11Prevention of delirium: pharmacological

Publication Details

11.1. Clinical introduction

The serious nature of delirium and its consequences makes all methods of prevention important to establish. Pharmacological agents are a recognised cause of delirium and so the use of these agents for prevention needs to be approached cautiously. Antipsychotic, benzodiazepines, acetylcholinesterase inhibitor classes of drugs in particular, and products that influence the immune system, may prove useful, based on early evidence from small studies, or from a theoretical perspective.

People at risk of delirium are already vulnerable to the adverse effects of pharmacological products. It will be essential to establish the efficacy and risks of preventative drug treatment from well conducted clinical trials before they might be considered for routine use in clinical practice.

11 A. Prevention in hospital

CLINICAL QUESTION

What are the most clinical and cost effective and safe pharmacological interventions for the prevention of delirium in people in hospital?

11.2. Description of studies

Details of included and excluded papers together with study design are reported in table 11.1.

Table 11.1. study inclusion, exclusion and design.

Table 11.1

study inclusion, exclusion and design.

Two Cochrane Reviews were identified (Lonergan 2007; Siddiqi 2007) and updated. The Lonergan (2007) review examined the effectiveness of cholinesterase inhibitors in one study (Liptzin 2005) and the Siddiqi (2007) review examined both pharmacological (Aizawa 2002; Berggren 1987; Diaz 2001; Kalisvaart 2005; Liptzin 2005) and non-pharmacological (Marcantonio 2001) interventions for the prevention of delirium. Studies which did not meet our search criteria (Berggren 1987) or examined interventions not licensed for use in the UK (Diaz 2001) were not inlcuded. One study reporting non pharmacological intervention (Marcantonio 2001) has been reported in Chapter 10B (multicomponent prevention). This evidence review also includes outcomes not reported within the Cochrane reviews and has been updated to include papers published up to 2009.

11.2.1. Study Design

None of the studies were conducted in the UK. Information on study size, geographical location and funding are described in table 11.2.

Table 11.2. study characteristics.

Table 11.2

study characteristics.

Study duration was reported in four studies (Aizawa 2002: 7 days; Gamberini 2009: 6 days postoperatively; Kalisvaart 2005: varied to a maximum of six days depending on the onset of delirium; Liptzin 2005: 28 days).

11.2.2. Population

The age range across the studies was 51 years to 90 years. All studies included men and women. The patients’ ethnicity was described as being 95% white and 5% other in one study (Liptzin 2005) and was not reported in the remaining studies.

All of the studies were conducted in hospital settings in patients undergoing surgery. The type of surgery included resection for gastric or colorectal cancer (Aizawa 2002); hip surgery for acute fractures or hip replacements (Kalisvaart 2005); gastrointestinal surgery (Kaneko 1999); total joint replacement surgery of the knee or hip (Liptizin 2005); cardiac surgery with cardiopulmonary bypass (Prakanrattana 2007), cardiac surgery (Gamberini 2009). The Kaneko (1999) study reported that all patients were admitted to an ICU before the scheduled surgery.

Cognitive status was not reported in two studies (Aizawa 2002; Prakanrattana 2007), one study (Liptzin 2005) reported that at baseline patients did not have dementia, and one study (Gamberini 2009) reported that patients with an MMSE score of less than 15 were excluded. Three studies reported that the method used to assess dementia was the Mini Mental State Examination (MMSE) (Gamberini 2009; Kalisvaart 2005; Liptzin 2005). The reported MMSE scores indicated that at least some of the patients had dementia. One study did not report the method used for the assessment of dementia (Kaneko (1999).

One study reported the risk of postoperative delirium (Kalisvaart 2005). In this study, 84% of the patients had an intermediate risk for postoperative delirium and 16% had a high risk for postoperative delirium (as based on four predictive risk factors not specifically described); low risk patients were excluded. Patients with delirium at hospital admission were excluded from the study.

The Kalisvaart (2005) study also described their patients as having light dehydration.

11.2.3. Interventions

Acetylcholinesterase

One study (Liptzin 2005) investigated the acetylcholinesterase inhibitor, donepezil.

One study (Gamberini 2009) investigated the acetylcholinesterase inhibitor, rivastigmine

  • 1.5 mg oral rivastigmine three times per day every 8 hours, starting on the evening preceding surgery and continuing until the sixth postoperative day; each patient received 22 doses in total.
11.2.3.1. Atypical antipsychotics

One study (Prakanrattana 2007) investigated the atypical antipsychotic, risperidone.

