Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2014 Dec 17.
Published in final edited form as: Am J Psychiatry. 2011 Oct 31;169(1):71–79. doi: 10.1176/appi.ajp.2011.11030347

Risk of Mortality Among Individual Antipsychotics in Patients with Dementia

Helen C Kales 1,2, Hyungjin Myra Kim 1,3, Kara Zivin 1,2, Marcia Valenstein 1,2, Lisa S Seyfried 2, Claire Chiang 1,2, Francesca Cunningham 4, Lon S Schneider 5, Frederic C Blow 1,2
PMCID: PMC4269551  NIHMSID: NIHMS638045  PMID: 22193526

Abstract

Objective

The use of antipsychotics to treat the behavioral symptoms of dementia is associated with increased mortality. However, there remains limited information regarding individual agents’ risks.

Method

This was a retrospective cohort study using national data from the US Department of Veterans Affairs (fiscal years 1999–2008) for patients ≥65 years old with dementia, beginning outpatient treatment with an antipsychotic (risperidone, olanzapine, quetiapine, and haloperidol) or valproic acid and its derivatives (as a non-antipsychotic comparison). The total sample included 33,604 patients. Individual drug groups were compared for 180-day mortality rates. Potential confounding was addressed using multivariate models and propensity adjustments.

Results

In covariate-adjusted intent to treat analyses, haloperidol users had the highest mortality rates (relative risk 1.54, 95% confidence interval 1.38–1.73) followed by risperidone (reference), olanzapine (RR 0.99, 95% CI 0.89–1.10), valproic acid and its derivatives (RR 0.91, 95% CI 0.78–1.06) and quetiapine (RR 0.73, 95% CI 0.67–0.80). Propensity-stratified and propensity-weighted models as well as analyses controlling for site of care and medication dosage showed similar patterns. Haloperidol risk was highest in the first 30 days and then significantly and sharply decreased. Among the other agents, mortality risk differences were most significant in the first 120 days and declined in the subsequent 60 days during 180-day follow-up.

Conclusions

There may be differences in mortality risks among individual antipsychotic agents. Further, the use of valproic acid and its derivatives as alternative agents to address the neuropsychiatric symptoms of dementia may carry associated risks as well.

INTRODUCTION

The US Food and Drug Administration (FDA) has not approved any medication for treating the neuropsychiatric symptoms of dementia. However, atypical antipsychotics are commonly used off-label for treatment [1]. In April 2005, the FDA warned that use of atypical antipsychotics for behavioral disturbances in patients with dementia was associated with increased mortality. Subsequently, research reports confirmed the mortality risks associated with both conventional and atypical antipsychotics in dementia patients [2–5]. An FDA warning for conventional antipsychotics followed in June 2008 [6].

Information is limited about mortality with individual antipsychotic agents in patients with dementia. An earlier study [7] found no significant mortality differences between olanzapine and risperidone. However, the number of deaths during this trial was small with wide confidence intervals. In a meta- analysis [2], no increased risk of death was found with any individual atypical antipsychotic; however, there may have been inadequate power to detect significant differences after controlling for confounding variables between trials. A study [8] comparing the most frequently prescribed antipsychotic drugs in Canada found increased 180-day mortality ratios for haloperidol and loxapine, but no difference for olanzapine compared to risperidone. The most recent study, using case-control methodology, found that patients with dementia taking haloperidol, olanzapine, and risperidone, but not quetiapine, had short-term increases in mortality [9] as compared to patients with dementia not taking these agents.

Large-scale comparisons of mortality with individual antipsychotic agents controlling for important confounders are currently lacking. Using multivariate and propensity-scoring methods, this study examined mortality risks in outpatients with dementia in the 6 months following a new antipsychotic start of the individual agents most commonly used for US Department of Veterans Affairs (VA) Healthcare System patients with dementia (risperidone, olanzapine, quetiapine, and haloperidol). Based on evidence from our earlier work that anticonvulsants had similar mortality risks to antipsychotics [3], and that there was a small but significant increase in the use of valproic acid and its derivatives following the black box warning [10], these agents were also included for comparison.

METHOD

Study cohort

Data were provided by national VA registries maintained by the Serious Mental Illness Treatment, Resource, and Evaluation Center (SMITREC) in Ann Arbor, Michigan, USA. Patients included: 1) were ≥ 65 years old; 2) had a dementia diagnosis between October 1,1998–September 30, 2008 (ICD 9 diagnoses 290.0, 290.1x, 290.2x, 290.3, 290.4x, 291.2, 294.10, 294.11, 331.0, 331.1, and 331.82); and 3) began outpatient treatment with a study medication after a 12-month “clean period” without antipsychotic or anticonvulsant exposure. Over 87% of patients in the sample had monotherapy (e.g. exposure to only the initial agent during 6-month follow-up). Given that switching to other antipsychotic agents might obscure risk profiles for individual antipsychotics, we restricted the final sample to these monotherapy patients. The final study sample included 33,604 patients.

This study was approved by the VA Ann Arbor Healthcare System IRB.

Medications

These included risperidone, olanzapine, quetiapine, and haloperidol, as well as valproic acid and its derivatives (an anticonvulsant group commonly used as a second-line treatment strategy for the neuropsychiatric symptoms of dementia). Patients taking valproic acid and its derivatives (sodium valproate or divalproex) who also had seizure disorders (n=337) were excluded from the sample as their anticonvulsant use would be less likely to be related to dementia.

