Introduction

As the complexity of pharmacotherapy has increased with increasing medication use, particularly among older adults with multiple morbidities [1], medication risk management has become an increasingly important area of research. In this field, potentially inappropriate medication (PIM) is a term used to describe the use of a medicine for which the associated risks outweigh the potential benefits, especially when more effective alternatives are available [2]. PIM use is an important public health challenge, with high prevalence rates (from 18 to > 40%) across a variety of healthcare settings [3,4,5,6]. Notably, elderly patients are more likely to be exposed to PIMs because they often deal with age-related pharmacokinetic and pharmacodynamic changes, which can result in increased adverse drug reactions and decreased efficacy [7, 8]. Additionally, older patients often suffer from multiple chronic-degenerative diseases and therefore use a higher number of drugs, compared to other age groups [9]. In this population, PIM use can lead to avoidable adverse drug events (ADEs) [3, 10], including falls, fractures, and delirium and is associated with hospitalization [11,12,13] and mortality [14,15,16].

In recent years, many strategies and tools have been developed to assess the appropriateness of medication use in older people [2, 17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51]. Explicit criteria (EC) focusing on a single medication/medication class can support improvements to the quality of drug therapy and help to monitor drug therapy. However, the development of evidence-based PIM lists specifically for older populations is problematic, as older people are typically underrepresented or excluded from most efficacy and safety trials [52, 53]. Accordingly, some investigators have used a consensus technique that synthesizes accumulated expert opinion to develop EC that would facilitate the formulation of recommendations for suitable treatments in older people [54]. This consensus technique could be used to determine which statements from the literature are applicable in clinical practice [22].

Many different expert panels, including pharmacists, geriatricians, and other health professionals, have developed lists of EC to identify PIM use among older people in different countries [2, 17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51]. Previous systematic reviews identified 7–25 different PIM lists applied to older people [55,56,57]. Although Luchetti et al. [57] summarized and described all medications classified as PIMs using 14 validated and nonvalidated PIM lists published between 2006 and 2015, the literature lacks a comprehensive evaluation of the most common drug–disease and drug–drug interactions described in these validated PIM lists. Notably, summaries of the items proven valid by many consensus panels may facilitate a translational comparison of the processes and provide information about the most important PIMs in clinical practice, which would inform the development of interventions aimed at improving the prescription of specific medications. Therefore, in this review, we aim to summarize and compare the validated potentially inappropriate medications lists for older people published in different countries between 1991 and 2017. Additionally, we aim to summarize the medications and drug–disease and drug–drug interactions listed in the different potentially inappropriate medications lists.

Methods

This review was performed according to a standard protocol for systematic reviews, which was based on the methodological manuals of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). The PRISMA checklist is available in Appendix 1.

Search strategy

We systematically identified studies published between January 1991 and April 2017 without any language restriction in the PUBMED and Academic Search Complete via EBSCOhost, Academic Search Premier via EBSCOhost, AgeLine via EBSCOhost, and CINAHL via EBSCOhost electronic databases. We selected this time limit for publication inclusion because the first instrument for PIM assessment was published in 1991 by Beers et al. [22]. The search included terms related to older adults or appropriate/inappropriate medication. Details of the full search strategy are included in Appendix 2.

Eligibility criteria

Original studies describing the EC used to determine potentially inappropriate medications were considered eligible for inclusion in this review if they involved individuals aged 65 years and older and described the development and validation of the methods used in the PIM list. Interventions and observational studies that evaluated PIMs were also retained if the abstract described potentially relevant PIM lists.

We applied the following exclusion criteria: medication review techniques using implicit criteria to evaluate PIMs and lists of PIMs restricted to specific therapeutic classes or specific diseases. Additionally, we excluded studies of PIMs not validated by expert consensus and guidelines or recommendations for the assessment of inappropriate prescriptions, as well as letters, editorials, and duplicate studies.

Study selection

Duplicate manuscripts were removed after exporting the search results to Endnote, version X6 (Clarivate Analytics, Philadelphia, PA, USA). Subsequently, two reviewers independently screened the titles and abstracts of the remaining manuscripts to identify potentially relevant studies describing the development and validation of PIM lists. Additional studies were identified by a manual search of the citation lists for studies that detailed potentially relevant PIM lists. Finally, full-text copies of studies that described either the validation or use of any of the potentially relevant measures were retrieved and considered for inclusion in this review. If a decision could not be reached regarding the ability of a manuscript to meet the inclusion criteria, a decision was reached during the following selection round.