  • 1 mg (orally disintegrating tablet) sublingually as a one-off dose when patients started to wake up in the ICU.
11.2.3.2. Typical antipsychotics

Two studies (Kalisvaart 2005; Kaneko 1999) investigated the typical antipsychotic drug haloperidol. The interventions included:

  • 0.5 mg haloperidol tablet three times per day, starting on hospital admission and continued until 3 days after surgery; a maximum delay from admission of 72 hours was permitted before surgery (Kalisvaart 2005)
  • 5 mg intravenous haloperidol once per day, starting on the first postoperative day (Kaneko 1999)
Benzodiazepines

One study (Aizawa 2002) investigated the use of a ‘Delirium Free Protocol (DFP)’ which was designed to address the risk factor of insomnia. The DFP included:

  • a combination of two benzodiazepines with pethidine: (diazepam 0.1 mg/kg per day intramuscularly given at 20.00h and a drip infusion of flunitrazepam 0.04 mg/kg) and pethidine 1 mg/kg (both given from 20.00 to 04.00h), for the first 3 days postoperatively, starting on the day of the operation.
  • The GDG expressed concern that the method of delivery of the drug (IM diazepam), and the addition of pethidine made the effect of benzodiazepines unclear, the study was addressing symptoms of improving insomnia, which in turn is a risk factor for delirium; this study was therefore not considered further.

11.2.4. Comparisons

The following comparisons were carried out:

Acetylcholinesterase inhibitors
  • Donepezil versus placebo (Liptzin 2005)
    • The intervention was given for 14 days preoperatively and a further 14 days postoperatively; patients were not admitted to hospital until the day before surgery.
    • The control group received placebo once a day at breakfast, and again, where symptoms of delirium were experienced, the placebo dose was doubled.
  • Rivastigmine versus placebo (Gamberini 2009)
    • The intervention was given the evening before surgery, three times per day every 8 hours thereafter until the evening of the sixth postoperative day.
    • The control group was administered the placebo (liquid identical to rivastigmine solution) following the same dosing scheme.
    • If postoperative delirium occurred, patients received haloperidol (starting with 0.5 mg every 6 to 8h) and lorazepam (1 mg per day)
11.2.4.1. Atypical antipsychotics
  • Risperidone (orally disintegrating tablet) versus placebo (an antiseptic strip applied sublingually). The interventions were a one-off dose. (Prakanrattana 2007)
11.2.4.2. Typical antipsychotics
  • Haloperidol versus placebo
    • 0.5 mg haloperidol tablet three times per day, up to 6 days pre and postoperatively (Kalisvaart 2005)
      • all patients received a proactive geriatric consultation (geriatric medical attention; enhancement of orientation and cognition; sensory and mobility improving advice; attention to pain and sleeping problems; extra attention to food and fluid intake; patient, family and nursing staff education). This study also gave the patients haloperidol and/or lorazepam 3 times a day if postoperative delirium occurred.
    • 5 mg intravenous haloperidol once per day, 5 day intervention period postoperatively (Kaneko 1999)

Concurrent medications were not reported in three studies (Liptzin 2005; Kalisvaart 2005; Kaneko 1999). Comorbidities were not reported in three studies (Kalisvaart 2005; Kaneko 1999; Liptzin 2005). One study (Prakanrattana 2007) reported that 67% of the patients were suffering from coexisting diseases including hypertension, diabetes mellitus, cerebrovascular accident, renal failure, or atrial fibrillation and another study (Gamberini 2009) reported that patients had arterial hypertension (78%) and were being treated for diabetes mellitus (7%) and for chronic pulmonary obstructive disease (4%).

11.3. Methodological quality

The Liptzin (2005) study reported that initially 1038 patients were contacted and 732 were not followed up or refused to participate. The remaining 306 were contacted 2–3 weeks before surgery and underwent screening. From these, 90 patients were randomised, although 10 were not operated on and the results are based upon 80 patients. The study reported there were no significant differences between the randomized patients and the non participants, in relation to age, gender, ethnicity, and site of operation (knee or hip joint surgery).

The method of sequence generation was adequate in three studies (computer generated blocks of 20: Gamberini 2009; computer-generated sequence: Kalisvaart 2005; Prakanrattana 2007). Sequence generation was not reported in two studies (Kaneko 1999; Liptzin 2005).

