Potentially inappropriate drug prescribing and polypharmacy in older patients can lead to drug–drug interactions and increased risk of drug-related adverse events.Reference Rochon, Petrovic, Cherubini, Onder, O'Mahony and Sternberg1 The use of multiple, concurrent medications that act in the central nervous system (CNS) can have larger impacts on older patients because of physiologic changes related to age (e.g. enhanced CNS penetration, decreased metabolisation).Reference Drenth-van Maanen, Wilting and Jansen2 For instance, prescribers should avoid the use of concurrent CNS-active medications in older patients because of their association with a higher risk of falls and accelerated cognitive decline.Reference Wright, Roumani, Boudreau, Newman, Ruby and Studenski3,Reference Johnell, Jonasdottir Bergman, Fastbom, Danielsson, Borg and Salmi4 Further, observational studies have reported that patients with CNS-active polypharmacy use present with a higher risk of cardiovascular events, unintentional overdoses, admissions to hospital and all-cause mortality compared with those without.Reference Johnell, Jonasdottir Bergman, Fastbom, Danielsson, Borg and Salmi4–Reference Bykov, Bateman, Franklin, Vine and Patorno8 While the underpinnings of the associations described for CNS-active polypharmacy and these adverse events remain to be completely elucidated, current evidence suggests that such medications are detrimental in older adults and their use should be minimised whenever possible.
Current evidence
Despite formal recommendations against the use of CNS-active polypharmacy in older individuals (e.g. Beers criteria and a black box warning for concurrent use of opioids with benzodiazepines, Z-drugs and other CNS depressants),9,10 recent evidence suggests that CNS-active polypharmacy remains frequent.Reference Maust, Strominger, Kim, Langa, Bynum and Chang11–Reference Slater, White and Frisher13 Notably, available data on the burden of CNS-active polypharmacy are limited to populations from the USA and Europe.Reference Maust, Strominger, Kim, Langa, Bynum and Chang11–Reference Slater, White and Frisher13 Medication use may diverge in Latin American countries such as Argentina, but studies measuring CNS-active polypharmacy are lacking. The healthcare system of Argentina differs from those of the USA and Europe, as it is characterised by the intersection of three sectors: (a) the state-funded public healthcare sector, which provides coverage mainly to uninsured individuals (40–45% of the population); (b) the social system healthcare sector, which provides labour union-run insurance services to formal workers (50–55%); and (c) the private sector, which includes both for-profit and not-for-profit organisations providing care to their clients (5–8%).Reference Marin14 Generally, patients in the private healthcare sector have a higher income, while fees for medication purchases are also higher than in the public and social sectors.Reference Marin14 Hence, patterns of CNS-active medication use in this setting may differ compared with the reports of previous studies.Reference Maust, Strominger, Kim, Langa, Bynum and Chang11–Reference Slater, White and Frisher13 Understanding the burden of CNS-polypharmacy and its drivers in diverse populations thus remains crucial to identify effective strategies to help providers avoid CNS-polypharmacy altogether and deploy safe deprescribing interventions.
Objectives
We conducted a retrospective cohort study to estimate the prevalence of out-of-hospital CNS-active polypharmacy among individuals aged 60 years or older in a health maintenance organisation in Argentina. Our objectives were to (a) describe CNS-active polypharmacy period prevalence during 2021 and (b) investigate sociodemographic and clinical factors associated with CNS-active polypharmacy.
Method
Data sources
The present cohort study was conducted using out-of-hospital, individual-patient-level data from the Hospital Italiano de Buenos Aires’ health maintenance organisation collected between 1 January 2017 and 31 December 2021. The hospital is one of the largest university teaching hospitals in Argentina and Latin America and it is accredited by the Joint Commission International.15 The institution manages its own private, not-for-profit, healthcare network, which provides care to over 170 000 affiliates in Argentina, with an extensive system of pharmacies, out-patient services and low- and high-complexity facilities (e.g. two university teaching hospitals, numerous medical offices).16 The Hospital Italiano de Buenos Aires’ health maintenance organisation uses integrated electronic health records that gather information on out-of-hospital and in-hospital diagnosis and procedures using SNOMED CT, Spanish 2020 edition for Windows (SNOMED international, London, UK; see https://www.snomed.org).Reference Chang and Mostafa17 Healthcare professionals input clinical terminologies into electronic health records, and SNOMED CT encodes these terminologies using standardised terms that are shared across medical specialties and healthcare systems.Reference Chang and Mostafa17 Finally, the health maintenance organisation has a comprehensive registry of medication purchases of patients. Filled prescriptions are encoded using the Anatomical Therapeutic Chemical (ATC) index codes and captures patients’ identifier, milligrams per pill, number of pills and number of boxes purchased.18 However, the database does not capture the days of supply or the clinical indication for the filled prescription.
