Substitution Patterns After Discontinuation of CNS‐Active Medications in Older Adults in Primary Care Authors: Shelly L Gray, Annalisa V Piccorelli, Laura A Hart, Andrea J Cook, Brian D Williamson, Benjamin H Balderson, Elizabeth A Phelan Published: 2026 ABSTRACT ABSTRACTBackgroundLittle is known about substitution of alternative medications in the context of deprescribing. The objectives were to: (1) determine the frequency of medication substitutions among those who discontinued a central nervous system (CNS)‐active medication, and (2) characterize substitutions as potentially inappropriate (as per the 2023 Beers Criteria) versus not.MethodsWe conducted a secondary analysis that combined data from the intervention and usual care arms from the STOP‐FALLS deprescribing trial that tested a health‐system‐embedded intervention designed to reduce prescription of CNS‐active medications. This analysis focused on participants followed for 360 days following baseline with chronic use of opioids, benzodiazepines, tricyclic antidepressants, skeletal muscle relaxants, or Z‐drugs. Discontinuation was defined as the first date when there was no evidence of a prescription fill for 90 days, thus only participants with a discontinuation that occurred in the first 270 days were included. A list of likely alternative treatments was developed for each target medication. A substitution was operationalized as a new alternative medication prescribed during the 30 days prior to or 60 days after discontinuation of a target medication.ResultsThe study sample included 2182 individuals (average age 70.5 years, 63.1% female). At baseline, a total of 2415 target medications were prescribed, of which 442 (18.3%) were discontinued. Discontinuation rates varied from 122 (8.0%) for users of opioids to 86 (49.7%) for users of skeletal muscle relaxants. Substitutions were made for 42 (9.5%) drug discontinuations. Of these substitutions, 11 of 42 (26.2%) were to a potentially inappropriate medication: tricyclic antidepressants (n = 5), benzodiazepines (n = 3) and hydroxyzine (n = 3). Other common substitutions included gabapentin, selective serotonin norepinephrine reuptake inhibitors, and trazodone.DiscussionOf medication discontinuations with a substitution, one‐quarter were to at least one potentially inappropriate medication. This finding highlights the need for additional guidance for prescribers to ensure safe deprescribing of CNS‐active medications.Keywords: aged, deprescriptions, potentially inappropriate medication list, primary health care Introduction Central nervous system (CNS)‐active medication use is prevalent in the older adult population [1], and many CNS‐active medications are considered potentially inappropriate due to their adverse safety profile. For example, benzodiazepines and nonbenzodiazepine benzodiazepine receptor agonist hypnotics (“Z drugs”) are associated with risks such as falls and fractures [2]. Anticholinergic medications are associated with risks such as confusion, delirium, and falls, with cumulative use associated with risk of dementia [2]. To date, several randomized trials have evaluated the impact of various interventions in the primary care setting to deprescribe CNS‐active medications in older adults, including opioids, benzodiazepines, or other sedative hypnotics [3, 4, 5, 6, 7], and anticholinergics [8, 9, 10]. A primary outcome in deprescribing trials is discontinuation of the target medication. In some cases, another CNS‐active medication may be substituted to manage the patient's underlying symptoms. Awareness of the patterns of medication substitution in the context of deprescribing is important because many medication substitutions may also be high‐risk and potentially inappropriate for older adults. For example, discontinuing a benzodiazepine and initiating a Z‐drug or hydroxyzine would continue to pose a concern for fall risk and other adverse outcomes. Evaluation of substitutions to other medication classes upon medication discontinuation, although infrequently described, has found that a small portion of participants that discontinued a benzodiazepine or Z‐drug had a substitution to other CNS‐active medications, including potentially inappropriate sedative‐hypnotics [3, 6]. Beyond these two trials, substitution to alternative medications following a deprescribing intervention has not been