Allocation concealment was partially met in all of the studies. Gamberini (2009) reported that optically identical solutions in identical bottles were delivered by the hospital pharmacy, labelled with a number. Kalisvaart (2005) used identical containers prepackaged by a hospital pharmacist, which were sequentially assigned; Kaneko (1999) used sealed envelopes. In the Liptizin (2005) study the patients were randomised by the research pharmacist, but no further details were given, and in the Prakanrattana (2007) study, a concealed envelope was used.

Four studies (Gamberini 2009; Kalisvaart 2005; Liptzin 2005; Prakanrattana 2007) were described as double-blind (Kalisvaart 2005: blinding was checked by interviewing the study assessors). Although in the Prakanrattana (2007) study the patients’ placebo was an antiseptic strip rather than tablet, the authors stated that the assessors were blind to treatment. The Kaneko (1999) study did not report on blinding, although a placebo was used.

An a priori sample size calculation was reported in three studies (Kalisvaart 2005; Liptizin 2005; Prakanrattana 2007). The Gamberini (2009) study reported that a sample size of 120 was required to detect a relative risk reduction of 50%, with 80% power at a 5% significance level. One study (Kalisvaart 2005) reported a sample size of 206 patients per group was required to detect a 13% decrease in risk with 80% power at a 5% significance level. The sample sizes included in this study (n= 430), slightly exceeded this sample size estimate. The Liptzin (2005) study reported that a sample of 80 was required to have an 80% power to detect a difference of 22% in the study groups at a one-sided significance level of 5% assuming a delirium rate of 44% in the placebo group. Another study (Prakanrattana 2007) required a sample size of 63 per group to detect a 30% reduction in risk with 90% power at a 5% significance level; 63 patients per group were recruited and completed the study.

All studies demonstrated baseline comparability.

The Kalisvaart (2005) study reported no significant differences in mean age, proportion of males to females, Mini-mental state examination scores, visual acuity, health scores, geriatric depression scores, Barthel Index, or baseline risk of delirium between treatment and control groups. The Kaneko (1999) study reported no differences in the proportion of males to females by group, pre-existing diseases, preoperative medicines, duration of operation and anesthesia. They did observe that fewer patients in the haloperidol group had premorbid cognitive impairment (5% versus 10% in the placebo group), but the difference was not statistically significant. In the Liptzin (2005) study patients were comparable at baseline for age, gender, ethnicity, the surgeon who operated, the joint operated on and the MMSE questionnaire and clock-drawing test scores. The Prakanrattana (2007), study demonstrated baseline comparability between intervention groups for age, proportion of males to females, weight, New York Heart Association functional class, coexisting disease, type of operation (coronary artery bypass graft, valve or others), anaesthesia time, cardiopulmonary bypass time, and aortic cross-clamp time. In the Gamberini (2009) study patients were comparable for age, gender, baseline MMSE, baseline clock-drawing test scores, pre-existing diseases, type of operation (CABG, valve repair).

One study (Prakanrattana 2007) reported no missing participants; all patients were included in the analysis.

Three studies (Gamberini 2009; Kalisvaart 2005; Kaneko 1999) reported acceptable missing levels of data (that is less than 20%).

  • The Gamberini (2009) study reported there was missing data for 25% (15/61) and 24% (14/59), in the intervention and control groups respectively. The study reported that only patients who were not assessed with CAM within 6 days after surgery (4/61: 3/59) were excluded from the analysis; however, the authors reported that an intention to treat analysis was carried out.
  • In the Kaneko (1999) study 5% (2/40) in the intervention group and 0% in control group were missing, and the authors analysed all available participants in their analyses (n = 78).
  • In the Kalisvaart (2005) study, 5% (11/212) were lost to follow-up in the treatment group and 11% (24/218) were lost to follow-up in the placebo group. However the authors analysed all patients who were randomised (ITT analysis).

One study (Liptzin 2005) had inadequate levels of missing data (more than 20% missing data in each group). Originally 90 patients were included in the study, but ten patients were not included in the final analyses because they were not operated on, or took no further part in the analysis; the groups to which they were assigned were not reported. Of the remaining 80 patients, a further 11/39 (28%) and 11/41 (27%) did not complete the study. A per protocol analysis was reported based on the 80 patients, although it was not clear what was assumed about the missing data.