Study population and study period
Eligible individuals were aged 60 years or older on 1 January 2021 (index date) and affiliated to the Hospital Italiano de Buenos Aires’ health maintenance organisation. Patients were followed from the index date until disaffiliation from the health maintenance organisation, death or end of the study period (31 December 2021).
CNS-active polypharmacy definition
CNS-active medications were defined as any out-of-hospital filled prescription for an agent from the following medication classes: anti-epileptics, benzodiazepines, Z-drugs, antipsychotics, opioids and antidepressants (i.e. selective serotonin receptor inhibitors, serotonin–noradrenaline reuptake inhibitors and tricyclic antidepressants)19 (Supplementary Table 1 available at https://doi.org/10.1192/bjo.2024.798). A CNS-active polypharmacy event was defined as the concurrent exposure to three or more CNS-active medications. We considered patients exposed to a CNS-active medication through the month of prescription fill and an additional period of 1 month after the prescription was not refilled to account for real-world practices in medication use (i.e. gaps in medication refilling such as stockpiling and variations in medication purchase practices).Reference Rasmussen, Wettermark, Steinke and Pottegård20,Reference Bharat, Degenhardt, Pearson, Buizen, Wilson and Dobbins21
There were no restrictions to the number of medications classes or combinations of medication classes involved in CNS-active polypharmacy. For example, one antipsychotic, one Z-drug and one benzodiazepine; two distinct antidepressants and one antipsychotic; or three distinct benzodiazepines could all be counted as a CNS-active polypharmacy event.Reference Maust, Strominger, Kim, Langa, Bynum and Chang11,19
Additional covariates
We captured baseline information (i.e. at index date) on age, gender, comorbidities (i.e. hypertension, cardiac failure, coronary heart disease, peripheral vascular disease, diabetes mellitus, chronic kidney disease, cirrhosis, cancer, chronic obstructive pulmonary disease, asthma, tobacco use, alcohol misuse, cognitive complaint, dementia and its subtypes, Parkinson's disease, anxiety and depressive disorders, bipolar disease, schizophrenia, sleep disorders and epilepsy) and the number of admissions to hospital in the preceding year at the individual patient level.
To estimate the positive predictive value of the clinical diagnoses, we reviewed the electronic medical charts of a probabilistic sample of each diagnosis (Supplementary Table 2). Five investigators (A.C., W.R., S.F., P.P., G.M.) reviewed the charts and recorded whether the diagnosis captured by the SNOMED CT term matched clinical criteria for the disease. Thus, we estimated the positive predictive value of each condition by calculating the number of patients classified by the medical doctors as true cases over the total number of patients classified by SNOMED CT terms. Any discrepancies were resolved by the lead author (A.F.).
Statistical analysis
To describe the CNS-active polypharmacy period prevalence during 2021, we calculated the sum of individuals with CNS-active polypharmacy: (a) on 1 January 2021 (i.e. prevalent users of CNS-active polypharmacy) and (b) during the study period (from 1 January 2021 to 31 December 2021) (i.e. incident users of CNS-active polypharmacy), divided by the total number of individuals in the cohort on 1 January 2021.Reference Rasmussen, Wettermark, Steinke and Pottegård20 We constructed 95% confidence intervals using the methods described by Clopper and Pearson (exact confidence intervals).Reference Clopper and Pearson22
To identify potential factors associated with CNS-active polypharmacy at the individual patient level, we used a multivariable logistic regression model to estimate odds ratios and 95% confidence intervals. Variables included in the exploratory model were demographics (i.e. age, gender), mental health diagnoses (i.e. anxiety and depressive disorders, sleep disorders, schizophrenia, bipolar disorder), neurodegenerative disorders (i.e. cognitive complaint, Parkinson's disease, Alzheimer's disease, vascular, frontotemporal, Lewy body or any type of dementia), other comorbidities (i.e. number of prior admissions to hospital in the preceding year as a continuous variable, epilepsy, hypertension, cardiac failure, coronary disease, peripheral vascular disease, diabetes mellitus, chronic kidney disease, cirrhosis, cancer, chronic obstructive pulmonary disease, asthma) and substance use (i.e. past or current alcohol misuse and tobacco use). Thus, for example, the estimated odds ratio of the exploratory multivariable model compared the odds of CNS-active polypharmacy for individuals with depressive disorders at baseline with the odds for those without, while keeping constant the remaining covariables included in the model. Furthermore, we conducted bivariate analyses for the individual association of variables included in the exploratory model and CNS-active polypharmacy.
