Longitudinal effects of using and discontinuing central nervous system medications on cognitive functioning

To investigate the longitudinal effect of using and discontinuing central nervous system (CNS) medications on cognitive performance.

decline as well as increased risk of developing dementia. Moreover, it has been suggested that cognitive problems related to CNS medication use may be reversed by adjusting or discontinuing these medications. However, the longitudinal effects of CNS medications on cognitive performance are still not fully understood. In the present study, we investigated if use of CNS medications (opioids, anxiolytics, hypnotics and sedatives) is associated with cognitive functioning over time. Using longitudinal cognitive data from population-representative adults aged 25-100 years, we investigated both the association between CNS medication use, and the association between discontinuing these medications and cognitive task performance. Our results suggest that long-term use of CNS medications may have negative effects on cognitive performance over time, whereas the discontinuation of these medications may partly reverse these effects.

| INTRODUCTION
Medications acting on the central nervous system (CNS) frequently cause adverse effects, including problems with mobility, falls, and cognition in older individuals. [1][2][3] Results from longitudinal studies investigating cognitive effects of CNS medication use suggest that CNS medications may accelerate cognitive decline in older adults, [4][5][6][7] and it has been suggested that cognitive problems related to CNS medication use may be reversed by adjusting or discontinuing these medications. 8 Accelerated cognitive decline as well as increased risk of developing dementia has been mostly related to CNS medications with strong anticholinergic properties 8,9 including opioids, antidepressants, and antipsychotics, but also benzodiazepines or related drugs (anxiolytics, hypnotics, and sedatives) as well as antiepileptics. [4][5][6][7] A recent systematic review 10 on the impact of opioid use on cognition in older adults showed that both improvements and impairments of cognition have been observed, especially in studies with higher mean opioid doses. Whereas most of these studies demonstrated no effect of opioid use on task performance in any cognitive domain, other studies showed that attention, language, orientation, psychomotor function, verbal working memory, and episodic memory were worsened, especially at high mean doses of opioids. 10 It has also been shown that individuals using opioids at high doses had slightly higher dementia risk than individuals with little or no use of opioids. 11 However, these results may reflect cognitive effects of chronic pain, 11 and it has been suggested that in patients taking opioids for chronic pain, severity and psychological measures are more predictive of cognitive decline than opioid use. 12 Benzodiazepines and related drugs enhance the effect of the inhibitory neurotransmitter gammaaminobutyric acid (GABA) 13 that may modulate cognitive performance, 14 and their acute negative effects on cognitive functioning are well established. [15][16][17] Moreover, there is mounting evidence of an association between long-term use of benzodiazepines and impairments in a range of neuropsychological functions, and some evidence suggests a higher risk of decline in cognitive functioning of various cognitive domains as a result of long-term benzodiazepine use. 18 Benzodiazepine use has also been found to be strongly associated with an increased risk of dementia. 19 Since insomnia and anxiety are prodromal symptoms of dementia, it is difficult to determine whether this association arises from the treatment of those early symptoms prior to diagnosis or if benzodiazepines and related drugs are causing cognitive decline and dementia. 20 The longitudinal effects of CNS medications on cognitive performance are still not fully understood, and discrepancies between studies are likely due to methodological differences and inconsistencies regarding the cognitive domains affected. 21 In the present study, we investigated if use of CNS medications (i.e. opioids, anxiolytics, hypnotics and sedatives) is associated with cognitive functioning (episodic memory, semantic memory, visuospatial ability) over time. Using longitudinal cognitive data from population-representative adults aged 25-100 years, we investigated both the association between CNS medication use, and the association between discontinuing these medications and cognitive task performance. Dementia diagnosis was done by a geropsychiatrist based on the DSM-IV criteria 24 (the diagnosis procedure has been described elsewhere). 22,25,26 The research was approved by the regional ethical review board at Umeå University (EPN), and all participants gave written informed consent.

| Cognitive tests
The analyses were based on z-transformed tests of visuospatial ability, episodic memory and semantic memory. For episodic memory, a composite score was calculated as the sum of z-transformed tests of episodic memory (two tests of free oral recall of verb-noun sentences, two tests of category-cued recall of nouns from the sentence recall, and one test of free recall of presented nouns). For semantic memory, a composite score was calculated as the sum of z-transformed tests of semantic knowledge/verbal fluency (three tests of verbal generation of as many words as possible during 1 min: words that begin with A; Words that begin with M, exactly five letters; professions that begin with B). Visuospatial ability was measured with the block design task from the Wechsler Adult Intelligence Scale (WAIS-R). 27 For details about cognitive tasks see Nilsson et al. 2004. 22

| RESULTS
Descriptive statistics for each test occasion (T3-T6) are shown in Table 1. Descriptive statistics per group (individuals using CNS medications, individuals discontinuing CNS medications, and individuals not using any CNS mediations) are shown in Table 2. Based on a cognition composite score that was calculated as the sum of z-transformed tests of visuospatial ability, episodic memory and semantic mem-

| Effect of using CNS medications on cognitive performance
To analyze the association between use of CNS medications on cognitive intercept and slope, we compared individuals taking the same CNS drug at all their visits with 1-3 matched controls per individual and visit. As shown in Table 3, using opioids was associated with cognitive decline of visuospatial ability, whereas no associations were seen for anxiolytics, hypnotics and sedatives, even at uncorrected levels (Table 3)