Methods to assess concordance were partially reported in Kalisvaart (2005). They stated that clinical staff recorded the level of adherence to the intervention, but it was not stated how this was done. Concordance was determined by patients keeping records of their medication usage, and this was assessed by a research assistant (Liptzin 2005). Methods to assess concordance were not reported in the remaining studies.

The method of delirium assessment was:

Method of delirium assessment was unclear in one study (Kaneko 1999). The DSM-IV and DSM III-R criteria were used for ‘psychotic diagnoses’ and also stated that delirium was ‘clinically diagnosed’. Data were collected from the patients and nursing charts on the fifth day after surgery; it was not clear if the charts were used to record delirium.

One study (Kalisvaart 2005) assessed severity using the DRS-R-98 [range 0 (no severity) to high 45 (high severity)], MMSE, and the Digit Span test [assessment of attention, range 0 (no attention) to 42 (good attention)].

All studies evaluated the incidence of delirium as a primary outcome. Secondary outcomes were: severity of delirium (Kalisvaart 2005), duration of delirium (Gamberini 2009; Kalisvaart 2005; Kaneko 1999; Liptzin 2005) and adverse events (Kalisvaart 2005; Kaneko 1999), length of hospital stay (Gamberini 2009; Kalisvaart 2005; Liptzin 2005; Prakanrattana 2007), length of ICU stay (Gamberini 2009; Prakanrattana 2007), and sleep-wakefulness rhythm (Kaneko 1999).

Overall two studies were considered to have a higher risk of bias for the following reasons:

The use of rescue medication in the Kalisvaart (2005) study may have led to confounding for the following outcomes: duration of delirium, severity of delirium and length of stay.

11.4. Results

11.4.1. Acetylcholinesterase inhibitor versus placebo

1. Incidence of postoperative delirium (endpoint 28 days)

Meta-analysis of two studies (Gamberini 2009; Liptzin 2005) with 193 patients, comparing acetylcholinesterase (ACH) with placebo showed_no significant difference in the incidence of delirium between the groups (RR 1.11 (95% CI 0.69 to 1.79)); although the results are very imprecise (figure 11.1, Appendix K);

2. Duration of postoperative delirium

Two studies (Gamberini 2009; Liptzin 2005) reported the duration of postoperative delirium.

The Gamberini (2009) study compared rivastigmine versus placebo, in 113 patients and reported there was no difference in the duration of delirium. The results from this study are not shown on the forest plot because study reported values for the median and range. The reported median and range were as follows: 2.5 days (range 1 to 5) and 2 days (range 1 to 6) for the rivastigmine and placebo groups respectively (reported p value= 0.3).

The remaining study (Liptzin 2005) comparing donepezil with placebo in 80 patients found no significant difference in the duration of postoperative delirium (end point) (figure 11.2, Appendix K); mean difference (MD) −0.30 days (95%CI −0.67 to 0.07), for a placebo group duration of 1.3 days; the results are imprecise. The standard deviation in the donepezil group was stated to be zero, but for the purposes of analysis this was assumed to be 0.001.

3. Length of hospital stay

Two studies (Gamberini 2009; Liptizin 2009) reported the length of stay. The Gamberini (2009) study reported the median and range and the results for this study are not shown on the forest plot. The (Gamberini 2009) study comparing rivastigmine versus placebo in 113 patients reported there was no difference in the length of hospital stay; the median and range was 13 days (range 7 to 39) for both the rivastigmine and placebo groups respectively (reported p value = 0.3).

One study (Liptzin 2005) comparing donepezil with placebo in 80 patients found no significant difference in the length of hospital stay (endpoint 28 days) between the groups (figure 11.3, Appendix K); MD 0.20 days (95%CI –0.10 to 0.50). There was imprecision because of the small sample size.

4. Length of ICU stay

One study (Gamberini 2009) comparing rivastigmine versus placebo in 113 patients reported there was no difference in the length of ICU stay; the median and range were as follows: 2 days (range 2 to 7) and 2 days (range 2 to 6) for the rivastigmine and placebo groups respectively (reported p value: 0.9). This outcome is not included in the GRADE evidence summary.

5. Number of patients discharged to a rehabilitation facility (endpoint 28 days)

Analysis of one study comparing donepezil with placebo in 80 patients found no significant difference between the groups for the number of patients discharged to ‘a rehabilitation facility’, but it was not clear what this facility was (figure 11.4, Appendix K); RR 0.87 (95%cl 0.68 to 1.10). There was some imprecision in this outcome.