Data extraction and synthesis

Two authors (FRM and JSF) independently extracted the data, after which the first author checked the completeness by reviewing the extraction tables generated by the second author and checking the extracted data in the full-text articles. Disagreements were resolved by discussion between the two authors; if no agreement could be reached, a third author was consulted (VMV).

The following data were extracted from the selected articles: country of origin, source of data used, and validation method (consensus technique, expert panel, literature based). We also extracted aspects evaluated in the lists of PIMs (medications, dosage, duration of therapy, duplication, drug–disease interactions, drug–drug interactions). We also analyzed the medication/medication class names and drug–disease interactions (medication or medication dosage or medication duration with consideration of diagnosis) and drug–drug interactions reported in all PIM lists. We considered all medications belonging to a class as inappropriate if the authors described concerns about the medication class and did not describe single medications. However, the anticholinergic medication class exhibited considerable variation in terms of the selection of specific drugs. Therefore, we included anticholinergic drugs described in a recent review of the literature for the EC that did not specifically state which medicines were considered anticholinergic [58]. Additionally, we also considered a medication class as inappropriate when the authors described single medications and raised concerns related to the medication class.

The data were entered into Excel (Microsoft Corp., Redmond, WA, USA), and all individual medications reported in the studies were subsequently grouped into Anatomical, Therapeutic and Chemical (ATC) classes (five levels).

Results

The search strategy produced 2933 potentially relevant publications (Fig. 1). After screening titles and abstracts, we retained 248 potentially relevant publications according to the inclusion criteria. After a full-text review, 214 articles were excluded according to the exclusion criteria. A manual search from the reference lists of the included articles produced two relevant publications not found in the previous systematic database search. Thus, 36 articles were included in this systematic review [2, 17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51].

Fig. 1
figure 1

PRISMA flow diagram of search strategy results

Full size image

Table 1 describes the characteristics of the PIM lists evaluated in this review. Most studies were conducted in Europe [2, 23, 25, 26, 28,29,30, 35, 36, 38, 39, 42, 43, 47, 49] and North America [17,18,19, 21, 22, 27, 32, 37, 41, 45, 46, 48, 51]. However, other countries from Asia [24, 31, 33, 34, 40, 50], Oceania [20], and South America [44] such as Taiwan [24], Pakistan [40], South Korea [33, 34], Thailand [50], Japan [31] Australia [20], and Chile [44] have also published lists of PIMs.

Table 1 Characteristics of potentially inappropriate medication (PIM) lists
Full size table

The majority of the PIM lists (23 PIM list, 63.9%) are aimed at the general population aged 65 years and older. The Norwegian General Practice (NORGEP) criteria [47] and its adaptation for nursing home residents were designed especially for individuals aged 70 years and older [42] and the French criteria [36] for those aged 75 years and older. Only three PIM lists (8.3%) were developed for nursing home residents [22, 32, 42], two (5.5%) for older hospitalized patients [28, 40] and one (2.7%) for use in community pharmacies [49].

Some PIM lists, such as Beers (1991, 1997, 2003, 2012 and 2015) [17, 18, 21, 22, 27], STOPP (Screening Tool of Older People’s Prescriptions) version 1 [29] (2008) and 2 (2015) [43], FORTA (Fit fOR The Aged) [35], Australian Prescribing Indicators Tool (2012) [20], Thailand criteria (2008) [50], and Lindblad criteria (2006) [37], used the current literature on efficacy and safety in older adults as an evidence base to develop their own list of PIMs. Other PIM lists, such as the McLeod criteria (1997) [41], Rancourt criteria (2004) [46], French criteria (2007) [36], NORGEP criteria (2009) [47], and PRISCUS (2010) [30], combined ECs previously published with a review of current literature. However, most PIM lists used previously published PIM lists to develop their lists of PIMs [2, 19, 23–26, 28, 31–34, 38–40, 42, 44–45, 48–49, 51]. Twenty-one (58.3%) of the 36 PIM lists were based on the Beers criteria and its updates, ten(27.8%) on the STOPP criteria and its update, and seven (19.4%) on the McLeod criteria. The tool developed by Tommelein et al. [49] was based on items derived from 14 different PIM lists (Table 1).

Of the 36 studies identified, 19 (52.8%) used the Delphi method and 14 (38.9%) used a modified Delphi method, to validate their ECs. Two studies used the RAND/UCLA [20, 49] (Research and Development/University of California, Los Angeles) process and the Italian criteria [38] used the Nominal Group Technique. The number of experts ranged from 4 to 62 and approximately 50.0% of the consensus panels included between 10 and 20 respondents. We observed a predominance of physicians and pharmacists whose practices concentrate on older adults and clinical pharmacology among the experts. Some studies included experts from different specialties such as psychiatrists [24, 30, 33, 38], cardiologists [24, 38], pulmonologists, gastroenterologists [24, 38], and urologists [24, 38] (Table 1).