To identify groups and combinations of drugs driving CNS-active polypharmacy, we described the individual medications, medication classes and combinations of medication classes more frequently involved in CNS-active polypharmacy. We allowed for multiple events of polypharmacy per patient, such that we recorded the individual medications and medication classes involved in each CNS-active polypharmacy event (Supplementary Table 1).
Sensitivity and subgroup analyses
To test the robustness of our period-prevalence analysis, we modified our outcome definition of CNS-active polypharmacy as having filled three or more concurrent CNS-active medications: (a) on the same calendar week (without using an additional gap period of 1 month after a prescription was not refilled) and (b) extending the gap period to 2 months after a prescription was not refilled.Reference Bharat, Degenhardt, Pearson, Buizen, Wilson and Dobbins21 Further, since CNS-active medications may be especially deleterious in patients with increasing age and in those with dementia, we conducted subgroup analyses to estimate the period prevalence of CNS-active polypharmacy by diagnosis (patients with and without dementia) and age group (older than 80 years; 70–79; 60–69).Reference Rochon, Petrovic, Cherubini, Onder, O'Mahony and Sternberg1,Reference Drenth-van Maanen, Wilting and Jansen2
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. All procedures involving human participants/patients were approved by institutional review board (Comité de Ética para Protocolos de Investigación del Hospital Italiano de Buenos Aires, protocol number: 7235), and patient consent was waived. The present report followed the REporting of studies Conducted using Observational Routinely-collected Data for PharmacoEpidemiology (RECORD-PE) statements for pharmacoepidemiologic studies.Reference Langan, Schmidt, Wing, Ehrenstein, Nicholls and Filion23 All analyses were performed using R statistical software for Windows, version 4.3.1 (R Foundation for Statistical Computing, Vienna, Austria; see https://www.r-project.org).
Results
Of the initial 77 171 patients available in the cohort, 63 857 were included in the present study (Fig. 1). Overall, 443 (0.7%) patients disaffiliated from the health maintenance organisation and 2332 (3.7%) patients died during follow-up. Baseline characteristics of patients included are summarised in Table 1. The cohort mainly comprised individuals aged 70 years or older and most patients were female. Anxiety (21%), depressive (14%) and sleep (11%) disorders were the most frequent mental health diagnoses in the entire cohort at baseline. Hypertension was the most frequent comorbid condition and roughly a fifth of the patients presented either current or previous tobacco use (Table 1).
Flowchart of study patients.
CNS, central nervous system.
Characteristics of adults aged 60 years and older on 1 January 2021 (Hospital Italiano de Buenos Aires, 2021)
COPD, chronic obstructive pulmonary disease.
Frequencies lower than 20 are shown as inequalities.
A total of 4535 individuals had CNS-active polypharmacy during 2021, for a period prevalence of 7.1% (95% CI: 6.9–7.3%). Of these, 1211 were prevalent users of CNS-active polypharmacy (Supplementary Figure 1). Prevalent users of CNS-active polypharmacy at baseline were on average older, more likely to be female and presented a higher prevalence of mental health, neurodegenerative and cardiovascular and respiratory diseases than those without use of CNS-active polypharmacy (Supplementary Table 3).
Sensitivity and subgroup analyses
The period prevalence of CNS-active polypharmacy was consistent when we changed our outcome definition. The period prevalence of CNS-active polypharmacy was 5.0% (95% CI: 4.9–5.2%) when we defined polypharmacy as filling prescriptions for three or more CNS-active medications on the same week without a gap period (Table 2). Conversely, when we extended the gap period to 2 months of exposure after a prescription was not refilled, the period prevalence of CNS-active polypharmacy was 7.6% (95% CI: 7.4–7.8%) (Table 2). Finally, among the pre-specified groups of interest, the period prevalence of CNS-active polypharmacy was more than four times higher in individuals with dementia compared with those without. The period prevalence of CNS-active polypharmacy was more than two times higher in individuals aged 80 years or older compared with the prevalence observed in patients aged 60–69 years (Table 2). Conversely, the prevalence of CNS-active polypharmacy in individuals aged 70–79 years was similar to that observed in patients aged 60–69 years (Table 2).