characterized in detail. Given the high prevalence of CNS‐active medication use in older adults and the importance of considering deprescribing of these medications, more information is needed regarding the patterns of substitution once these medications are discontinued. Data from the STOP‐FALLS deprescribing trial offered a unique opportunity to examine medication substitutions occurring in a real‐world practice setting [11]. The objectives of the present study were to: (1) determine the frequency of medication substitutions among those who discontinued a CNS‐active medication, and (2) characterize substitutions as potentially inappropriate (as per the 2023 Beers Criteria [2]) vs. not. Methods Data Source We used a prospective cohort design, combining data from the intervention and control arms of the STOP‐FALLs trial. In brief, STOP‐FALLS was a pragmatic, cluster‐randomized, parallel‐group, controlled trial that tested the effectiveness of a health‐system‐embedded deprescribing intervention in an integrated US healthcare system (ClinicalTrials.gov Identifier: NCT05689554) [11]. Eligible participants were chronic users of one or more target medications, defined as pharmacy dispensing of the medication for a minimum of 70 of the prior 90 days, aged 60 years or older for opioid and sedative‐hypnotic medications and aged 65 years or older for skeletal muscle relaxants, tricyclic antidepressants, and first‐generation prescribed antihistamines (i.e., hydroxyzine). The intervention was comprised of provider decision support and patient education via medication‐specific educational brochures and self‐care/symptom management handouts. The intervention did not reduce injurious falls over the follow‐up but did result in higher discontinuation of tricyclic antidepressants for up to 15 months compared with the control group. Details of the STOP‐FALLS study can be found elsewhere [11, 12]. Study Design and Sample We followed participants for 360 days after baseline, defined as the date the intervention was initiated in the STOP‐FALLs trial. Of the 2367 participants in the STOP‐FALLs trial, we excluded 37 that used first‐generation antihistamines as the only target medication, as it would be difficult to determine substitutions as these could be over‐the‐counter products with incomplete capture in the pharmacy data that served as the data source for the analyses herein. A full list of the target medications included in this analysis is listed in Supporting Information File, Table S1. Our definition of discontinuation was the first date when there was no evidence of a prescription fill for 90 days (Figure 1) [11]. Thus, we excluded participants that did not have a discontinuation of the target medications within 270 days of follow‐up (n = 148), as we were not able to evaluate medication discontinuation per our definition. Study Design and Definitions for Medication Discontinuation and Substitution Time 0 is baseline and participants were followed for 360 days to identify medication discontinuation. Our definition of discontinuation was the first date when there was no fill for 90 days. A substitution was operationalized as a new alternative medication prescribed during either the 30 days prior to or the 60 days following discontinuation of a target medication, with no fill in the period of 30–90 days prior to discontinuation to ensure “new use”. In this example, for a target medication discontinued at 180 days, substitution to a new alternative medication was defined as a fill of that medication between 150 days (30 days prior) and 240 days (60 days following) with no evidence of a fill between 90 days and 150 days (period of 30–90 days prior). Substitutions to Alternative Medications A priori, two members of the team (SG and EP) developed a list of medications for each target medication that were likely alternative treatments based on the conditions each medication is used to treat [Supporting Information File, Table S2]. We further classified substitutions into potentially inappropriate medications vs. not based on the AGS Beers Criteria [2], with the following medications considered potentially inappropriate: benzodiazepines, tricyclic antidepressants, Z‐drugs, and hydroxyzine. We did not include nonsteroidal anti‐inflammatory drugs (NSAIDs) as a potentially inappropriate medication because this would require knowing the