Mortality

Data were obtained from the US National Death Index (National Center for Health Statistics, Hyattsville, MD).

Other variables

These included age, gender, ethnicity, marital status and indicators of psychiatric and medical comorbidity {the latter using a modified version of the Charlson comorbidity index [11] based on 18 medical comorbidities (excluding dementia) in the year prior to new medication start}. Also, as delirium occurs frequently among patients with dementia and is an independent mortality risk factor [12], and antipsychotics are often prescribed for delirium, we also assessed for the presence of a delirium diagnosis at the time of prescription, using a coding scheme for acute confusional states developed for a prior study [13]; this included the following codes: 290.3, 291.0, 292.0, 292.1, 292.2, 292.9, 293.0, 293.1, 293.9, 294.8, 294.9, 348.3, 437.2, 572.2, 290.11, 290.41, 292.81, 293.31, 293.82, 293.83, 293.89,349.82. To control for potential changes in health care, particularly given the impact of the black box warning [10], calendar time at the new medication start was included as a covariate. The model also included the following variables: inpatient and nursing home days in the year prior to new medication start; and size, rurality and academic affiliation of the VAMC where the medication was prescribed.

Statistical Analysis

Descriptive statistics were used to characterize patient characteristics by type of medication prescribed. A 180-day follow-up period was chosen based on the duration of trials in the FDA’s analysis as well as the follow-up period used in prior studies [8, 14]. Analyses accounted for medication exposure days in two ways- “intent to treat” and exposure. For the intent to treat analyses, exposure-days were the length of time from the first filled prescription until death or 6 months, whichever was earlier. For the exposure analyses, exposure-days to a specific antipsychotic or valproic acid and its derivatives began on the date of the first fill; exposure was censored at the end of the exposure period, at 6 months, or at time of death, whichever was earlier. As in a prior study [2], the exposure period continued for the number of days’ supply of medication received plus 30 days. Any gaps in fills of less than 30 days were considered continued exposures. This accounts for some level of continued exposure and biological effect among patients who missed doses or used lower than prescribed doses.

For each of the medication types, mortality during 180-day follow-up was calculated as per 100 person years, and distribution of time to death since index prescription was estimated using the Kaplan-Meier survival analysis method.

We used a variety of approaches to deal with potential selection biases. Initially, we used multivariate analyses that included potential confounders available in administrative data. Additionally, we used propensity-weighted and propensity-stratified methods. Both methods attempt to control for “treatment by indication” in observational studies by adjusting for the predicted probability that a patient will receive a specific treatment conditional on the patient’s baseline covariate values. The propensity-weighted analyses estimated hazard ratios using Cox’s regression model with observations weighted inversely by the propensity estimates obtained using multinomial models, permitting comparisons across multiple medications based on the one model [15]. For the propensity-stratified analyses, comparisons were made between pairs of medications, with each medication compared against risperidone. For each pair-wise comparison, propensity scores were estimated using logistic regression, and hazard ratio estimates were obtained using Cox’s regression model, stratified by the estimated propensity quintiles. In both propensity-weighted and propensity-stratified methods, models used to obtain propensity scores were optimally fit to be highly predictable without consideration for parsimony.

Secondary analyses included site of care examination (psychiatric vs. non-psychiatric) and adjustment for antipsychotic dose which was standardized to haloperidol equivalent dose [16]. After a visual inspection of the smoothed hazards revealed decreasing hazards in time for haloperidol, we also extended the Cox regression model to test for non-proportional hazards using logarithmically transformed time by medication indicator interaction terms. Upon finding significantly decreasing risks in time for haloperidol, we divided time since medication start into 30-day intervals and used a piece- wise exponential model to compare relative risks between medications at different time intervals.

To confirm that our conclusion was not biased by the inclusion of only monotherapy patients, we also did a true intent to treat analysis where patients who switched or augmented their initial medication were also included in the analysis and were analyzed as exposed to their initial medication.

Lastly, we did two additional analyses to further examine mortality risk differences: 1) an exploration of whether the relatively larger proportion of Parkinson’s disease (PD) patients in the quetiapine cohort may have resulted in a lower mortality risk; and 2) an analysis where we sought to further confirm haloperidol’s role as the agent with the highest mortality by comparing haloperidol and risperidone users after individually matching each haloperidol patient with up to two risperidone patients on a number of key variables.

RESULTS

Characteristics of the Study Population

Table 1 shows demographic and clinical characteristics of the individual medication groups. Haloperidol users were significantly older and sicker (as evidenced by the highest Charlson scores, highest rates of concurrent delirium, and having more inpatient days in the prior year) than users of the other study medications. A higher percentage of African-American patients used haloperidol as compared to the other agents. Those taking haloperidol also were significantly more likely to have used opiods or benzodiazepines and less likely to have used antidepressants during the year prior to the new antipsychotic start. Users of the various atypical agents had similar rates of medical and psychiatric comorbidities with the exception of significantly higher rates of Parkinson’s disease in users of quetiapine. Users of valproic acid and derivatives tended to be younger, less likely to be African American, more likely to have comorbid bipolar disorder and other psychiatric illnesses than users of other agents.

Table 1.

