Verbally Agitated Behavior scores of groups 1 and 2 for each assessment (evening shift).
Cohen-Mansfield J, Lipson S, Werner P, Billig N, Taylor L, Woosley R. Withdrawal of Haloperidol, Thioridazine, and Lorazepam in the Nursing HomeA Controlled, Double-blind Study. Arch Intern Med. 1999;159(15):1733-1740. doi:10.1001/archinte.159.15.1733
Copyright 1999 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.1999
Ongoing regimens of haloperidol, thioridazine, and lorazepam are commonly administered to manage behavior problems in nursing home residents. Nevertheless, there is controversy over whether periodic drug withdrawal should be attempted when those medications are prescribed. This study addressed that issue by examining the effects of discontinuing treatment with haloperidol, thioridazine, and lorazepam among residents of a large suburban nursing home.
In a double-blind, crossover study, half of 58 nursing home residents (43 women and 15 men with a mean age of 86 years) continued to take the psychotropic medication they had been prescribed, whereas the other half were tapered to placebo. After 6 weeks of taking placebo or original drug, patients were tapered to the reverse schedule and remained on it for 6 weeks. Assessments included informant ratings by the nursing staff who completed the Brief Psychiatric Rating Scale and the Cohen-Mansfield Agitation Inventory.
Analyses comparing residents taking placebo to those taking medication after completion of the first phase showed no impact of drug therapy discontinuation on their behavior. Similarly, using the crossover design to compare residents' behaviors while taking placebo vs taking drugs, withdrawal of medication had no impact on Cohen-Mansfield Agitation Inventory or Brief Psychiatric Rating Scale scores.
Results of this work suggest that long-term use of haloperidol, thioridazine, and lorazepam in nursing homes to manage agitation should be closely monitored for their efficacy. Furthermore, routine attempts at drug withdrawal should be considered for most residents taking psychotropic medication.
THE COMMON administration of haloperidol, thioridazine, and lorazepam to manage behavioral problems in nursing homes has been amply documented.1- 4 As a result of the recognition of this high (or excessive) use and potential adverse effects, the Nursing Home Reform Amendments of the Omnibus Budget Reconciliation Act (OBRA-87) were mandated. These amendments aimed to ensure that psychotropic medication is prescribed only when a specific diagnosis has been made and that dose reductions and behavioral modifications be attempted. The regulations required trials of discontinuation of treatment with psychotropic drugs. This approach has proven successful, with several studies5- 8 documenting the accomplishment of psychotropic drug withdrawal in nursing home residents. Most of this success could, however, be attributed to excessive or inappropriate prescribing.
Although OBRA regulations have had an effect on psychotropic drug use in nursing homes, significant levels of these drugs continue to be prescribed.9 Whereas the magnitude of efficacy of psychotropic drugs in treating agitation is under debate, there seems to be an initial effect from the use of neuroleptics and benzodiazepines.10,11 The question therefore remains, when such drugs are prescribed within currently accepted standards of therapeutics, should drug withdrawal be attempted? If so, under what conditions and at what times? These issues are at the heart of much controversy, portions of which this article aims to address.
Several studies examined psychotropic drug withdrawal as a result of OBRA-87 regulations in nursing homes, where there may have been overuse of such medications and reported beneficial results to drug withdrawal. Fitz and Mallya6 examined the impact of dosage reductions in psychotropic medications among nursing home residents within a program of "assessment, therapeutic programming, and behavior and medication management." They found that reduction of psychotropics resulted in a 2-week increase, followed by an overall reduction in behavioral difficulties (as measured by use of restraints, incident reports, and pro re nata medication use) 8 weeks later. A control group without the program did not experience drug reductions. They concluded that successful withdrawal of psychotropic medications can occur in a nursing home only in the presence of psychogeriatric programming. Avorn et al5 showed that an educational program targeted to physicians, nurses, and aides resulted in a reduced use of psychotropic drugs in nursing homes, without adversely affecting overall behavior or level of functioning. There were, however, increased reports of depressive symptoms. Thapa et al8 examined the effect of antipsychotic drug withdrawal in elderly nursing home residents of 12 community nursing homes that participated in a randomized controlled trial of an educational program designed to reduce antipsychotic drug use. Compared with patients who continued to take the medication, those whose treatment with antipsychotic medications was discontinued showed a significant reduction in overall psychiatric symptoms as measured by the Brief Psychiatric Rating Scale (BPRS). Furthermore, they also reported that these patients manifested less agitation and aggressive behavior, although not statistically significant.