duration of use and whether the participant received gastroprotection; this information may not be available as NSAIDs and proton pump inhibitors are often self‐administered without a prescription. There is not a gold standard for defining medication substitution using automated pharmacy data. For this study, a substitution was operationalized as a new alternative medication prescribed during either the 30 days prior to or the 60 days following discontinuation of a target medication and there were no fills for the new medication in the period of 30 to 90 days prior to discontinuation (Figure 1). Analysis For participants for multiple target medications, we focused on the first medication discontinuation. We present descriptive statistics of the study sample and the proportion of individuals who discontinued a target medication over the study period, the proportion of individuals who had a substituted alternative medication, and whether the new medication was potentially inappropriate or not. Results Study Sample The study sample included 2182 participants with an average age of 70.5 (SD 7.5) years; 63.1% were female and the majority White. About one‐third had a fall requiring medical care in the prior 12 months, and 3.3% were considered frail. At baseline, opioids were the most prescribed target medication, with 70.2% of the sample using them, followed by benzodiazepines (13.9%) and tricyclic antidepressants (11.8%). Other baseline characteristics are shown in Table 1. TABLE 1Baseline characteristics.CharacteristicOverall (N = 2182) a Demographic variablesAge, years, mean (SD)70.5 (7.5)Female1377 (63.1)Self‐reported raceAmerican Indian or Alaska Native27 (1.2)Asian35 (1.6)Black or African American61 (2.8)Native Hawaiian or Pacific Islander10 (0.5)White1904 (87.3)Other race b 30 (1.4)Multiple races50 (2.3)Race unknown or not reported65 (3.0)Hispanic ethnicity70 (3.2)Health StatusUse of mobility aid28 (1.3)Difficulty walking20 (0.9)Number of chronic conditions, mean (SD)2.2 (1.5)Frailty status c Non‐frail997 (45.7)Pre‐frail1114 (51.1)Frail71 (3.3)Musculoskeletal pain1230 (56.4)Medical care for a fall in past 12 months623 (28.6)Medication classOpioids1531 (70.2)Benzodiazepines304 (13.9)Tricyclic antidepressants257 (11.8)Muscle relaxants173 (7.9)Z‐drugs153 (7.0) a For categorical variables, we report n (%); for continuous variables, we report mean (SD). b Other race as reported by the participant in their electronic medical record field (e.g., Middle Eastern; North African). c Frailty assessed based on Kim's Claims‐Based Frailty Index. Baseline characteristics. For categorical variables, we report n (%); for continuous variables, we report mean (SD). Other race as reported by the participant in their electronic medical record field (e.g., Middle Eastern; North African). Frailty assessed based on Kim's Claims‐Based Frailty Index. Discontinuation of CNS‐Active Medications Participants were prescribed 2415 target medications at baseline, of which 442 (18.3%) were discontinued during follow‐up. Discontinuation rates varied by medication class, ranging from 8.0% to 49.7%. Patterns of medication discontinuation and substitution among users of the target CNS‐active medication classes are depicted in Table 2. Medication discontinuation was highest for users of skeletal muscle relaxants (86 of 173 users, 49.7%), followed by users of benzodiazepines (138 of 303 users, 45.5%), users of tricyclic antidepressants (66 of 257 users, 25.7%), users of Z‐drugs (30 of 153 users, 19.6%), and users of opioids (122 of 1529 users, 8.0%). TABLE 2CNS‐active medication use, discontinuation, and therapeutic substitutions (n = 2182).Use at baseline N (%)Discontinued N (%)Discontinued drugs with a substitution N (%)Substituted medications (N) a , b Opioids1529122 (8.0)13 (10.7)ASD (1); Gabapentinoids (4); NSAID (2); SSRI (1); SNRI (3); TCA (3); Topical analgesic (2)Benzodiazepines303138 (45.5)7 (5.1)Gabapentinoids (1); Hydroxyzine (2); Mirtazapine (1); SSRI (2); SNRI (2); Trazodone (1)Tricyclic antidepressants25766 (25.7)6 (9.1)Mirtazapine (1); SSRI (1); SNRI (2); Trazodone (2); Topical analgesic (1)Skeletal muscle relaxants17386 (49.7)11 (12.4) Benzodiazepines (3); Gabapentinoid (2); NSAID (1); SNRI (1); Topical analgesic (4)Z‐drugs15330 (19.6)5 (16.7)Gabapentinoid (1); Hydroxyzine (1); TCA (2); Trazodone (2)Abbreviations: ASD = Antiseizure