Characteristics of Patients with Dementia who Started One of Five Psychotropic Medications

Variable Names Haldol
N=2855		 Olanzapine
N=4716		 Quetiapine
N=10651		 Risperidone
N=13356		 Valproic acid and derivatives
N=2026		 p-value
N % N % N % N % N %
Age
 65–69 138 4.8 288 6.1 559 5.2 755 5.7 192 9.5 <0.0001
 70–74 368 12.9 648 13.7 1592 14.9 1827 13.7 303 15.0
 75–79 749 26.2 1303 27.6 2885 27.1 3756 28.1 546 27.0
 80–84 949 33.2 1571 33.3 3484 32.7 4342 32.5 601 29.7
 85+ 651 22.8 906 19.2 2131 20.0 2676 20.0 384 19.0
Female 66 2.3 134 2.8 222 2.1 370 2.8 37 1.8 0.0012
Race:
 White 1950 68.3 3232 68.5 7565 71.0 9024 67.6 1470 72.6 <0.0001
 Black 431 15.1 397 8.4 974 9.1 1595 11.9 142 7.0
 Other 32 1.1 49 1.0 128 1.2 163 1.2 29 1.4
 Unknown 442 15.5 1038 22.0 1984 18.6 2574 19.3 385 19.0
Married 1964 68.8 3170 67.2 7726 72.5 9130 68.4 1426 70.4 <0.0001
Benzodiazepine use 630 22.1 883 18.7 1971 18.5 2307 17.3 374 18.5 <0.0001
Antidepressant use 1184 41.5 2603 55.2 5708 53.6 6618 49.6 1161 57.3 <0.0001
Opioid use 889 31.1 1190 25.2 3119 29.3 3702 27.7 554 27.3 <0.0001
Delirium 1245 43.6 1735 36.8 4496 42.2 5408 40.5 870 42.9 <0.0001
Depression 578 20.2 1399 29.7 3199 30.0 3638 27.2 697 34.4 <0.0001
Schizophrenia spectrum* 55 1.9 145 3.1 179 1.7 307 2.3 17 0.8 <0.0001
Bipolar 1 10 0.4 66 1.4 68 0.6 102 0.8 100 4.9 <0.0001
Bipolar 2 0 0.0 18 0.4 19 0.2 38 0.3 39 1.9 <0.0001
Other psychoses 648 22.7 955 20.3 2372 22.3 3036 22.7 330 16.3
Parkinsons Disease 132 4.6 339 7.2 1775 16.7 591 4.4 128 6.3 <0.0001
Psych IllnessesΨ
 0 2467 86.4 3894 82.6 8804 82.7 11184 83.7 1606 79.3 <0.0001
 1 345 12.1 712 15.1 1621 15.2 1870 14.0 330 16.3
 ≥2 43 1.5 110 2.4 236 2.1 302 2.2 90 4.4
Any Substance Abuse 159 5.6 240 5.1 464 4.4 641 4.8 95 4.7 0.0575
Alcohol Abuse 105 3.7 156 3.3 272 2.6 442 3.3 63 3.1 0.0025
Drug Abuse 98 3.4 148 3.1 289 2.7 386 2.9 60 3.0 0.2773
PTSD 82 2.9 206 4.4 559 5.2 551 4.1 105 5.2 <0.0001
Other Anxiety Disorders 177 6.2 375 8.0 917 8.6 1103 8.3 174 8.6 0.0010
Personality Disorder 8 0.3 41 0.9 60 0.6 73 0.5 24 1.2 0.0002
Charlson Comorbidities‡
 0 942 33.0 2064 43.8 4461 41.9 5281 39.5 763 37.7
 1 667 23.4 1107 23.5 2427 22.8 3136 23.5 546 27.0 <0.0001
 >1 1246 43.6 1545 32.8 3763 35.3 4939 37.0 717 35.4
Inpatient Days
 0 1857 65.0 3746 79.4 8441 79.3 10250 76.7 1593 78.6
 1–5 285 10.0 284 6.0 713 6.7 994 7.4 135 6.7 <0.0001
 >5 713 25.0 686 14.5 1497 14.1 2112 15.8 298 14.7
Nursing Home Days:
 0 2744 96.1 4562 96.7 10299 96.7 12777 95.7 1950 96.2
 1–30 68 2.4 89 1.9 224 2.1 378 2.8 50 2.5 0.0018
 >30 43 1.5 65 1.4 128 1.2 201 1.5 26 1.3
Fiscal Year
 2001 618 21.6 782 16.6 451 4.2 2018 15.1 212 10.5
 2002 416 14.6 968 20.5 852 8.0 2203 16.5 226 11.2 <0.0001
 2003 296 10.4 950 20.1 1284 12.1 2116 15.8 188 9.3
 2004 275 9.6 787 16.7 1693 15.9 1872 14.0 222 11.0
 2005 275 9.6 448 9.5 1777 16.7 1617 12.1 253 12.5
 2006 343 12.0 311 6.6 1661 15.6 1371 10.3 299 14.8
 2007 319 11.2 245 5.2 1420 13.3 1148 8.6 295 14.6
 2008 313 11.0 225 4.8 1513 14.2 1011 7.6 331 16.3
Urban facility 2507 87.8 4037 85.6 9761 91.6 11952 89.5 1781 87.9 <0.0001
Academic Affiliation:
 low 599 21.0 1150 24.4 2131 20.0 3447 25.8 467 23.1
 limited 1011 35.4 1432 30.4 3388 31.8 4519 33.8 724 35.7 <0.0001
 high 1245 43.6 2134 45.3 5132 48.2 5390 40.4 835 41.2
Facility Size (# beds)
 <=200 758 26.6 1182 25.1 2154 20.2 3001 22.5 491 24.2
 201–400 766 26.8 1060 22.5 2769 26.0 3552 26.6 618 30.5 <0.0001
 401–600 752 26.3 1424 30.2 3511 33.0 3943 29.5 490 24.2
 >600 579 20.3 1050 22.3 2217 20.8 2860 21.4 427 21.1
Open in a new tab
*