Elderly persons are particularly susceptible to adverse effects of psychotropic medication, including akathisias, tremors, bradycardia, depression, orthostatic hypotension, sedation, and tardive dyskinesia, and are more vulnerable to adverse drug interactions. Furthermore, it is possible that some agitated behaviors are themselves induced by adverse effects or interactions.12
The uncertain long-term efficacy of haloperidol, thioridazine, and lorazepam for agitation, combined with the possibility that their adverse effects or interaction with other drugs might actually induce the behaviors for which they are administered, underscores the need for further study of their utility in the treatment of agitation. Even when the use of those psychoactive medications is appropriate, it is unclear how long that treatment should be continued because the natural course of agitation and the pharmacologic response are unknown. The present study addresses these issues in a random assignment, placebo-controlled, double-blind, crossover study in which the withdrawal of psychoactive medication—specifically administered for agitation—was examined among residents of a nursing home. The research questions were as follows: (1) When haloperidol, thioridazine, and lorazepam are prescribed according to current standards of care, are they effective in managing agitation? What is the impact of withdrawal on agitated behaviors? (2) What functions (if any), other than agitation, are affected by withdrawal?
The study was reviewed and approved by the institutional review board of the Hebrew Home of Greater Washington, Rockville, Md.
Participants were residents of a 550-bed, skilled, nonprofit nursing facility. The facility was chosen specifically because its use of psychotropic medication conforms to OBRA regulations, the staff practice continuous follow-up and monitoring of residents who take psychotropic drugs, and all staff had undergone considerable education concerning OBRA regulations before initiation of the study.13- 15 Medical care was given by 5 full-time, salaried, board-certified physicians. The residents of this nursing home reflect national norms for residents of skilled nursing facilities in terms of distribution of age, sex, and functional impairment.
All participants of the study fulfilled the inclusion criteria. For at least 4 weeks, they were receiving either haloperidol, thioridazine, or lorazepam for agitation, and were residents of the facility so that nursing staff knew them well enough to assess them. Residents were included only if older than 70 years, given that more than 90% of nursing home residents are older than 70 years and that younger age is indicative of an unusual condition and is, therefore, of limited generalizability.
Potential participants were not entered into the study if they met the following exclusion criteria: concomitant administration of other antipsychotic or antianxiety drugs other than low-dose trazodone hydrochloride for sleep; life expectancy of less than 3 months due to obvious causes, as judged by the nursing staff member responsible for direct care; psychiatric diagnosis of a major affective disorder of schizophrenia, according to the Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition16; or an acute infection within 10 days before entry, because data17 and clinical observations suggest that acute infection is correlated with increased levels of agitation (these patients were eligible for inclusion after the infection was effectively treated). Furthermore, residents were not included if they were expected to leave the nursing home (either to the hospital or to home) within 3 months, or if they had uncontrolled hyperglycemia or hypoglycemia.
A written statement of informed consent was obtained for all participants. Any resident whose level of cognitive function was sufficiently intact (as judged by the nursing staff member who was well acquainted with the participant) was asked to provide consent for himself of herself. For residents unable to provide informed consent, a legally appointed guardian or family member responsible for their care was contacted and asked to provide consent for them.
Of 362 residents originally considered for the study, 251 were excluded (Table 1). The most common reasons for exclusion were infrequent pro re nata medication use (73), medication use discontinued before entering the study (45), diagnosis of major affective disorder (37), and death before starting the study (30). Of those eligible for the study, 53 refused to participate. Fifty-eight nursing home residents (43 women and 15 men with a mean age of 86 years) participated in the study. Participants were similar in their demographic characteristics to those who were considered but did not participate (Table 2).
Eligibility for participation in the study was ascertained monthly based on reviews of pharmacy records and charts for these individuals. Eligibility was then confirmed with the care team, after which consent was sought.