Classification systems for PIMs varied between the studies. The majority of PIM lists provide an explicit listing of individual drugs. Eleven (30.6%) tools focused on PIMs to avoid in older adults independent of disease/condition; 22 (61.1%) included PIMs to avoid in older adults for specific diseases or conditions, and 20 (55.6%) mentioned relevant drug–drug interactions. Fourteen (38.8%) tools presented alternative therapies and 10 (27.8%) provided information about special considerations of use. Furthermore, 29 PIM lists (80.5%) also described doses or durations of medications, which should not be exceeded. Avoiding unnecessary duplication was mentioned in eight PIM lists (22.2%) (Table 2).

Table 2 Summary of evaluated aspects in the potentially inappropriate medication (PIM) lists
Full size table

The 36 PIM lists identified a total of 907 different medications/ medication classes. Among them, only 4 classes and 44 medications were reported by more than 69% of PIM lists. The most prevalent class of medication identified as inappropriate was benzodiazepines, which were included in 33 (91.7%) of the 36 PIM lists. Other medication classes of PIMs identified included nonsteroidal anti-inflammatory drugs (NSAIDs) (28/36 PIM lists; 77.8%) followed by tricyclic antidepressants and antihistamines (27/36 PIM lists, 75.0%) (Table 3). Regarding the medications, only diazepam, chlordiazepoxide, indomethacin, and amitriptyline were considered inappropriate by 35 (97.2%) of the 36 PIM lists (Table 4).

Table 3 Most common medication classes reported in potentially inappropriate medication (PIM) lists
Full size table
Table 4 Most common medications reported in potentially inappropriate medication (PIM) lists
Full size table

Similar results were observed when we considered the 33 PIM lists (91.7%) that evaluated PIM independent of disease/condition. Benzodiazepines (29/33 PIM lists, 87.9%) and antihistamines (23/33 PIM lists, 69.7%) were the most common medication classes reported followed by tricyclic antidepressants (19/33 PIM lists, 57.6%) (Table 3). The commonest medications included were diazepam, chlordiazepoxide (31/33 PIM lists, 93.9%), amitriptyline, and chlorpheniramine (28/33 PIM lists, 84.8%) (Table 4).

Medications that can be avoided in individual diseases/conditions are specified in 22 PIM lists (61.1%). The most common medication classes implicated were NSAIDs (20/22 PIM lists, 90.9%), tricyclic antidepressants (19/ 22 PIM lists, 86.4%), followed by urologic spasmolytics, and long-acting benzodiazepines (18/22 PIM lists, 81.8%) (Table 3). Oxybutynin, diazepam, and chlordiazepoxide were the most common medications; they were reported as PIMs in specific diseases/conditions in 20 (90.9%) of 22 PIM lists (Table 4).

We identified 536 different drug–disease interactions involving 84 diseases/conditions. Among them, only 38 (7.0%) drug–disease interactions were cited in more than 25% of PIM lists. The most common conditions cited were constipation/chronic constipation (42/536 drug–disease interactions, 7.8%), dementia/cognitive impairment (41/536 drug–disease interactions, 7.6%), insomnia (36/536 drug–disease interactions, 6.7%), lower urinary tract symptoms/benign prostatic hyperplasia (28/536 drug–disease interactions, 5.2%), heart failure (19/536 drug–disease interactions, 3.5%), and history of falls/fractures (19/536 drug–disease interactions,3.5%). Table 5 summarizes the most common drug–disease interactions identified. The use of NSAIDs in patients with renal insufficiency (15/22 PIM lists, 68.1%) and heart failure (13/22 PIM lists, 59.1%)) and the use of metoclopramide in patients with Parkinson’s disease (13/22 PIM lists, 59.1%) were the most commonly reported. Other prevalent drug–disease interactions included anticholinergic drugs in those with dementia/cognitive impairment (12/22 PIM lists, 54.5%), benzodiazepines in those with a history of falls/fractures (11/22 PIM lists, 50.0%), and urologic spasmolytics in those with lower urinary tract symptoms/benign prostatic hyperplasia (10/22 PIM lists, 45.4%).