includes Schizophrenia and Schizoaffective Disorder

Ψ

Total number of psychiatric illnesses

‡

Includes myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, COPD, rheumatologic disease, peptic ulcer disease, cirrhosis, hepatic failure, diabetes mellitus, diabetes mellitus with complications, hemiplegia, chronic renal disease, malignant neoplasm, leukemia, lymphomas, metastatic solid tumor, and AIDS.

All use and diagnoses data are based on one year prior to the initiation of the medication.

Individual Medication Use and Mortality

The crude 6-month mortality rates were as follows: haloperidol 20.0%; olanzapine 12.6%; risperidone 12.5%; valproic acid and its derivatives 9.8%; and quetiapine 8.8% (X2 = 294.4, df=4, p<0.0001). The mortality rate rankings were also consistent in the intent to treat and exposure analyses (Table 2).

Table 2.

Crude death rates for dementia patients with new medication starts after a 12-month clean period

Medication Intent-to-Treat Exposure
N #Deaths within 180 days Total person-years Death rate per 100 person-years 95%CI #Deaths in exposure Total person-years Death rate per 100 person-years 95%CI
haloperidol 2855 570 1245.0 45.8 (42.1–49.7) 294 710.9 41.4 (36.8–46.4)
olanzapine 4716 596 2169.4 27.5 (25.3–29.8) 371 1521.1 24.4 (22.0–27.0)
quetiapine 10651 933 5019.2 18.6 (17.4–19.8) 531 3484.9 15.2 (14.0–16.6)
risperidone 13356 1669 6162.4 27.1 (25.8–28.4) 935 4165.5 22.4 (21.0–23.9)
valproic acid and derivatives 2026 199 948.5 21.0 (18.2–24.1) 123 669.3 18.4 (15.3–21.9)
Open in a new tab

Multivariate adjustment, as well as the propensity-weighted and propensity-stratified adjustments yielded similar results. Adjusted RRs (Table 3) averaged over 180-day period showed consistently that haloperidol had the highest mortality risk and quetiapine the lowest. In all but one analysis, valproic acid and its derivatives showed a risk higher than quetiapine, but lower than the other antipsychotics. Figure 1 shows covariate-adjusted survival function by days of exposure.

Table 3.

Relative risks of 180-day mortality for dementia patients with new medication starts after a 12-month clean period

Intent to Treat Adjusted, unweighted Propensity-weighted Propensity-stratified
Medication Hazard Ratio 95%CI P-VALUE Hazard Ratio 95%CI P-VALUE Hazard Ratio 95%CI P-VALUE
risperidone 1.0 1.0 1.0
haloperidol 1.54 (1.38–1.73) <.0001 1.57 (1.39–1.78) <.0001 1.54 (1.38–1.73) <.0001
olanzapine 0.99 (0.89–1.10) 0.8748 1.03 (0.92–1.16) 0.6194 1.00 (0.90–1.12) 0.9941
quetiapine 0.73 (0.67–0.80) <.0001 0.74 (0.67–0.81) <.0001 0.74 (0.68–0.81) <.0001
valproic acid and derivatives 0.91 (0.78–1.06) 0.2468 0.97 (0.83–1.14) 0.7220 0.93 (0.80–1.09) 0.3831
Exposure Adjusted, unweighted Propensity-weighted Propensity-stratified
risperidone 1.0 1.0
haloperidol 1.59 (1.36–1.85) <.0001 1.61 (1.37–1.89) <.0001 1.56 (1.34–1.81) <.0001
olanzapine 1.06 (0.93–1.22) 0.3954 1.10 (0.95–1.28) 0.2062 1.07 (0.93–1.23) 0.3347
quetiapine 0.74 (0.65–0.83) <.0001 0.74 (0.65–0.85) <.0001 0.74 (0.66–0.84) <.0001
valproic acid and derivatives 0.96 (0.79–1.17) 0.6996 1.04 (0.84–1.29) 0.7057 0.99 (0.82–1.21) 0.9568
Open in a new tab

Note: All relative risks were based on Cox regression adjusted for gender, age, race, marital status, delirium, depression, schizophrenia, bipolar I, bipolar II, other psychoses, parkinson’s disease, substance abuse, PTSD, other anxiety, personality disorder, use of benzodiazepine, antidepressent, opiod, days in hospitalization, days in nursing home, fiscal year of index drug use, rurality of facility, facility size, academic affiliation of facility, Charlson’s comorbidity index, myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, COPD, rheumatologic disease, peptic ulcer disease, cirrhosis, hepatic failure, diabetes mellitus, diabetes mellitus with complications, hemiplegia, chronic renal disease, malignant neoplasm, leukemia, lymphomas, metastatic solid tumor, and AIDS.

Figure 1.