Each participating resident was assessed for cognitive functioning, agitation, and adverse effects. At this stage, the first phase of the study was initiated, wherein half the residents were randomly assigned to have their medication dose tapered during a 3-week period, followed by receipt of a placebo (the other half continued their usual medication dosage). After 7 weeks of taking either placebo or medication, the second phase of the crossover study was undertaken, in which the group taking placebo was titrated back to the initial dose and the group taking the drug was tapered to placebo. Each group, therefore, continued to take the new drug-placebo dose for 7 weeks. Residents were randomly assigned to the placebo vs medication group and stratified both by level of cognitive functioning (1-3 vs 4-7 on the BCRS18) and by psychotropic medication (haloperidol, thioridazine, or lorazepam). Study medications (usual medication and placebo) were administered as identical liquids to ensure blindness by the care team. Only the dispensing pharmacist, who was not an employee of the nursing home, knew which medication was administered. The care team, residents, family caregivers, and research team were blinded to which group a participant was assigned.
All participants (ie, medication and withdrawal groups) were monitored by nursing staff to ensure that withdrawal did not produce effects that threatened their well-being. In addition to daily evaluations, residents were formally assessed by the research team during the second week of baseline, 1 week after the beginning of dosage taper, and after 6 weeks of full withdrawal for a total of 5 assessments: baseline, phase 1 tapering, phase 1 end point, phase 2 tapering, and phase 2 end point.
Assessments concerned primary outcomes in the domain of behavior and agitation, and secondary outcomes in the domains of resident functioning, adverse effects, and global impressions of functioning.
The BPRS19,20 was used because of its wide use in clinical psychopharmacologic trials; its psychometric properties for young and elderly persons have been documented.21,22 The BPRS was assessed by daytime and evening nursing staff and by the unit's social worker. Symptoms assessed included the following: somatic concern, anxiety, emotional withdrawal, conceptual disorganization, guilt feelings, tension, mannerisms and posturing, grandiosity, depressive mood, hostility, suspiciousness, hallucinatory behavior, motor retardation, uncooperativeness, unusual thought content, and blunted affect. Each symptom was rated on a severity scale ranging from 1 (not present) to 7 (extremely severe).
Agitation was measured via the Cohen-Mansfield Agitation Inventory (CMAI).23 This nurses' rating questionnaire consists of 29 agitated behaviors, each rated on a 7-point scale of frequency (1 indicates the participant never engages in that specific behavior, and 7 indicates the participant manifests that behavior an average of several times an hour). The period rated was the week before administration of the CMAI, and 3 subtypes of agitation were used (further details, reliability, and factor analyses described by Cohen-Mansfield et al23): aggressive behavior, physically nonaggressive behavior, and verbally agitated behavior. The CMAI was independently completed by daytime and evening shift nursing staff members most familiar with the resident, and by social workers.
Direct measurement of cognitive functioning was obtained via the Mini-Mental State Examination.24 Scores range from 0 to 30, with higher scores indicating better cognitive functioning.
We used the same measures as in our previous research25Daytime sleep was an average of the items "How often does the resident appear drowsy or sleepy during the day?" and "How frequently does the resident actually sleep during the day?" Both items were rated by the daytime nursing staff member most familiar with the resident on a 7-point frequency scale ranging from 1 (never) to 7 (several times an hour). The Pearson correlation between these 2 items was 0.66 at baseline, 0.85 at week 12, and 0.76 at week 22 (all significant at P<.001). Time to fall asleep was rated by evening shift nursing staff and asked, "On the average, how long did it take the resident to fall asleep at night (from the time he/she went to bed until the time the resident fell asleep)?" This was rated on a 6-point scale ranging from 1 (falls asleep immediately) to 6 (nearly never sleeps at night). Activity level was an average of 2 items: "How often did the resident socialize or participate in social activities with friends, acquaintances, family, or other residents?" and "How frequently was the resident involved in activities, such as music, occupational therapy, art, needlework, taking a walk, or engaging in any other activity which is meaningful for his/her level of functioning?" Items were rated by daytime and evening staff on a 6-point frequency scale ranging from 1 (never) to 6 (several times a day). Pearson correlations between the 2 items were 0.51, 0.64, and 0.63 for daytime staff and 0.58, 0.68, and 0.50 for evening staff for the 3 time points described herein (P<.01).