drugs; NSAID, nonsteroidal anti‐inflammatory drugs; SNRI, selective serotonin norepinephrine reuptake inhibitors; SSRI, Selective serotonin reuptake inhibitors; TCA, tricyclic antidepressant. a Out of number that discontinued drugs and had a substitution (prior column). b Substitutions that were considered potentially inappropriate medications according to the 2023 Beers Criteria are bolded: tricyclic antidepressants, hydroxyzine, benzodiazepines, and Z‐drugs. CNS‐active medication use, discontinuation, and therapeutic substitutions (n = 2182). Abbreviations: ASD = Antiseizure drugs; NSAID, nonsteroidal anti‐inflammatory drugs; SNRI, selective serotonin norepinephrine reuptake inhibitors; SSRI, Selective serotonin reuptake inhibitors; TCA, tricyclic antidepressant. Out of number that discontinued drugs and had a substitution (prior column). Substitutions that were considered potentially inappropriate medications according to the 2023 Beers Criteria are bolded: tricyclic antidepressants, hydroxyzine, benzodiazepines, and Z‐drugs. Substitutions to Alternative Medications Substitutions were made for 42 of 442 (9.5%) drug discontinuations and included 49 medications. Substitutions varied from 5.1% to 16.7% according to medication class. For participants with medication discontinuation, substitution to alternative medications occurred for 7 of 138 (5.1%) discontinuations of benzodiazepines, 6 of 66 (9.1%) discontinuations of tricyclic antidepressants, 5 of 30 (16.7%) discontinuations of Z‐drugs, and 13 of 122 (10.7%) discontinuations of opioids. Of the 42 medication discontinuations with substitutions, 11 (26.2%) were to a potentially inappropriate medication: tricyclic antidepressants (n = 5), benzodiazepines (n = 3) and hydroxyzine (= 3). Other common substitutions included gabapentin (8 substitutions, 19.0%), selective serotonin norepinephrine reuptake inhibitors (8 substitutions, 19.0%), and trazodone (5 substitutions, 11.9%). Discussion We performed a secondary analysis of data collected from a real‐world deprescribing trial to describe medication substitutions for those who discontinued a CNS‐active medication over 270 days. Discontinuation rates varied considerably across the target medication classes, from 8.0% for those prescribed opioids to 49.7% of those prescribed skeletal muscle relaxants. Substitutions were made for 42 (9.5%) drug discontinuations. Of these medication discontinuations with a substitution, 11 of 42 (26.2%) were to at least one potentially inappropriate medication. Literature regarding substitution of a medication following medication discontinuation is sparse. Two prior deprescribing trials have described the extent of medication substitution after CNS‐active medication discontinuation, with a specific focus on benzodiazepines and Z‐drugs [3, 6]. The EMPOWER trial found that, among participants who discontinued benzodiazepines in the intervention group, 13% experienced substitution with an antidepressant medication, specifically trazodone, paroxetine, or amitriptyline, within 6 months [3]. The D‐PRESCRIBE trial found that, among participants who discontinued a sedative‐hypnotic medication, 8.6% received substitution of another potentially inappropriate sedative‐hypnotic medication [6]. We found that 1.4% of participants who discontinued a benzodiazepine and 10.0% of participants who discontinued a Z‐drug experienced a medication substitution that was considered potentially inappropriate; neither group received the alternative treatment (i.e., Z‐drug substituted for a benzodiazepine or vice versa). Of participants that had a medication discontinuation with a substitution, the most common medications were gabapentin (19%) or selective serotonin norepinephrine reuptake inhibitors (19%). Guidance from the American Geriatrics Society publication regarding Beers Criteria alternatives lists selective serotonin norepinephrine reuptake inhibitors as alternatives to benzodiazepines for anxiety and tricyclic antidepressants for neuropathic pain. Gabapentin is also considered an alternative to tricyclic antidepressants as long as it is not used concurrently with an opioid or the patient does not have a history of falls [2, 13]. Gabapentinoids have been increasingly used, particularly as an alternative analgesic to opioids or as an alternative sedative‐hypnotic to benzodiazepines and Z‐drugs [14, 15, 