Figure 1

Open in a new tab

Covariate Adjusted Survival Function by Days of Exposure

Secondary Analyses

Site of Care

Haloperidol users had the highest proportion of non-psychiatric visits associated with the prescription (77.6%) as compared to 52.4–57.8% of the other medications. An analysis stratified by script location produced results consistent with those from the main analyses, with haloperidol having the highest risk in both settings (propensity stratified results: non-psychiatric script RR 1.42, 95% CI 1.19–1.70, p< 0.001 and psychiatric script RR 1.41, 95% CI 0.93–2.13, p=.1068) and quetiapine the lowest (propensity stratified results: non-psychiatric script RR 0.75, 95% CI 0.64–0.88, p=0.0006; psychiatric prescription RR 0.71, 95% CI 0.55–0.91, p=0.006). Antipsychotic Dose Patients taking only “as needed” (PRN) antipsychotics (n=3,613) or valproic acid and its derivatives were not included in the analyses adjusting for dose. Table 4 shows summary statistics of initial prescribed doses and haloperidol equivalent doses. The majority of patients (81.6%) had initial haldol equivalent doses less than 1.5 mg, while 5.4% had prescribed doses ≥3 mg and 13.0% had prescribed doses between 1.5 to <3 mg. RR estimates adjusted for dose showed mortality risk order consistent with the main analyses.

Table 4.

Prescribed initial daily dose and haloperidol equivalent daily dose of antipsychotic medications

Average Initial Dose
Antipsychotic N Mean Min 25th Percentile 50th Percentile 75th Percentile 95th Percentile Max
haloperidol 1809 1.758 0.250 0.500 1.000 2.000 5.000 25.000
olanzapine 4446 4.715 0.625 2.500 5.000 5.000 10.000 40.000
quetiapine 9371 51.871 0.750 25.000 25.000 50.000 150.000 1600.000
risperidone 11962 0.818 0.100 0.500 0.500 1.000 2.000 8.000
Haloperidol Equivalent Average Initial Dose
Antipsychotic N Mean Min 25th Percentile 50th Percentile 75th Percentile 95th Percentile Max
haloperidol 1809 1.758 0.250 0.500 1.000 2.000 5.000 25.000
olanzapine 4446 1.937 0.149 0.832 1.967 1.967 4.654 26.045
quetiapine 9371 0.585 0.002 0.201 0.201 0.486 1.966 39.957
risperidone 11962 1.096 0.088 0.584 0.584 1.319 2.979 15.188
Open in a new tab

Changes in Mortality Risks Over Time

Using a piece-wise exponential model, mortality risk was found to be on average 1.5 times higher in the first 120 days than for the subsequent 60 days across all medications, except haloperidol. Haloperidol risk was highest in the first 30 days (RR compared to risperidone in 150–180 day period was 2.24, p<0.001), and then the risk significantly decreased to no difference by 90–120 day period (RR=1.11 compared to risperidone in 150–180 day period, p=0.65). We note, however, that the exposure days were significantly shorter for haloperidol than for other medications (median of 60 days for haloperidol versus 111 days or longer for other medication groups). The RRs between olanzapine and risperidone and between valproic acid and its derivatives and risperidone were not significantly different during the 180-day period. Quetiapine risk was consistently lower than that of risperidone, with RRs of 0.67 (p <0.001) for 0–30 days, 0.76 (p < 0.01) for 30–60 days, 0.74 (p = 0.02) for 60–90 days, and 0.72 (p = 0.02) for 90–120 days, each relative to risperidone risk in 150–180 day period. After 120 days, there were no longer significant mortality risk differences between any of the medications.

Additional Sensitivity Analyses

Two additional analyses were performed to further understand mortality risk differences. First, we explored whether the relatively larger proportion of Parkinson’s disease (PD) patients in the quetiapine cohort may have resulted in a lower mortality risk. Compared with non-PD patients taking quetiapine, PD patients tended to receive lower quetiapine doses, and also had less medical burden, but were more likely to have depression. However, after covariate adjustment, PD patients actually had higher mortality rates than non-PD patients in the quetiapine cohort (RR 1.39, 95% CI 1.18–1.64, p< 0.001).

Secondly, we sought to further confirm haloperidol’s role as the agent with the highest mortality by comparing haloperidol and risperidone users after individually matching each haloperidol patient with up to two risperidone patients including age, site of care (psychiatric vs. non-psychiatric prescription), race and medical comorbidity (Charlson score, presence of delirium diagnosis, inpatient hospitalization in the prior year). This analysis based on 2,757 patients (n=1056 haloperidol patients and n=1691 matching risperidone patients) showed that haloperidol users were at higher risk of mortality than risperidone users with an adjusted RR of 1.45 (p = 0.06) for the exposure analysis and 1.57 (p < 0.001) for the intent to treat analysis.

Finally, a true intent to treat analysis (including patients who subsequently switched from their initial medication) did not yield different conclusions: haloperidol users had the highest covariate-adjusted mortality rates (RR 1.50, 95% CI 1.35–1.67) followed by olanzapine (RR 1.02, 95% CI 0.92–1.12), risperidone (reference), valproic acid and its derivatives (RR 0.95, 95% CI 0.82–1.10) and quetiapine (RR 0.76, 95% CI 0.70–0.82).

DISCUSSION

In this large US national sample of outpatients with dementia newly started on an antipsychotic or valproic acid and its derivatives, we examined differences in mortality among individual medications. Consistent across analyses was the finding that haloperidol had the highest mortality risk and quetiapine the lowest. Valproic acid and its derivatives, included as a non-antipsychotic comparison, generally had morality risks higher than quetiapine and similar to risperidone. Across all medications other than haloperidol, mortality risk was found to be on average 1.5 times higher in the first 120 days than for the subsequent period; for haloperidol, risk was highest in the first 30 days and then significantly and sharply decreased.