Positive mood was measured by 2 items: "How often does he/she seem content (ie, happy enough with what one has or is, satisfied, pleased, content) during the past week?" and "How often does he/she seem to be in good spirits (ie, happy, smiling, laughing, cheerful) during the past 2 weeks?" Each item was rated by daytime and evening nursing staff on a 7-point frequency scale ranging from 1 (never) to 7 (all the time). Pearson correlations between the items were 0.42, 0.45, and 0.46 for daytime nursing staff and 0.49, 0.46, and 0.48 for evening staff for the time points described herein (P<.01).
Weight was recorded as the last weight in the resident's chart; residents were weighed by nursing staff every 4 weeks.
The Abnormal Involuntary Movement Scale26 assesses neurologic and physical side effects associated with psychotropic medication. It was measured by a trained research assistant and includes 9 items concerning movement of the face and oral cavity, extremities, and trunk, and global judgments of abnormal movements.
Adverse effects of psychotropic medication include sedation, extrapyramidal reactions, orthostatic hypotension, and anticholinergic effects, eg, dry mouth, constipation, and blurred vision. A list of these symptoms was provided to nursing staff, who indicated their frequency of occurrence.
Nurse managers checked lists of movement adverse effects, including 13 items describing pseudoparkinsonism, akathisia, acute distonic reactions, and tardive dyskinesia.
A Global Clinical Impression Scale was rated by daytime and evening nursing staff and social workers. The Global Clinical Impression Scale was "How would you rate the general condition/disposition of the resident this week," with a 7-point scale ranging from 1 (extremely positive) to 7 (extremely negative).
The BCRS17 was used at the beginning of the study to determine the participants' stage of dementia, to characterize the study population, and to stratify groups. It was rated by nursing staff members most familiar with the participants. The BCRS was revised for use in the nursing home (ie, using only examples that are applicable in the nursing home) and included only the first 4 axes (concentration, recent memory, past memory, and orientation), since self-care was addressed specifically during assessment of daily activities. This revised instrument has been used in previous research27 with excellent interrater agreement. The BCRS is rated on a 7-point scale ranging from 1 (normal cognitive functioning) to 7 (complete deterioration).
Nursing staff completed Lawton and Brody's Physical Self-Maintenance Scale28 subscales of toileting, eating, dressing, grooming and bathing, for which an average rating was calculated. Scores ranged from 1, indicating complete independence in performing the activity, to 5, denoting complete dependence. Good reliability and validity was established.29
Demographic variables were recorded from each resident's chart by the research assistant. Medical variables were obtained from the resident's physician.
Nursing staff expectations regarding withdrawal of the drug were solicited at the beginning of the study by asking: "In what way do you predict the resident's behavior will change?"
The main question of our study was whether the behavior of the resident was less agitated when taking the drug compared with taking placebo. To answer this question, several approaches were taken. First, paired t tests were conducted for drug vs placebo conditions during assessments 3 and 5 (ie, comparison of end point assessments of each phase in the crossover design). This approach was undertaken to maximize the power of the analysis. Second, only the first portion of the study was considered, and a 2-way repeated-measures analysis of variance (ANOVA) was conducted, in which the within-subject variable was assessment 1 (basement) vs assessment 3 (phase 1 end point) and the between-subject variable was the randomization group. This approach maximized sample size, as in a parallel group design. Third, a repeated-measures ANOVA was conducted with a within-subject comparison of assessment 3 vs assessment 5 and a between-subject comparison of randomization groups. Fourth, a repeated-measures ANOVA was conducted as just described, with adjustment for BCRS and number of medical diagnoses.
Twenty-three participants discontinued participation in the study before completion for the following reasons: death or dying (3), hospitalization (1), not eating or weight loss (3), increased agitation (9), lethargy (2), withdrawal of consent (4), facial asymmetry (1), and a fall (3); some had multiple reasons (Table 3). For 12 participants, discontinuation occurred during the original drug dosage, for 9 while taking placebo, and for 2 during titration from drug to placebo. Most discontinuations (20 of 23) occurred in the first part of the study, before the crossover stage.