6. Use of rescue medications

The Gamberini (2009) study reported the use of haloperidol and lorazepam rescue medications. 32%: and 30% of the patients receiving rivastigmine and placebo respectively were given haloperidol (p=0.9). 61% and 68%, of the patients receiving rivastigmine and placebo, respectively were given lorazepam; p=0.3). There were no significant differences between the two groups in the number of patients who received the rescue medications.

Please refer to table 11.3 for the GRADE evidence summary for cholinesterase inhibitors.

Table 11.3. GRADE evidence summary - cholinesterase inhibitors vs placebo.

Table 11.3

GRADE evidence summary - cholinesterase inhibitors vs placebo.

Typical antipsychotics

Typical antipsychotics versus placebo
1. Incidence of postoperative delirium

Two studies (Kaalisvaart 2005; Kaneko 1999) reported the use of haloperidol versus placebo on incidence of postoperative delirium. The Kaalisvart (2005) study reported that all patients received a proactive geriatric consultation, thus the study was investigating the adjunctive effect of haloperidol. Therefore, these two studies are reported separately on the forest plots (figure 11.5, Appendix K)

  • One study (Kalisvaart 2005) with 440 patients showed no significant difference in the incidence of postoperative delirium; RR 0.91 (95% CI 0.59 to 1.42).
  • The Kaneko (1999) study with 78 patients showed a small significant effect [0.32 (95% CI 0.12 to 0.91)]. We note this study was at higher risk of bias.
2. Severity of delirium

Two studies (Kalisvaart 2005; Kankeo 1999) evaluated the severity of delirium, and only Kalisvaart (2005) presented data for analysis. In 78 patients who had delirium, Kalisvaart (2005) used the highest value obtained during delirium, on the DRS-R-98 scale, (maximum value on this scale is 39) to assess the severity of delirium. The analysis demonstrates a significant effect in favour of haloperidol: MD −4.01 (95% CI −5.87 to −2.15; figure 11.6, Appendix K). It is noted that the severity of delirium may have been confounded by the use of rescue medication.

The Kaneko (1999) study reported that the postoperative delirium was more severe in the placebo group (no data or statistical analyses were presented).

3. Duration of delirium

Two studies (Kalisvaart 2005; Kaneko 1999) evaluated the duration of delirium, and only Kalisvaart (2005) presented data for analysis. The analysis demonstrates that patients who received haloperidol, had, on average, significantly fewer days of delirium (of those who had delirium): MD −6.40 (95% CI −9.38 to −3.42; figure 11.7, Appendix K). It is noted that the duration of delirium may have been confounded by the use of rescue medication and that results were reported only for those with delirium. We also note that the distribution for the duration of delirium is skewed for both the intervention and placebo groups (mean values less than twice the standard deviation). The Kaneko (1999) study reported that the duration of postoperative delirium was longer in the placebo group (no data or statistical analyses were presented).

4. Length of hospital stay

The Kalisvaart (2005) study demonstrated that the number of days spent in hospital was significantly shorter in patients who received haloperidol compared to patients who received placebo in addition to the proactive geriatric consultation; MD −5.50 (−8.17 to −2.83; figure 11.8, Appendix K). The study included the results for hospital length of stay in a table that was stated to apply to patients with delirium only. However, we have assumed this should refer to all patients; we also note that the summary statistics are incorrectly noted in the table in the report (the upper confidence limit is lower than the mean). Furthermore, the distribution for length of stay is skewed for both intervention and placebo groups.

5. Adverse events

Two studies (Kalisvaart 2005; Kaneko 1999) evaluated adverse events. Kalisvaart (2005) reported that there were no drug–related side effects and no sedation events were reported, other than those related to morphinomimetics. Only Kaneko (1999) presented data for analyses; they observed that one patient in the treatment group developed transient tachycardia. The results are very imprecise (figure 11.9, Appendix K).

Please refer to table 11.4 for the GRADE evidence summary for typical antipsychotics.

Table 11.4. GRADE evidence summary: Typical antipsychotics vs placebo.

Table 11.4

GRADE evidence summary: Typical antipsychotics vs placebo.

11.4.2. Atypical antipsychotics

Atypical antipsychotics versus placebo
1. Incidence of delirium

In the Prakanratta (2007) study, delirium was recorded twice daily in the ICU and once daily on discharge from the ICU. The study reported results as percentages, so we calculated the number of patients with delirium.