Haloperidol’s association with the highest mortality risks in this study is not surprising and is confirmatory of prior findings. A number of prior observational studies have reported that conventional antipsychotics are associated with higher mortality risks than atypical antipsychotics. [8,14] In addition, Schneider and colleagues’ meta-analysis of atypical antipsychotics [2] showed haloperidol to have a higher relative risk of mortality compared to placebo (RR=1.68) than did atypicals antipsychotics (RR=1.54). The relationship between haloperidol and mortality may be confounded by selection issues and underlying user characteristics, particularly given secular trends in which atypical antipsychotics largely replaced conventionals in the 1990’s.[10, 18, 19] The shift from conventional to atypical antipsychotics during this period is thought to be due to several factors: 1) efficacy evidence from early clinical trials; 2) perceived safety advantages; and 3) published expert consensus guidelines [20]. In this study, we found that patients receiving haloperidol were older, sicker (highest Charlson scores, most inpatient days and highest concurrent delirium diagnoses), and more likely to be African American than users of atypicals. After controlling for those confounding factors, the haloperidol- associated risks remained significant, although it should be noted that the main risk of mortality with this agent appeared to be in the first month of treatment with a rapid decrease in mortality over time. The majority of haloperidol users (approximately 78%) received their prescriptions at non-psychiatric visits; this result paired with likelihood that haloperidol is used for delirium in inpatient settings and therefore on discharge these could be picked up in observational data as “new prescriptions”, might again suggest that the mortality difference could be influenced by unmeasured medical confounders such as unrecorded delirium episodes. However, our sensitivity analysis where we matched users of haloperidol to users of risperidone (chosen as the most relevant clinical comparison) using variables including age, race, medical comorbidities and site of prescription (psychiatric vs. non-psychiatric) did not corroborate this concern; here, although only marginally significant due to the reduction in sample size from matching, haloperidol showed increased mortality risk over risperidone (RR=1.45).

What about differences in risk among the atypicals? A recent case-control study [9] comparing information about antipsychotic users with non-users found that quetiapine was not associated with short-term increases in mortality, while the other atypical antipsychotics studied were. The study did not directly compare antipsychotics to each other to assess differential risk among these agents. Additionally the comparison of antipsychotic users to nonusers may have been problematic, as the underlying behavioral and frailty issues prompting medication use may be linked inextricably with mortality and may substantially overestimate the mortality risk of antipsychotics [21]. Using a variety of approaches to control for potential selection bias, our study focused on head-to-head antipsychotic comparisons to find differential risks among the atypical agents.

Notably, we found that quetiapine had the lowest risk of mortality across all analyses. Clinically, quetiapine is often prescribed in low doses by providers for sedation and hypnotic purposes, thus, we also performed analyses controlling for antipsychotic dose. In these analyses as well, quetiapine was associated with significantly lower risk. It is not entirely clear why quetiapine would have lower risk than the other atypicals. Some of the lower risk could have to do with quetiapine’s receptor or side effect profile. A significantly higher proportion of quetiapine users had Parkinson’s disease; however, our sensitivity analyses did not indicate that this was a likely explanation for quetiapine’s lower mortality risk. An alternative explanation might be that quetiapine’s lower mortality risk has more to do with the patients it is prescribed for, perhaps patients with milder dementia or behavioral disturbances. Notably, there is no rapid-acting form of quetiapine as there are for other atypicals, thus, this agent is likely used less in urgent situations.

In most analyses, valproic acid and its derivatives showed a risk higher than quetiapine but no different than that of risperidone or olanzapine. Thus, the use of valproic acid and its derivatives as an alternative to antipsychotics to address neuropsychiatric symptoms of dementia may not be without risks as well. Studies have linked anticonvulsant use in the elderly with fracture [22], somnolence and thrombocytopenia [23]. In addition, although valproic acid and its derivatives have been touted for behavioral stabilization and even as potentially neuroprotective, the evidence has been lacking for such efficacy [24, 25].

The greatest mortality risk for haloperidol was found to be within the first 30 days whereas for the other medications, mortality risk was higher in the first 120 days than the subsequent period. As noted above, the exposure period for haloperidol was considerably shorter than for the other agents, and thus, the haloperidol result may in part relate to its selection for older and sicker patients, a number of whom may have this agent started during an inpatient stay for delirium or in non-psychiatric settings. The clinical implications of increased mortality risk for atypical antipsychotics and valproic acid and derivatives in the first four months of treatment than the later period may be several-fold: 1) if these agents are prescribed then they should be used in conjunction with a risk-benefit approach [26, 27] with consideration of the established efficacy of each agent [2–4, 6, 20, 26]; 2) patients should be monitored during the acute treatment period for side effects and adverse reactions; and 3) periodic attempts at discontinuation should be attempted particularly in light of the DART-AD trial results. In the DART-AD trial [5], the investigators randomized patients with dementia taking antipsychotics to either continuation of antipsychotics {the majority taking risperidone (67%) or haloperidol (26%)} vs. antipsychotic discontinuation/placebo and found a significantly higher mortality rate for patients who continued antipsychotics. Difference in length of mortality risk for antipsychotics between DART-AD (>12 months) and our study (120 or less) may relate to exposure periods and differences in study samples.