Participants who discontinued the study were similar in demographic characteristic to those who stayed (Table 4). Although their levels of agitation at baseline were higher than those who stayed in the study, these differences did not reach statistical significance (the only difference close to statistical significance was physically nonaggressive behaviors; P=.07 with a 2-tailed t test).
All 4 approaches described in the analytic section of this article were used to compare primary behavioral outcome under the drug and placebo conditions. To enhance the power of the analysis, all medications were grouped, as were the randomization groups, and the main outcome variables (BPRS and CMAI) were compared during drug vs during placebo phases via a paired t test analysis, with 35 individuals. Results for the first and third approaches are described in Table 5. All approaches resulted in nonsignificant differences for all variables. Furthermore, no consistent trends were observed. For example, the BPRS score was actually higher during the drug phase than during placebo, and, though mean differences in agitated behaviors differed in their direction by type of agitation and by rater, none were significant.
Comparison of the variables assessing functioning, adverse effects, and global impression all resulted in nonsignificant differences between drug and placebo conditions, whether compared via paired t tests or by the interaction term of an ANOVA with stratification group as the between-subjects factor and study phase as the within-subject factor.
If the psychotropic drugs have effects described in the literature, the following outcomes would be expected: behavior problems would increase while taking placebo in comparison with taking the drug, and adverse effects would decrease and functioning level would improve while taking placebo compared with taking the drug.
Given that these notions refer to a specific type of change (increase or decrease), a 1-tailed probability could be used with the paired t test. Accordingly, the medications were effective in controlling verbal agitation during the evening (P=.03), and cognitive functioning improved during the taking of a placebo compared with the taking of a drug (P=.05). Given the large number of comparisons made, the significance level used should be lower than .05, and these results should be regarded only as suggesting the need for additional study in this area.
The number of residents included in the study were nearly as many as those excluded because of discontinuation by their physicians outside the framework of this study (n=45; Table 1). To explore the residents' responses to discontinuation, information regarding the psychotropic medications they took during 22-week follow-up to discontinuation of the original drug was obtained from their medical charts. Results showed that less than a third (13) remained free of all psychotropic drugs for the follow-up period; 20% (9) had the drug reinstated; and 67% (30) received another psychotropic medication. Seven (16%) of the 45 received a combination of the original drug and another psychotropic medication. For the 9 participants for whom treatment with the original medication was restarted, the average hiatus was less than 4 weeks (24 days), with a range of 2 to 134 days. The time to receiving another psychotropic drug averaged 21 days, with a range of 0 to 123 days (n=30).
At initiation of the study, more than half (54%) of the nursing staff predicted a deterioration in the residents' behavior following drug withdrawal; 19% predicted no change; and the rest claimed they could not predict the effects of withdrawal.
The major finding of this study is that drug withdrawal trials are essential for most nursing home residents who have been taking haloperidol, thioridazine, or lorazepam, even when prescription and follow-up are within current standards of care. Participants were taking medication for a median of 9 months (average, 16.5 months) before starting the study (range, 1.6-95.2 months). The failure of these medications to exert an effect was found both in comparisons of behavior ratings and in analyses of residents who discontinued the study before its termination. The lack of effectiveness was especially significant to the investigators because study discontinuations usually occurred when a staff member called the researchers and expressed a sense of emergency. Most often, the nursing staff implied that it was probably the discontinuation of medication that caused any worsening of behavior. It was therefore an unanticipated finding after breaking the code at the end of the study to find that as many discontinuations of the study occurred with the original medication and dosage as with drug withdrawal.
Were the drugs ineffective? As shown by Table 5, even a trend in the direction of effectiveness was not shown by most of the behavioral indicators studied; the only possible exception to this pattern of outcome being that less verbal agitation occurred during the evening nursing shift when participants were taking the drug than when they were taking the placebo. A closer look at the means of all assessments for this group reveals that the effect was caused by an unexplained decline in verbal behavior by the group that continued to take the drug (Figure 1) during the first phase of the study. These participants showed a decrease in verbal agitation while continuing to take the original drug. They resumed their baseline levels while taking placebo. Relatively little change was seen in the other subgroup.