Prevalence of central nervous system (CNS)-active medications according to different measurement strategies and subgroups of patients
a. We estimated 95% confidence intervals using Clopper–Pearson (exact) approach for a binomial.
b. Using a gap period of 1 month after a prescription was filled (main analysis) to estimate CNS-active polypharmacy.
c. Counting only prescriptions occurring within the same week, without a gap period to estimate CNS-active polypharmacy.
Factors associated with CNS-active polypharmacy
Several baseline characteristics were associated with CNS-active polypharmacy (Table 3). We found the largest association between CNS-active polypharmacy with frontotemporal dementia, but our estimates were imprecise. Similarly, severe mental illnesses (i.e. bipolar disorder, major depression and schizophrenia) and a diagnosis of cognitive complaint were positively associated, while cirrhosis was inversely associated with the prevalence of CNS-active polypharmacy (Table 3).
Baseline factors associated with central nervous system (CNS)-active medication polypharmacy
COPD, chronic obstructive pulmonary disease.
a. Bivariate analyses for the association of the listed baseline variables (individually) and central nervous system-active polypharmacy.
b. Multivariable logistic regression model with 95% confidence intervals estimated using Wald's formula.
c. The reported odds ratio used indicator variable coding to compare age categories 70–79 and 80 years and older with the reference group of 60–69 years. Odds ratios were estimated including all age categories using indicator variable coding.
d. Hospital admission was included in the model as a discrete variable, and the reported odds ratio compared the odds of central nervous system-active polypharmacy of two groups of individuals that differ by one hospital admission in the preceding year, with individuals with a higher number of admissions to hospital presenting a higher odds of central nervous system-active polypharmacy.
Statistically non-significant associations at the 0.05 level for the multivariable analysis are showed in bold format.
Prevalence of CNS-active medication, medication classes and individual medications
A total of 24 596 CNS-active polypharmacy events that involved 80 775 CNS-active medications were recorded among 4535 patients with CNS-active polypharmacy during follow-up. The median time from index date until the first CNS-active polypharmacy event was 3 (interquartile range, 0–7) months (Supplementary Figure 1). Antidepressant, benzodiazepine and antipsychotic classes accounted for 72% of the total prescriptions involved in CNS-active polypharmacy events. Overall, antidepressants were present in 9 of the 10 most frequent medication classes combinations involved in CNS-active polypharmacy events (Table 4). The most frequent combinations of medication classes were as follows: an antidepressant, an antipsychotic and a benzodiazepine in 21% of CNS-active polypharmacy events; an antidepressant, an anti-epileptic and a benzodiazepine in 9% of CNS-active polypharmacy events; and an antidepressant, a benzodiazepine and a Z-drug in 8% of CNS-active polypharmacy events (Table 4). Finally, medications most frequently involved in CNS-active polypharmacy events were clonazepam (12.9% of prescriptions), quetiapine (11.1% of prescriptions), alprazolam (6.5% of prescriptions), pregabalin (6.3% of prescriptions) and escitalopram (6.2% of prescriptions) (Supplementary Table 4).
Frequency of medication classes and combinations of medication classes involved in 24 596 central nervous system-active polypharmacy events
a. Frequencies and percentage estimates are based on a total of 80 775 individual medications used in 24 596 central nervous system-active polypharmacy events among 4535 patients with central nervous system-active polypharmacy. Percentages may not add age exactly 100% owing to rounding.