16]. Although gabapentinoids are often viewed as safer alternatives to these agents, their use has been associated with adverse outcomes such as cognitive changes, delirium, pneumonia, and increased antipsychotic use [14, 17]. Overall, more attention should be paid to judicious use of gabapentinoids, especially among older adults with a history of falls or fractures [18]. In our study, trazodone represented 11.9% of substitutions. Trazodone has long been prescribed off‐label for management of insomnia in both the inpatient and outpatient setting and, even among geriatrics‐trained providers, has been considered safe. However, common side effects include dizziness, weakness, blurred vision, and sedation. At higher doses, trazodone may cause orthostatic hypotension, a modifiable risk factor for falls. Research has found that the risk of falls with use of trazodone is comparable to that of benzodiazepines and antipsychotics in high‐risk subgroups [19, 20]. Therefore, consideration of alternative approaches to insomnia management in older adults at increased risk of falls is prudent [13]. When a trial of cognitive behavioral therapy for insomnia has not been effective for improving sleep, judicious short term use of pharmacologic agents can be considered, including oral doxepin 3 or 6 mg, ramelteon, or an orexin receptor antagonist [13]. A strength of the present study is that it provides further insight into medication substitution following discontinuation of high‐risk medications, notably those not evaluated in prior literature, including opioids, skeletal muscle relaxants, and tricyclics. A priori, we developed a list of alternatives for each target medication and leveraged automated pharmacy data to systematically identify substitutions. We examined substitutions of benzodiazepines and Z‐drugs separately to understand nuances in the deprescribing process. However, there are some limitations. Defining medication substitution based on pharmacy refill patterns is challenging as our definition of substitution is based on the timing of new medications in relation to the discontinuation of the target medication and may not truly represent a substitution. We combined data from the intervention and control arms; thus, discontinuation could reflect the effect of the intervention or the natural course of medication use. Nonetheless, understanding substitutions to alternative medications remains important irrespective of the process. It is likely that our findings represent an underestimate of substitutions to potentially inappropriate medications, as we were not able to capture participants' use of over‐the‐counter sleep aids (first‐generation antihistamines). Our sample had medical insurance and was predominantly White; therefore, results may not generalize to more diverse populations. In summary, we found that, although medication substitutions occurred infrequently, for medication discontinuations that had a substitution, over one‐quarter were to a potentially inappropriate medication. This finding suggests that additional guidance for prescribers regarding safer alternative medications is warranted. The recent American Geriatrics Society publication regarding Beers Criteria alternatives may serve as a useful guidance document for those seeking to optimize management of common symptoms in older adults and avoid use of high‐risk CNS‐active medications [13]. Author Contributions Study concept and design: S.L.G., E.A.P., A.V.P., A.J.C.; Acquisition of subjects and/or data: B.D.W., B.H.B.; Analysis and interpretation of data: S.L.G., E.A.P., A.V.P., L.A.H., B.D.W., B.H.B.; Preparation of manuscript: All authors. Funding This study was funded by the Centers for Disease Control and Prevention (U01CE002967). All statements in this report, including its findings and conclusions, are solely those of the authors and do not necessarily represent the views of the CDC. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Table S1: Target Medications in STOP‐FALLS Examined in this Analysis. Table S2: Alternative medications are denoted by x for each STOP‐FALLs Target Medication. Acknowledgments The authors have nothing to report. References Associated Data Supplementary Materials Table S1: Target Medications in STOP‐FALLS Examined in this Analysis. Table S2: Alternative medications are denoted by x for each STOP‐FALLs Target Medication.