Why were drug effects nonexistent? We monitored whether the residents were indeed taking the medications in a sample of drug administration records and found a drug administration rate of 100%. Drug dosages were relatively low, with means of 1.03 mg/d of haloperidol (range, 0.125-5.0 mg), 25.75 mg/d of thioridazine (range, 5.0-100.0 mg), and 0.79 mg/d of lorazepam (range, 0.25-1.5 mg). However, low dosages are endorsed in the literature and are currently prescribed by geriatric physicians. It is possible that, even when an initial effect is displayed, it is of limited duration. Hence, drug therapy should be discontinued after an initial period to examine whether the beneficial effect continues. This research therefore supports OBRA regulations' mandate of trials of psychotropic drug withdrawal. Such a withdrawal should be undertaken, even in cases in which monitoring is available and the care team is convinced of the drugs' benefit.
Although haloperidol, thioridazine, and lorazepam did not improve behavior, they had very little if any negative effect on functioning or adverse effects. Withdrawal of the medications did not improve functioning, with the possible exception of a marginally significant improvement on the Mini-Mental State Examination. The lack of impact on functioning can be explained in 2 ways: (1) Had any adverse events or significant deterioration in function occurred, the care team would have discontinued treatment with the medication before the resident became a candidate for this study. (2) When medication has been taken for so long, the effects on functioning are no longer reversible.
This study highlighted some of the difficulties in reducing haloperidol, thioridazine, and lorazepam doses in this population. The care team, including nursing staff, physicians, and family members, has a great deal of faith in the utility of these drugs to control agitation in this population. As seen from the responses, more than half of staff were convinced that a trial withdrawal would result in a worsening of behavior problems. In several cases, the care team refused to allow the research because their belief in the utility of the drug was so strong. In many cases, participants were allowed to take part in the study only after much effort was invested in assuring staff that such a trial was ethical and could benefit the residents. Early study discontinuations also seemed to be affected by staff fears that psychotropic discontinuation might be harmful to the residents. Furthermore, the care team was extremely protective of the residents. All were reluctant to change the status quo, fearing it could interfere with the resident's life.
A limitation of the study is its small sample size. Only 35 participants completed the trial. That number of subjects, however, is consonant with almost all the studies reviewed by Schneider et al30 in their meta-analysis of controlled trials of neuroleptic treatment in dementia. Furthermore, our rate of recruitment is somewhat higher than the 11% (of 195 members of a potential subject pool) reported by Treves et al31 who agreed to take part in a study of recruitment rates for drug trials for dementia of the Alzheimer type. This result underscores the difficulty of conducting research on a population with high mortality, comorbidity, and frailty, especially when the cooperation of multiple caregivers is essential. On the other hand, that last requirement also facilitated an in-depth investigation, with input from multiple raters. Overall, the findings from all caregivers converged to produce a consistent picture of the impact of the trial.
Finally, the study demonstrates that long-standing behavior problems are a challenge that we do not yet know how to face. Most residents whose drug therapy was discontinued at the physician's initiation were administered other psychotropic medications. Those for whom treatment with current medication was discontinued during the study continued to manifest about the same levels of agitated behavior, with or without the medication. Haloperidol, thioridazine, and lorazepam failed to influence the very behaviors they were administered to treat. Research demonstrating effective nonpharmacologic treatments and pharmacologic agents for the treatment of agitation is still sorely needed. Duration of effectiveness needs to be an integral part of this research. It is possible that new generations of psychotropic medication might provide a more effective means of controlling agitation among the elderly than the drugs considered in this study. The need to explore that possibility is underscored by the outcome of the present findings.
Corresponding author: Jiska Cohen-Mansfield, PhD, Research Institute of the Hebrew Home of Greater Washington, 6121 Montrose Rd, Rockville, MD 20852 (e-mail: email@example.com).
Accepted for publication November 16, 1998.
This study was supported by grants AG00547 and AG10172 from the National Institute on Aging, Bethesda, Md.