Table 5 Most common drug–disease interactions described in 22 potentially inappropriate medication (PIM) lists
Full size table

We also identified 159 potential drug–drug interactions described in 20 PIM lists. Among them, only 16 (10.1%) drug–drug interactions were cited in more than 20% of PIM lists. The most common medication classes implicated were the NSAIDs (19/20 PIM lists, 95.0%), tricyclic antidepressants (14/20 PIM lists, 70.0%), followed by angiotensin-converting-enzyme inhibitors (12/20 PIM lists, 55.6%), and selective serotonin reuptake inhibitors (10/20 PIM lists, 50.0%). Regarding single medications, warfarin was the most common medication reported; it was included in 18(90.0%) of 20 PIM lists that evaluated drug–drug interactions. Table 6 describes the most common drug–drug interactions described. The concomitant use of warfarin with NSAIDs (11/20 PIM lists, 55.5%) and aspirin (7/20 PIM lists, 35.0%) was the most common drug–drug interactions reported followed by the concomitant use of NSAIDs and ACE inhibitors (7/20 PIM lists, 35.0%) and the concomitant use of beta blockers and verapamil (7/20 PIM lists, 35.0%).

Table 6 Most common drug–drug interactions described in 20 potentially inappropriate medication (PIM) lists
Full size table

Discussion

This systematic review presents data from 36 PIM lists (published between 1991 and April 2017) that developed and validated EC for identification of PIMs. The aggregation and comparison of studies showed a wide variability of PIMs, and we identified different 907 medications/medication classes reported in all PIM lists. A previous systematic review identified 729 different medications/classes described in 14 different PIM lists published between 2006 and 2015 [57]. The higher number of medications/classes in our review is justified by the fact that we included more years and other baseline data in our search strategies, and we did not exclude PIM lists for institutionalized or hospitalized patients or criteria that reported only drug–disease interactions. Furthermore, aspects such as different settings and prescribing cultures, differences in medication availability/formulary between countries, and ethnopharmacology may have contributed to these results.

Conversely, we observed that less than half of PIM lists developed their own EC based on literature reviews. The development of evidence-based PIM lists is a dynamic and complex process, because older participants are commonly excluded from well-designed clinical trials [52, 53]. Thus, a majority of the studies used prior PIM lists to develop their own lists of PIMs [2, 19, 23–26, 28, 31–34, 38–40, 42, 44–45, 48–49, 51]. However, some of these authors have combined different PIM lists with drug references [2], pharmacoepidemiologic studies [25, 28, 33, 44], or prescribing indicators [28, 49] in order to include some medications and other instances of PIM use in older people (e.g., drug–drug interactions, drug class prescription duplication, special considerations of use, alternative therapies) which were not described in prior PIM lists.

We also verified that the majority of studies were developed for general practice. Few PIM lists focused on specific populations such as nursing home residents [22, 32, 42] and hospitalized patients [28, 40]. These were adaptations from existing PIM lists and included some new PIMs in their evaluation. For instance, some lists did not account for drugs frequently used during inpatient stays such as antibiotics. Thus, this result suggests that more work is needed to develop PIM lists for these populations and that some PIM lists originally designed for general practice could later be externally validated in these settings.

The Delphi technique was used to validate EC in the majority of the studies. This method is defined as an exercise in group communication that brings together and synthesizes the knowledge of a group of geographically distributed participants who have never meet [59]. Although there is no agreement on the definition of an expert, number of experts used, the number of rounds, and the consensus level in the literature, Delphi technique has some advantages over other consensus techniques such as the lack of discussion domination by any one panel member [60]. However, in this review, some studies [17, 18, 21, 27] modified the Delphi technique; these studies used a physical panel meeting at the end of consensus procedure in order to exchange views and resolve uncertainties.

We found that benzodiazepines and NSAIDs were the most common drugs reported as PIMs for older adults in all PIM lists. Previous systematic review also verified that these medication classes are among the most common reported in PIM lists [57]. However, these authors considered the number of indications of each medication class in each PIM lists while we evaluated the medication class included in each PIM lists. Of the 36 PIM lists evaluated, 33 described benzodiazepines as inappropriate. There is good observational data on the association between the use of benzodiazepine by older adults and serious ADEs, including impaired cognitive function [61, 62], delirium [63], respiratory insufficiency [64], falls [65], and fall-related injuries such as hip fractures [66]. Thus, they have the potential to create serious public health problems including hospitalization and death. Despite these risks, benzodiazepines are commonly used in the treatment of anxiety, depression, and insomnia in older patients around the world. Patients and providers hesitate to discontinue benzodiazepines because of the fear of withdrawal symptoms or relapse [67]. Studies show that there is a high prevalence of long-term use of this class in this age group, ranging 12 to 43% [68, 69].