The use of administrative data for pharmacoepidemiologic work has several limitations. Prescription fills can be an imprecise measure of actual drug exposure; medication fills may not reflect day-to-day usage. Data on dementia severity were also lacking, although not accounting for this confounder may actually contribute to an underestimation of the mortality risk of haloperidol [21]. Finally, while the large integrated VA health system offers us the opportunity to examine pharmacoepidemiologic changes, the findings may not be completely generalizable. Consistent with the demographic characteristics of the VA patient population, the study cohort was primarily male. However, we note that there are striking similarities on many key variables that might affect provider antipsychotic prescribing practices (e.g. race mix, prevalence of key psychiatric and medical conditions) between our data and other national data [28, 29]. Finally, the issue of concurrent delirium in dementia is a key one, and as with other observational studies, we used diagnostic codes to denote the presence of delirium. Given the lack of recognition of delirium in many cases, underdiagnosis is a problem; however, we have no evidence to suggest that such underdiagnosis varies by antipsychotic agent. Despite these limitations, our results using various analytic methods consistently indicated differences in mortality risks among individual antipsychotic agents. Further, the use of valproic acid and its derivatives as alternative agents to address the neuropsychiatric symptoms of dementia may carry associated risks as well.

Acknowledgments

This research was supported by a grant from the National Institute of Mental Health, R01-MH081070. Resources were also contributed by the Serious Mental Illness Treatment, Resource, and Evaluation Center, Ann Arbor, MI.

Footnotes

The views expressed in this article are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs.