Discussion
The present study evaluated the prevalence and factors associated with CNS-active polypharmacy in a cohort of adults aged 60 years or older in a health maintenance organisation in Argentina. We found that one in 14 adults aged 60 years or older had CNS-active polypharmacy during 2021. The year 2021 avoided the secular influence of the COVID-19 pandemic onset in our estimations, which presented specific challenges to accessing healthcare among older adults in Argentina.Reference Loza, Baez, Valverdi, Pisula, Salas Apaza and Discacciati25 In addition, we provided the most recent estimate of CNS-active polypharmacy available in the cohort. Our estimation of CNS-active polypharmacy burden remained consistent in multiple sensitivity analyses. We also found a higher prevalence of CNS-active polypharmacy in groups of interest, such as patients aged 80 or older and patients with a diagnosis of dementia. This study provides insight into the concurrent use of CNS-active medications in real-world settings and represents a first step towards understanding the drivers of CNS-active polypharmacy in Latin American populations. Further evidence is required to identify groups of patients that may benefit from the implementation of potential deprescribing interventions adapted to the particular characteristics of the region.Reference Rochon, Petrovic, Cherubini, Onder, O'Mahony and Sternberg1,Reference Schapira, Calabró, Montero-Odasso, Osman, Guajardo and Martínez26
Previous studies evaluated CNS-active medication use across several populations. In the USA, a 2017 study found a prevalence of CNS-active polypharmacy (defined as the use of three or more CNS-active medications simultaneously) of 1.4% among out-patients aged 65 or older.Reference Maust, Gerlach, Gibson, Kales, Blow and Olfson12 The study by Maust et alReference Maust, Gerlach, Gibson, Kales, Blow and Olfson12 reported a noticeably lower prevalence of polypharmacy as compared with our study. Reasons for this discrepancy include differences in (a) analytical approach (i.e. week versus annual period prevalence), (b) medication use ascertainment (i.e. physician's reports versus pharmacy claims), (c) comorbidity burden (i.e. a higher prevalence of anxiety, sleep and depressive disorders in this study population) and (d) healthcare system characteristics (i.e. possibly better access to medications in our study population). In addition, a 2021 study found a prevalence of CNS-active polypharmacy (defined as three or more CNS-active medications for 30 consecutive days or longer) of nearly 14% among patients with a diagnosis of dementia.Reference Maust, Strominger, Kim, Langa, Bynum and Chang11 In Denmark, a 2016 study reported a prevalence (defined as the concurrent use of two or more CNS-active medication classes) of roughly 25% among patients with dementia.Reference Nørgaard, Jensen-Dahm, Gasse, Hansen and Waldemar27 These findings underscore how, despite variations in the definition of polypharmacy,Reference Masnoon, Shakib, Kalisch-Ellett and Caughey28 CNS-active medication use remained consistently high across different settings. Our study expands the literature by providing estimates from a health maintenance organisation in Argentina.
We also explored clinical factors related to CNS-active polypharmacy. The largest associations were found for mental health disorders that frequently present with psychotic features, that is, frontotemporal dementia, bipolar disorder and schizophrenia. However, our estimates were imprecise because of the low number of cases observed and should be interpreted with caution. More prevalent conditions, such as anxiety and depressive disorders, were also associated with CNS-active polypharmacy. Our findings suggest that older adults with mental health conditions may represent an attractive target to direct efforts to reduce the prescribing of multiple concomitant CNS-active agents. These patients are also at a heightened risk of poor clinical outcomes, including opioid-related overdoses,Reference Szmulewicz, Bateman, Levin and Huybrechts6,Reference Szmulewicz, Bateman, Levin and Huybrechts7 accelerated cognitive decline, higher risk of falls and higher risk of all-cause mortality.Reference Wright, Roumani, Boudreau, Newman, Ruby and Studenski3,Reference Huang, Steptoe, Wei and Zaninotto5,Reference Weiner, Hanlon and Studenski29,Reference Gray, Marcum, Dublin, Walker, Golchin and Rosenberg30 Previous interventions within this same health maintenance organisation have used alarms set in the electronic health records to identify patients at high risk of poor clinical outcomes to improve their clinical trajectories within the health system network.Reference Basquiera, Aguirre, Serra, Vaca, Brulc and Perusini31 A similar approach using risk factors of CNS-active polypharmacy could be implemented, for example, by creating a warning when initiation of treatments would lead to CNS-active polypharmacy in older adults with dementia. Thus, tailored deprescribing interventionsReference Schapira, Calabró, Montero-Odasso, Osman, Guajardo and Martínez26 for these patients and their caregiversReference Chock, Wee, Gan, Teoh, Ng and Lee32 could potentially have a large clinical impact. In addition, minimising the duration of CNS-active polypharmacy when its use is required should also be considered. Monitoring the appropriate duration of CNS-active treatments could be an initial measure to reduce CNS-active medication burden, especially when treatments are intended for short-term use (e.g. benzodiazepines).Reference O'Mahony, Cherubini, Guiteras, Denkinger, Beuscart and Onder33 Until the results of ongoing high-quality randomised clinical trials evaluating deprescribing interventions for CNS-active medications on broader populations are published,Reference Balderson, Gray, Fujii, Nakata, Williamson and Cook34 resource allocation to patients at the highest risk of adverse clinical outcomes (e.g. patients in nursing home residencies) could also be considered.Reference Lee, Negm, Peters, Wong and Holbrook24,Reference Gedde, Husebo, Mannseth, Kjome, Naik and Berge35