There was very limited overlap between the PIM lists that we described in this study. Among all PIMs, only diazepam, chlordiazepoxide, indomethacin, and amitriptyline were considered inappropriate by 35 of the 36 PIM lists. Furthermore, only 44 medications and 4 medication classes were present in 69.0% or more of PIM lists. Prior systematic reviews also reported that only a few drugs are common to all the lists of PIMs published [57]. The heterogeneity in the lists of medications reflects the fact that medication management in older adults is extremely complex with a very limited evidence base to guide it. Additionally, health professionals from various fields were involved in the development of the PIM lists and they would, therefore, have different approaches and attitudes. As a consequence, the list of medications can vary widely.

We compiled all drug–disease interactions and drug–drug interactions included in the different PIM lists. It is interesting to note that NSAIDs were the most common medication class in both types of drug interactions. Despite the consistent recommendations to avoid the use of this medication class in different situations, it is estimated that 40% of people aged 65 years and older fill one or more prescriptions for a NSAIDs each year [70] with additional users accessing NSAIDs over the counter [71]. This, like the high utilization of benzodiazepines, may highlight the limited impact of the consensus on PIMs or that, while potentially inappropriate, the benefit may frequently be determined to outweigh the risk for the individual.

We identified the drug–drug interactions described in 20 PIM lists. Although a considerable proportion of adverse drug reactions is caused by interactions between drugs [72, 73], drug–drug interactions are still underreported in the criteria for assessing inappropriate prescriptions in older adults. Of the 159 drug–drug interactions identified, only 16 are described in more than 20% of the PIM lists. The concomitant use of NSAIDs and aspirin with warfarin was the most frequent drug–drug interaction described. Many studies have provided an increased risk of hospitalization in elderly adults using this combination of drugs [72]. Additionally, the warfarin was the most common single medication reported among the drug–drug interactions lists. Despite this medication is highly effective in the prevention of stroke in atrial fibrillation, it is known for its interaction with many drugs [72–73], which is the leading cause of adverse drug event-related hospitalizations in older adults and can lead to fatal outcomes in this population [74].

Strengths

This is the first study that systematically compiled all drug–disease interactions and drug–drug interactions included in validated PIM lists since 1991. This systematic review used a comprehensive search strategy applied by the reviewers without language limitations. Furthermore, the study followed the PRISMA methodology, including study selection performed by two independent reviewers with arbitration by a third party if necessary. This reduced the risk of studies being omitted and also reduced the risk of selection bias.

Limitations

Our review had some important limitations. EC are limited in that they do not address individual differences among patients or the complexity or appropriateness of entire medication regimens. Furthermore, they need to be regularly updated in line with the evidence, and country-specific adaptations are necessary where countries differ in their guidelines, standards, and approved medications. It is important to recognize that a detailed description of the consensus method was not included in some studies [26, 42, 43]. To our knowledge, there is no formal method for quality assessment or risk of bias for consensus studies, so a rigorous assessment of the quality/bias of each study could not be performed as required by the PRISMA criteria [58].

Conclusion

Appropriate mediation management among older adults can help prevent serious adverse drug events [3, 10] which are associated with the increase of hospitalization and mortality in this population. For this reason, approaches aimed at detecting inappropriate prescriptions have intensified in the last decades with the development and validation of a number of strategies, particularly PIM lists. These PIM lists are important educational tools and should be included in the comprehensive assessment of every older patient who requires medication. We identified 36 different PIM lists. Different medication/medication classes, drug–disease interactions, and drug–drug interactions were included in different lists, with limited overlap between the PIM lists presented. These results demonstrate that the use of medications in older people is complex field and that more evidence is required to be able to generate consistent expert recommendations and to implement them.

Our review highlights the most common PIMs, drug–disease interactions, and drug–drug interactions validated by expert consensus for over 26 years. These results can help health professionals to elaborate strategies to minimize use of PIMS in many different settings. Although benzodiazepines and NSAIDs were the most common medications classified as being inappropriate, they are still commonly used in older adults. Avoiding medication in which the risks outweigh the benefits in the elderly patient continues to be a challenge for health professionals. Some PIM lists are complex and did not provide special considerations of use and alternative medications to avoid those considered potentially inappropriate. In addition, few PIM lists provide information that supports safely tapering or withdrawing PIM. These facts may compromise the use of PIM lists in clinical practice. Future PIM lists should integrate information about alternative therapies and special considerations of use in order to help clinicians to make decisions about drug prescription.