References

  • 1.Saha S, Robinson S, Bainco T, Schechtel M, Hickam D. In A Summary for Clinicians and Policy Makers. Agency for Healthcare Research and Quality; 2007. Off-Label Use of Atypical Antipsychotic Drugs; pp. 1–4. [PubMed] [Google Scholar]
  • 2.Schneider LS, Dagerman KS, Insel P. Risk of death with atypical antipsychotic drug treatment for dementia: meta-analysis of randomized placebo-controlled trials. JAMA. 2005;294(15):1934–1943. doi: 10.1001/jama.294.15.1934. [DOI] [PubMed] [Google Scholar]
  • 3.Kales HC, Valenstein M, Kim HM, McCarthy JF, Ganoczy D, Cunningham F, Blow FC. Mortality risk in patients with dementia treated with antipsychotics versus other psychiatric medications. The American journal of psychiatry. 2007;164(10):1568–1576. doi: 10.1176/appi.ajp.2007.06101710. quiz 1623. [DOI] [PubMed] [Google Scholar]
  • 4.Gill SS, Bronskill SE, Normand SL, Anderson GM, Sykora K, Lam K, Bell CM, Lee PE, Fischer HD, Herrmann N, et al. Antipsychotic drug use and mortality in older adults with dementia. Ann Intern Med. 2007;146(11):775–786. doi: 10.7326/0003-4819-146-11-200706050-00006. [DOI] [PubMed] [Google Scholar]
  • 5.Ballard C, Hanney M, Theodoulou M, Douglas S, McShane R, Kossakowski K, Gill R, Juszczak E, Yu L-M, Jacoby R. The dementia antipsychotic withdrawal trial (DART-AD): long-term follow- up of a randomised placebo-controlled trial. Lancet neurology. 2009;8(2):151–157. doi: 10.1016/S1474-4422(08)70295-3. [DOI] [PubMed] [Google Scholar]
  • 6.FDA. Safety Labeling Changes for Antipsychotic Drugs. 2008. [Google Scholar]
  • 7.Kryzhanovskaya L, Jeste D, Young C, Polzer J, Roddy T, Jansen J, Carlson J, Cavazzoni P. A review of treatment-emergent adverse events during olanzapine clinical trials in elderly patients with dementia. The Journal of clinical psychiatry. 2006;67(6):933–945. doi: 10.4088/jcp.v67n0610. [DOI] [PubMed] [Google Scholar]
  • 8.Schneeweiss S, Setoguchi S, Brookhart A, Dormuth C, Wang PS. Risk of death associated with the use of conventional versus atypical antipsychotic drugs among elderly patients. CMAJ. 2007;176(5):627–632. doi: 10.1503/cmaj.061250. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Rossom RC, Rector TS, Lederle FA, Dysken MW. Are All Commonly Prescribed Antipsychotics Associated with Greater Mortality in Elderly Male Veterans with Dementia? Journal of the American Geriatrics Society. 2010;58(6):1027–1034. doi: 10.1111/j.1532-5415.2010.02873.x. [DOI] [PubMed] [Google Scholar]
  • 10.Kales HC, Zivin K, Kim HM, Valenstein M, Chiang C, Ignacio RV, Ganoczy D, Cunningham F, Schneider LS, Blow FC. Trends in Antipsychotic Use in Dementia 1999–2007. Arch Gen Psychiatry. 2010 doi: 10.1001/archgenpsychiatry.2010.200. In Press. [DOI] [PubMed] [Google Scholar]
  • 11.Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373–383. doi: 10.1016/0021-9681(87)90171-8. [DOI] [PubMed] [Google Scholar]
  • 12.Adamis D, Treloar A, Martin FC, Macdonald AJ. Recovery and outcome of delirium in elderly medical inpatients. Arch Gerontol Geriatr. 2006;43(2):289–298. doi: 10.1016/j.archger.2005.11.005. [DOI] [PubMed] [Google Scholar]
  • 13.Kales HC, Kamholz BA, Visnic SG, Blow FC. Recorded delirium in a national sample of elderly inpatients: potential implications for recognition. Journal of geriatric psychiatry and neurology. 2003;16(1):32–38. doi: 10.1177/0891988702250535. [DOI] [PubMed] [Google Scholar]
  • 14.Wang PS, Schneeweiss S, Avorn J, Fischer MA, Mogun H, Solomon DH, Brookhart MA. Risk of death in elderly users of conventional vs. atypical antipsychotic medications. N Engl J Med. 2005;353(22):2335–2341. doi: 10.1056/NEJMoa052827. [DOI] [PubMed] [Google Scholar]
  • 15.Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Statistical Assoc. 1958;53:457–481. [Google Scholar]
  • 16.Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70(1):41–55. [Google Scholar]
  • 17.Andreasen N, Pressler M, Nopoulos P, Miller D, Ho B-C. Antipsychotic dose equivalents and dose-years: a standardized method for comparing exposure to different drugs. Biological psychiatry. 67(3):255–262. doi: 10.1016/j.biopsych.2009.08.040. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Lopez O, Becker J, Sweet R, Klunk W, Kaufer D, Saxton J, DeKosky S. Patterns of change in the treatment of psychiatric symptoms in patients with probable Alzheimer’s disease from 1983 to 2000. The Journal of Neuropsychiatry and Clinical Neurosciences. 2003;15(1):67–7. doi: 10.1176/jnp.15.1.67. [DOI] [PubMed] [Google Scholar]
  • 19.Rapoport M, Mamdani M, Shulman KI, Herrmann N, Rochon PA. Antipsychotic use in the elderly: shifting trends and increasing costs. International journal of geriatric psychiatry. 2005;20(8):749–753. doi: 10.1002/gps.1358. [DOI] [PubMed] [Google Scholar]
  • 20.Schneider LS, Dagerman K, Insel PS. Efficacy and adverse effects of atypical antipsychotics for dementia: meta-analysis of randomized, placebo-controlled trials. Am J Geriatr Psychiatry. 2006;14(3):191–210. doi: 10.1097/01.JGP.0000200589.01396.6d. [DOI] [PubMed] [Google Scholar]
  • 21.Schneeweiss S, Setoguchi S, Brookhart MA, Kaci L, Wang P. Assessing residual confounding of the association between antipsychotic medications and risk of death using survey data. CNS drugs. 2009;23(2):171–180. doi: 10.2165/00023210-200923020-00006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Mezuk B, Morden N, Ganoczy D, Post E, Kilbourne A. Anticonvulsant use, bipolar disorder, and risk of fracture among older adults in the Veterans Health Administration. The American Journal of Geriatric Psychiatry. 18(3):245–255. doi: 10.1097/JGP.0b013e3181bf9ebd. [DOI] [PubMed] [Google Scholar]
  • 23.Porsteinsson AP, Tariot PN, Erb R, Cox C, Smith E, Jakimovich L, Noviasky J, Kowalski N, Holt CJ, Irvine C. Placebo-controlled study of divalproex sodium for agitation in dementia. The American Journal of Geriatric Psychiatry. 2001;9(1):58–66. [PubMed] [Google Scholar]
  • 24.Tariot P, Loy R, Ryan JM, Porsteinsson A, Ismail S. Mood stabilizers in Alzheimer’s disease: symptomatic and neuroprotective rationales. Advanced Drug Delivery Reviews. 2002;54(12):1567–1577. doi: 10.1016/s0169-409x(02)00153-9. [DOI] [PubMed] [Google Scholar]
  • 25.Tariot P, Raman R, Jakimovich L, Schneider L, Porsteinsson A, Thomas R, Mintzer J, Brenner R, Schafer K, Thal L. Divalproex sodium in nursing home residents with possible or probable Alzheimer Disease complicated by agitation: a randomized, controlled trial. The American Journal of Geriatric Psychiatry. 2005;13(11):942–949. doi: 10.1176/appi.ajgp.13.11.942. [DOI] [PubMed] [Google Scholar]
  • 26.US Food and Drug Administration. Public Health Advisory. US Department of Health and Human Services; 2005. Deaths with Antipsychotics in Elderly Patients with Behavioral Disturbances. [Google Scholar]
  • 27.Lyketsos C, Colenda C, Beck C, Blank K, Doriaswamy M, Kalunian D, Yaffe K. Directors Bo: American Association for Geriatric Psychiatry, editor. American Association for Geriatric Psychiatry Position Statement. 2005. Principles of Care for Patients With Dementia Resulting From Alzheimer Disease. [DOI] [PubMed] [Google Scholar]
  • 28.Langa K, Plassman B, Wallace R, Herzog AR, Heeringa S, Ofstedal M, Burke J, Fisher G, Fultz N, Hurd M, et al. The Aging, Demographics, and Memory Study: study design and methods. Neuroepidemiology. 2005;25(4):181–191. doi: 10.1159/000087448. [DOI] [PubMed] [Google Scholar]
  • 29.Plassman BL, Langa KM, Fisher GG, Heeringa SG, Weir DR, Ofstedal MB, Burke JR, Hurd MD, Potter GG, Rodgers WL, et al. Prevalence of dementia in the United States: the aging, demographics, and memory study. Neuroepidemiology. 2007;29(1–2):125–132. doi: 10.1159/000109998. [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES