity (44%, 32/73) Thornton) index ATC: Anatomical Therapeutic Chemical classification where B is Blood and blood forming organs, L is Antineoplastic and immunomodulating agents, M is Musculoskeletal System, and N is Nervous system; DRD: drug related death; NR: not reported; WHO: World Health Organization; WHO - UMC: World Health Organization - Uppsala Monitoring Centre; *: calculated This article is protected by copyright. All rights reserved. Table 3: Study characteristics and incidences of drug-related death (DRD) and hospital-acquired DRD Hospital-acquired Hospital- Number of Number of Number of DRD occurrence Hospital-acquired Number of DRD incidence of DRD occurrence acquired DRD Study author Study period deceased deceased Number of hospital- rate of deceased DRD incidence of Country admitted hospitalized patients rate of deceased Incidence of year inpatients inpatients DRD (D) acquired DRD inpatients inpatients patients (A) (D/A) inpatients DRD (B) assessed (C) (E) (D/C) (E/A) (E/C) (E/D) 0.053% 4.96% Not Juntti-Patinen [30] Finland 2000 141,484 1,547 1,511 75 NR Not calculable Not calculable (75/141,484) (75/1,511) calculable 0.08% 5.88% 0.08% 5.88% 100% Pardo Cabello [32] Spain 2004 21,541 820 289 17 17 (17/21,541) (17/289) (17/21,541) (17/289) (17/17) Lapatto- near 0.013% 3.04% 0.0025% 0.58% 19.2% Finland 2012 1,708 1,708 52 10 Reiniluoto [31] 400,000 (52/400,000) (52/1,708) (10/400,000) (10/1,708) (10/52) 2009 0.74% 18.44% 0.5% 11.6% 63% Pardo Cabello [15] Spain 34,590 1,400 1,388 256 161** (256/34,590) (256/1,388) (161/34,590) (161/1,388) (161/256) 01-10/2010* 0.32% 11.18% 0.12% 4.19% 37.5% Grévy [29] France 2014 57,450 1,646 1,646 184 69 (184/57,450) (184/1,646) (69/57,450) (69/1,646) (69/184) 0.34% 7.05% 0.028% 0.58% 8.2% Montané [14] Spain 2015 21,483 1,135 1,036 73 6 (73/21,483) (73/1,036) (6/21,483) (6/1,036) (6/73) Meta-analysis 0.1% 7.3% 0.1% 2.7% 37.4% Europe 2000-2015 676,548 8,256 7,578 657 263 % (CI 95%) (0.0 – 0.4) (4.1-12.5) (0.0 – 0.2) (1.0 – 6.9) (17.7-62.4) *: 22 months; **: calculated DRD: drug-related death; CI: confidence interval This article is protected by copyright. All rights reserved.