In one study (Prakanrattana 2007) comparing risperidone with placebo in 126 patients, there were significantly fewer patients with delirium in the risperidone group compared with placebo, although the result was imprecise (figure 11.10, Appendix K); RR 0.35 (95%CI 0.16 to 0.77) which corresponds to a number needed to treat of 5 (95%CI 3 to 14), for a control group rate of 32%. The authors reported that all episodes of delirium occurred within the first three postoperative days.

2. Length of ICU stay

There was no significant difference between the treatment groups for the number of days spent in ICU; MD 0.10 (95% CI −0.64 to 0.84; figure 11.11, Appendix K). The results are very imprecise (clinically important difference: 0.5 days).

3. Length of hospital stay

There was no significant difference between the treatment groups for the number of days spent in hospital; MD 0.20 (95% CI −1.74 to 2.14; figure 11.12, Appendix K). The results are very imprecise.

Please refer to table 11.5 for the GRADE evidence summary for atypical antipsychotics.

Table 11.5. GRADE evidence summary - Atypical antipsychotics vs placebo.

Table 11.5

GRADE evidence summary - Atypical antipsychotics vs placebo.

Overall summary of results

Overall summary of results for pharmacological prevention of delirium in hospital setting are reported in table 11.6.

Table 11.6. summary of results: pharmacological prevention of delirium in hospital setting.

Table 11.6

summary of results: pharmacological prevention of delirium in hospital setting.

11.5. Health economic evidence

11.5.1. Pharmacological interventions for the prevention of delirium in a hospital setting

One economic evaluation study was included as evidence (Bracco 2007). This was a non-randomised clinical trial of 1293 patients who underwent cardiac surgery in Canada. The objective was to examine outcomes and use of intensive care resources for a cohort of consecutive patients who underwent cardiac surgery with or without thoracic epidural anaesthesia. The intervention group received thoracic epidural anaesthesia for cardiac surgery. The control group did not receive thoracic epidural anaesthesia. Detailed description of intervention and control strategies is given in Appendix J (table J1). The intervention shortened ventilation time and the length of stay in the ICU by 9.6 hours and 12.7 hours respectively after adjusting for type of surgery in a multivariate analysis. This reduction decreased the ICU and mechanical ventilation costs by US$2700 and US$700 respectively, per patient. The additional cost of thoracic epidural use was given as US$82. Post-operative delirium complication rate was reported as 24/506 in the intervention arm, and 20/787 in the control arm. This was measured using CAM-ICU scale. A relative risk of 0.3 was reported. Intensive care unit mortality rate of 2/506 was also reported in the intervention arm and 14/787, in the control arm. A multivariate analysis for mortality was not statistically significant. Cost data was taken from the literature and QALY estimates were not reported. The study sample was not randomised and there was no sensitivity analysis on variables whose values will probably be uncertain. The results are not directly applicable and should be cautiously interpreted.

11.6. Clinical evidence statements

11.6.1. Acetylcholinesterase inhibitor versus placebo

  • Meta-analysis of 2 RCTs comparing acetylcholinesterase with placebo showed:
    • no significant effect on the incidence of delirium (very low quality).
11.6.1.1. Donepezil versus placebo
  • 1 RCT comparing donepezil with placebo showed:
    • no significant effect on the length of hospital stay and the number of patients discharged to a rehabilitation facility (low quality).

11.6.2. Typical antipsychotics

Haloperidol versus placebo
  • 1 RCT comparing haloperidol with placebo as an adjunct to a proactive geriatric consultation (non-pharmacological intervention) showed:
    • no significant effect on the incidence of postoperative delirium (low quality).
    • a significantly lower severity of delirium and fewer days of delirium in favour of the haloperidol group ( low quality)
    • a significantly shorter length of hospital stay in patients who received haloperidol (low quality).
  • 1 RCT comparing haloperidol with placebo showed:
    • no significant effect on the incidence of postoperative delirium (low quality)
    • no difference between the groups for the number of adverse events (transient tachycardia); (insufficient evidence).
11.6.2.1. Atypical antipsychotics versus placebo
  • 1 RCT conducted in ICU, comparing risperidone with placebo showed:
    • a lower incidence of delirium in patients receiving risperidone (moderate quality).
  • 1 RCT comparing risperidone with placebo showed: o no significant difference between the groups for length of stay in ICU and hospital (low quality).
    • no significant difference between the groups for length of stay in ICU and hospital (low quality).

11.7. From evidence to recommendations

The GDG discussed the evidence from the pharmacological prevention reviews and noted that it was limited and of low quality. The evidence was mainly from single studies and each of these had risk of bias issues; in addition, the evidence was often imprecise, sometimes indirect and showed inconsistency where there was more than one study.

  • Donepezil: the study was unrepresentative of the population (patients were fit and healthy with no cognitive impairment)
  • Risperidone: the study was unrepresentative of the intervention or the population [the dose used was very different from that used in clinical practice, and the study included a relatively young population (age range: 51 to 71 years) undergoing cardiac surgery].
  • Haloperidol: two studies investigated haloperidol. One study had a high risk of bias and the other assessed haloperidol as an adjunct to a proactive geriatric consultation intervention. There was explained inconsistency between the studies.

The GDG was not confident in the evidence and did not make a recommendation but agreed a research recommendation for typical antipsychotics, atypical antipsychotics, acetylcholinesterase inhibitors and benzodiazepines (and Appendix H). For ethical reasons, research should only be carried out in a population at high risk of delirium.

Future research recommendation

In hospital patients at high risk of delirium, which medication (atypical antipsychotics, typical antipsychotics, benzodiazepines, or acetycholinesterase) compared with placebo or each other is more clinically and cost effective, in preventing the development of delirium?

11.8. Recommendations

There are no recommendations for this section. In light of the evidence the GDG did not wish to make recommendations.

11 B. Prevention in long-term care

CLINICAL QUESTION

What are the most clinical and cost effective and safe pharmacological interventions for the prevention of delirium in people in long-term care?

11.9. Description of studies

One paper was evaluated for inclusion Moretti (2004). The study was an RCT.

11.9.1. Study Design

The RCT was conducted in Italy in a community based setting; this was treated as an indirect setting for long-term care. Patients without reliable carers were excluded from the trial. The funding source was not reported. Two hundred and forty six patients were randomised; the unit of randomisation was the patient.

11.9.2. Population

The patients all had an MMSE score of at least 14, indicating patients had mild to moderate dementia. All patients met the DSM-IV criteria for dementia. Patients also satisfied the criteria for probable vascular dementia, or multi-infarct dementia with the NINDS-AIREN criteria (National Institute of Neurological Disorders and Stroke and Association Internationale pour la Recherché et l’Enseignement en Neurosciences). Their ages ranged from 65–80 years with a mean age of 76 years. One hundred and sixteen men and 130 women were included in the study, although 12 patients died during the study and four refused to participate; all data were based on the remaining groups of 115 in the rivastigmine group and 115 in the aspirin group. All were ambulatory outpatients living in the community. Their delirium risk was not stated in the study. The comorbidity was vascular dementia, although other comorbidities were implied because of the drugs patients were taking; patients with previous psychiatric illness or central nervous system disorders or alcoholism were excluded.

11.9.3. Interventions

The included study investigated rivastigmine, a cholinesterase inhibitor, compared with cardio-aspirin (considered as usual care). Participants were ambulatory outpatients and were given the interventions for 2 years after randomisation. Rivastigmine was titrated to the higher dose after the first 16 weeks. The interventions included:

It was assumed that the cardio-aspirin was representing usual care and was not an active intervention.

11.9.4. Comparisons

The following comparison was carried out:

The patients were allowed to continue taking their existing drug therapies, anti-hypertensives, anti-dyslipidemic, anti-diabetic drugs, diuretics and bronchodilators.

Patients received benzodiazepines or neuroleptic drugs during delirium, which were significantly less in the intervention group. This may have led to confounding for some outcomes, but would serve to underestimate the size of the effect.

11.10. Methodological quality

The methods of sequence generation and allocation concealment were not described, although the patients were matched for age and education level. It was not reported if all eligible patients were recruited.

The study did not report whether patients and investigators were blinded to treatment allocation. An a priori sample size calculation was not reported.

Originally 246 patients were included in the study, but 16 were not included in the final analyses (7% missing data; 12 patients died during the follow up and four refused to participate in the follow up). The groups to which they were assigned were not reported. The remaining 230 patients completed the two year follow up. Patients were found to be comparable at baseline on the following scales: BEHAVE-AD (Behavioural Pathology in Alzheimer’s Disease Rating); Clinical Dementia Rating; and the Cumulative Illness Rating Scale. Concordance was monitored by care givers, who controlled the intake of drugs.