This study presents several limitations. First, we used medication purchases documented in out-of-hospital settings to measure the occurrence of CNS-active polypharmacy. While misclassification of CNS-active medication use may still be possible, we expected to capture all out-of-hospital medication purchase events, since the medications included in the present study are unavailable over the counter, require reports to the national regulatory authorities upon purchase and clients have re-imbursement benefits when using the hospital's pharmacy network.16 Similarly, we could not identify musculoskeletal relaxant use, a medication class recently added to the CNS-active polypharmacy definition.9 For both reasons, our calculations may be a conservative estimate of the true CNS-active polypharmacy prevalence under the new definitions in this setting. Second, we calculated the period prevalence, a measure of disease occurrence that includes both prevalent and incident CNS-active medication users. The inclusion of prevalent CNS-active polypharmacy users casts doubt on causality claims with respect to the association between predictors and polypharmacy.Reference Rasmussen, Wettermark, Steinke and Pottegård20 However, our analyses were exploratory and focused on identifying groups of individuals with high prevalence of CNS-active medication use that may benefit from deprescribing interventions. Furthermore, these groups of patients can be subject of future research to evaluate the impact of such extensive CNS-active polypharmacy exposure on health outcomes. Third, we did not have information on days of supply and we could not evaluate the duration of CNS-active polypharmacy, which may be related to the occurrence of adverse events. Finally, our study was conducted in a health maintenance organisation, and our findings may not be representative of the national or regional trends in CNS-active polypharmacy occurrence. Future studies reproducing our methods in populations with dissimilar health-seeking behaviours, burden of comorbidities and distribution of social determinants of health remain warranted.
In conclusion, one in 14 individuals aged 60 or older presented CNS-active polypharmacy during follow-up in a large health maintenance organisation in Argentina in 2021. CNS-active polypharmacy use was found to be higher in patients aged 80 years or older and in patients with a diagnosis of dementia. Our findings underscore the need for reducing concurrent CNS-active medication use in older patients whenever feasible.
Supplementary material
Supplementary material is available online at https://doi.org/10.1192/bjo.2024.798
Data availability
The data-set from this study is held securely in coded form. The data are not publicly available because of the presence of sensitive information that could compromise the privacy of research participants. Data sharing agreements from the Hospital Italiano de Buenos Aires prohibit making the data-set publicly available. Furthermore, the creation of this data-set involved the use of several data sources from administrative records that were linked at the individual patient level. While the final data-set cannot be shared, the full data-set creation plan and underlying analytic code are available from the main author (A.F.) upon request.
Acknowledgements
The authors would like to thank Nicolas Tentoni MD and the Internal Medicine Research Unit team from the Hospital Italiano de Buenos Aires for their support in the development of the present study. The analyses, conclusions, opinions and statements expressed herein are solely those of the authors of the study.
Author contributions
All authors meet International Committee of Medical Journal Editors (ICMJE) recommendations for authorship and have made substantial contributions to the concept or design of the work. A.F., A.G.S., F.A.: study conception, analysis of the data, critical revisions. T.B., J.P.: study conception, interpretation of data, critical revisions. P.P., S.F., W.R., A.C., G.M.: study conception, data acquisition, critical revisions. A.F. and A.G.S. drafted the first version of this manuscript. All authors gave approval of the final version to be published.
Funding
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
Declaration of interest
None.
Transparency declaration
A.F. guarantees that this manuscript is an honest, accurate and transparent account of the study being reported; no important aspects of the study have been omitted; and any discrepancies from the study as planned have been explained.
Open access
eLetters
No eLetters have been published for this article.