| Effect of discontinuing CNS medications on cognitive performance
To analyze the effect of discontinuing CNS medications on cognitive performance over time, we included the two time-points of individuals that were taking a CNS drug at one visit but not at their next visit.
In relation to individuals not taking any CNS medications at both these time-points, discontinuing anxiolytics, hypnotics and sedatives was significantly associated with cognitive improvement in both episodic memory and visuospatial ability (Table 4). Analyses restricted to hypnotics and sedatives (21 cases and 58 controls) showed that the association between discontinuation and cognitive improvement was no longer significant for visuospatial ability (t value = 1.234, p value = 0.220), and only trend significant for episodic memory (t value = 1.935, p value = 0.055). Individuals discontinuing opioids significantly declined in visuospatial ability compared to individuals not using any CNS medications (Table 4). In relation to individuals using the same CNS medication at both time-points, discontinuation of both opioids and anxiolytics, hypnotics and sedatives was significantly associated with cognitive improvement in visuospatial ability (Table 5)   discontinuing these medications on cognitive task performance. Using longitudinal cognitive data from population representative adults aged 25-100 years, we found associations of CNS medication use and cognitive decline that were specific to both cognitive domains and medication class. Opioid use was associated with decline in visuospatial ability, and individuals that discontinued taking opioids had a more positive cognitive slope than individuals that continued using opioids 5 years lateryet more negative than those never taking opioids. For anxiolytics, hypnotics and sedatives there was no difference between continued users and non-users, but drug discontinuation was associated with more positive cognitive development both in relation to those that continued using these medications and non-users. In addition, we were able to show that concomitant use of drugs with effects on the CNS was associated with cognitive decline in a general population, which is in line with previous studies. 4,7 This effect may for example be mediated by anticholinergic properties of drugs such as tricyclic antidepressants. 15 Consistent with previous studies, we found no differences in memory decline between individuals using opioids and matched controls. 10 The observed greater cognitive decline in visuospatial ability in opioid users may be novel, as we are not aware of any longitudinal studies investigating the effects of opioids on cognitive decline specifically in tasks of visuospatial ability. This novel effect of cognitive decline in visuospatial ability is confirmed by our observation that individuals discontinuing using opioids show cognitive improvement in visuospatial ability compared to individuals continuing using opioids. Our observation of greater decline in visuospatial ability in individuals discontinuing using opioids compared to individuals not using opioids may indicate that the negative cognitive effects on visuospatial ability as a result of opioid use may remain to some degree after discontinuing these medications. However, we lacked measures of pain, and it has been suggested that chronic pain may be associated with impairments in cognitive functions. 28,29 Opioids are efficient analgesics and could improve cognitive function as a result of pain reduction. 30 On the other hand, it has been shown that opioids have negative effects on cognitive functioning also in healthy volunteers, 31 and that opioids induce apoptosis of microglia 32 pain and opioids on cognitive functioning, our data suggest that longterm opioid use might contribute to cognitive decline in visuospatial ability.
Whereas cognitive impairments have consistently been reported in users of both benzodiazepines 15,34 and related drugs increasing GABA transmission, 15,35,36 the current and most other longitudinal studies did not find benzodiazepine use to be associated with cognitive decline. 18,37 Although evidence regarding the association between benzodiazepine use and cognitive decline is conflicting, stronger links between benzodiazepine use and cognitive decline have emerged from studies investigating long-acting benzodiazepines. 37 A longitudinal study investigating the use of sedative drugs and incident cognitive decline in older individuals did not find significant associations between use of sedatives and cognitive decline. 38 However, we found that discontinuing anxiolytics and/or hypnotics and sedatives was significantly associated with improvement in visuospatial ability in relation to individuals continuing using these medications, and with improvement in both episodic memory and visuospatial ability in relation to individuals not using these medications. This indicates that any negative cognitive effects of using these medications may be reversed after withdrawal. However, analyses restricted to hypnotics and sedatives showed that the associations between discontinuation and cognitive improvement were no longer significant. This could be related to power issues due to the lower sample size, or the effects were possibly driven by anxiolytics.
Some limitations of the present study should be acknowledged.
Potential confounders that could influence the relationship between CNS medication use and cognitive task performance such as sleep problems or anxiety were not controlled for, as these data were not available. Further, a large part of individuals could not answer for how long they were using medications. Thus, data on duration of CNS mediation use were not included in the present study. Furthermore, data on self-reported medication use was limited to collection at the four time-points, potentially increasing the risk for recall bias. However, it can be viewed as a strength that medication use was based upon participants' reported actual medication use rather than pharmacy dispensing or records of medications prescribed to participants, as participants may not use all their prescribed medications. Another strength of the present study is the long follow-up time of 15 years. Most importantly, our analyses to investigate the effect of CNS medication use on cognitive decline were based on CNS medication use at all four occasions rather than using baseline medication data. Individuals that discontinued the respective CNS medication during the follow-up period were excluded from these analyses, and instead included in our analyses to investigate the effect of discontinuing CNS medications on cognitive change. It should also be noted that drugs may affect various cognitive domains and related brain areas in different ways. In our study, we include measures of episodic memory and semantic memory that capture both type of material to-be-remembered as well as both recall and recognition, and we also include measures of visuospatial ability. However, longitudinal effects of using and discontinuing CNS medications on other cognitive domains require further investigation in large population-representative samples.
In summary, although our results may be confounded by subjacent conditions, they suggest that long-term use of CNS medications may have domain-specific negative effects on cognitive performance over time, whereas the discontinuation of these medications may partly reverse these effects. These results open up for future studies that address subjacent conditions on cognition to develop a more complete understanding of the cognitive effects of CNS medications.

AUTHOR CONTRIBUTIONS
EK designed the study, performed analyses and wrote the manuscript.
All authors provided feedback on the manuscript and approved the final version.

FUNDING INFORMATION
This work was supported by a grant to KK from the Swedish Research Council (Grant no 2017-03011).