Delirium was assessed using the Confusion Assessment Method (CAM).

Overall, the study may have been at a higher risk of bias because allocation concealment and blinding were unclear; appear to have a higher potential for bias, although the differential use of rescue medication may have led to confounding for some outcomes.

11.11. Results

11.11.1. Rivastigmine versus usual care (aspirin)

Incidence of delirium (endpoint 2 years)

Analysis of one study in 230 patients showed that the incidence of delirium was significantly lower in the rivastigmine group compared with usual care; RR 0.65 (95%CI 0.50 to 0.85), which corresponds to a number needed to treat of 5 (95%CI 4 to 12), for a control group rate of 62%. The result was imprecise (figure 11.13, Appendix K).

Duration of delirium

Analysis of one study in 230 patients showed that the duration of delirium was significantly shorter in the rivastigmine group compared with usual care (figure 11.14a, Appendix K); MD −3.86 days (95%CI −4.44 to −3.28), for a control group duration of 7.86 days. It was unclear whether the duration of delirium was reported just for those who had delirium or was a mean across all patients: the paper describes ‘the main duration of the delirium’. In addition, the different standard deviations across the groups, indicates the mean may just be for those with delirium. Figure 11.14b (Appendix K) shows the analysis with this assumption; the only difference is a slightly wider CI; MD −3.86 days (95%CI −4.66 to −3.06).

Cognitive impairment

The study assessed global performance using the Clinical Dementia Rating (scale 0–3), and reported the change from baseline at 12 months. Analysis of 230 patients showed there was no significant difference between the groups, although the table in Moretti (2004) stated the difference was significant (figure 11.15, Appendix K);

Behavioural disturbance (change score at 1 year)

Analysis of one study in 230 patients showed that behavioural disturbance was significantly lower in the rivastigmine group compared with usual care (figure 11.16a, Appendix K). The study used the BEHAVE-AD to assess individual behavioural items on this scale (delusions, hallucinations, activity alterations, aggressiveness, anxiety/phobia, sleep disturbances, affective disturbancesand anxiety). All individual items were stated to be statistically significant, with the exception of delusions. The overall score showed a statistically significant mean difference, favouring the intervention; MD −39.66 (95%CI −40.06 to −39.26). This seems to be a very narrow CI, even for a change score from baseline, but if these were standard errors, rather than standard deviations (despite what was reported in the text), the standard deviations would be rather large for the intervention group (figure 11.16b, Appendix K). The assumption of a standard error gave a large significant mean difference of −39.66 (95% CI −43.91 to −35.41), favouring the intervention group.

Please refer to table 11.7 for the GRADE evidence summary for cholinesterase inhibitors.

Table 11.7. GRADE evidence summary:Cholinesterase inhibitors vs usual care.

Table 11.7

GRADE evidence summary:Cholinesterase inhibitors vs usual care.

Overall summary of results for pharmacological prevention of delirium in long term care setting are reported in table 11.8.

Table 11.8. summary of results: pharmacological prevention of delirium in long-term care setting.

Table 11.8

summary of results: pharmacological prevention of delirium in long-term care setting.

11.12. Health economic evidence

No relevant health economic papers were identified.

11.13. Clinical evidence statements

  • 1 RCT of very low quality comparing rivastigmine with usual care (indirect evidence) showed that the rivastigmine group had significantly:
    • lower incidence of delirium (endpoint 2 years).
    • fewer days of delirium.
    • lower behaviour disturbances (change score at 1 year).
    • However at 1 year there was no significant difference between the groups for change in cognitive impairment from baseline.

11.14. From evidence to recommendations

There was one very low quality study in an indirect population for the long term care setting (the community). The GDG were not confident in the evidence to make a recommendation on the basis of this study.

11.15. Recommendations

There are no recommendations for this section. In light of the evidence the GDG did not wish to make recommendations.

Image appkf109
Image appkf110
Image appkf111
Image appkf112
Image appkf113
Image appkf114
Image appkf15
Image appkf116
Image appkf117
Image appkf118
Image appkf119
Image appkf120
Image appkf121
Image appkf122
Image appkf123
Image appkf124
Image appkf125
Image appkf126