Multiple Versus Single Antipsychotic Agents for Hospitalized Psychiatric Patients: Case-Control Study of Risks Versus Benefits

We screened the medical records of all inpatients with DSM-IV psychotic disorders admitted to McLean Hospital during a 3-month period (March–May) in 1998. Computerized prescription drug records were used to identify “cases” given two or more antipsychotic drugs simultaneously for at least 3 consecutive days during the index hospitalization, with no exclusion criteria applied. We then matched each case with a “comparator” subject hospitalized within the same 3 months but treated with only one antipsychotic drug at one time. Matching was by sex, DSM-IV category (schizophrenia or schizophreniform disorder, bipolar disorder with psychotic features, or other disorders), age (within 5 years), and illness severity at admission as estimated by two retrospective psychiatric ratings: Clinical Global Impression (CGI) (24) (within 1 point) and Global Assessment of Functioning (GAF) (DSM-IV, p. 32) (within 5 points).

We determined the initial, maximum, and final or discharge total daily antipsychotic dose for each subject and recorded the days of use of each agent. In addition, the percentage increase from the initial to maximum daily dose was computed as a within-subject index of antipsychotic dosing. Reported doses (mg/day) were based on conversion to the approximate daily chlorpromazine equivalent, using recently developed conversion factors reported elsewhere (10) for all FDA-approved antipsychotic agents.

Based on independent assessments of data provided in medical records by two clinician-investigators (J.L.G., P.S.), estimated baseline and discharge ratings for the GAF, CGI, and Positive and Negative Syndrome Scale (25) were made (with consensus ratings in cases of discrepancies) independent of knowledge of dosing. Percentage changes in the three clinical assessments between hospital admission and discharge were computed for each subject. We also defined three subscales from the Positive and Negative Syndrome Scale to identify relationships of specific symptom complexes in cases versus comparators. These ad hoc subscales were: disorganization (Positive and Negative Syndrome Scale items P2, N5, N7, G10, G11, G12), anxiety/agitation (P4, P7, G3, G4, G14), and paranoia/hypervigilance (P6, G8, G9, G16).

Also from medical records data, we determined prior hospitalization count, approximate age at onset of the primary psychiatric diagnosis, length of stay (days) for the index hospitalization, and the type and number of all other psychotropic drugs used. We identified adverse effects recorded during hospitalization, by category, including neurological and general medical events, determined if they were present at admission or emerged during hospitalization, and supplemented this information with incident use of drugs employed to manage extrapyramidal side effects (anticholinergics and lipophilic beta-blockers in the absence of a cardiovascular indication). Both total adverse events and percentage of patients with at least one adverse event were recorded.

We examined factors and clinical outcomes associated with the treatment classification (antipsychotic polytherapy versus monotherapy) other than those used to match the groups, with particular attention to contrasting antipsychotic doses (initial, maximum, and final), incidence and types of medically adverse effects, hospital length of stay, and improvements in clinical ratings. Contrasts between antipsychotic polytherapy cases and monotherapy comparators were carried out by using conditional logistic regression methods for binary outcomes (adverse effects), negative binomial regression for prior hospitalization counts, generalized least-squares regression with binomial family and log-link to estimate relative risk for adverse effects, and generalized least-squares regression methods (Gaussian family) for the remaining outcomes (changes in clinical rating scale scores from baseline, chlorpromazine-equivalent daily doses, and length of stay). Regression analyses included adjustment for clustering on matched pairs. For selected contrasts, several covariates were considered in additional modeling; these included sex, current age, diagnostic category, number of previous hospitalizations, and baseline CGI and GAF ratings. Postmodeling chi-square tests were carried out for selected contrasts. For some bivariate comparisons involving continuous measures, Spearman rank correlations (r_s) were used. Chlorpromazine-equivalent dose and length of stay data were positively skewed (few patients with unusually high doses and long hospitalizations) and so were log-normalized before analysis. We checked modeling fits using partial-residual plotting methods. Robust estimates of standard errors were obtained when feasible.

Averaged continuous data are reported as means with standard deviations, except for the skewed dose and length of stay data, for which medians and standard deviations are reported. Statistical significance required two-tailed p<0.05. Analyses employed Stata (Stata Corp, College Station, Tex.) and Statview-5 (SAS Corp., Cary, N.C.) microcomputer programs.

A total of 140 inpatients in 70 matched pairs were studied. Matching of polytherapy cases and monotherapy comparators was based on current age, sex, diagnostic category, and initial GAF and CGI ratings, but not other factors, all of which are summarized in Table 1. Of all factors considered, the only significant difference found was that patients receving antipsychotic polytherapy were on average 4.2 years younger at illness onset than those receiving antipsychotic monotherapy (Table 1). On two severity indicators not used in matching, Positive and Negative Syndrome Scale scores (including total, positive, and negative scores) and the number of prior hospitalizations, the treatment groups did not differ at baseline (Table 1). The total number of simultaneously administered psychotropic medicines ranged from one to seven, with polytherapy cases averaging 4.29 medications (SD=1.23) and monotherapy comparators averaging 3.47 (SD=1.46). This difference, while large, basically reflects the circumstance that comparator subjects were prescribed one antipsychotic agent, while all polytherapy cases received at least two such drugs at the same time, and six received three different antipsychotic agents at some point during the inpatient stay.

Principal findings in this retrospective case-control comparison of psychiatric inpatients receiving antipsychotic monotherapy versus polytherapy, based on chart reviews, include the following points. 1) Maximum and final total chlorpromazine-equivalent antipsychotic doses, as well as the percentage increase in total daily doses during hospitalization, all were much higher among polytherapy patients. 2) Adverse effects occurred more frequently among polytherapy patients in association with higher total antipsychotic doses. 3) Hospitalization was much longer in antipsychotic polytherapy cases, particularly among patients with adverse effects. 4) Clinical outcomes, based on changes in CGI and GAF ratings and Positive and Negative Syndrome Scale score, did not differ between the matched treatment groups even after adjustment for age, sex, diagnosis, and initial CGI and GAF ratings. 5) Patients treated with antipsychotic monotherapy or polytherapy did not differ in the total number, or types, of other coadministered psychotropic agents. 6) Patients treated with antipsychotic polytherapy were 4.2 years younger at age at onset of illness than those treated with antipsychotic monotherapy. Too few patients were enrolled to make valid comparisons in outcomes associated with specific antipsychotic agents.

These findings are based on relatively brief inpatient treatment, when use of multiple agents of the same type is most likely to occur, and may not generalize to other circumstances (3–7). In this setting, several summary measures of clinical change during hospital stays lasting a median of 20 days (SD=25.9) did not differ between the antipsychotic polytherapy and monotherapy groups, whereas hospitalization was more than a week longer with polytherapy. The findings suggest a lack of obvious “benefit” of inpatient antipsychotic polytherapy in either superior clinical outcomes or shorter hospitalizations. Moreover, the relevant “costs” include a somewhat greater clinical burden of adverse effects, as well as the burden of inconvenience, life disruption, and substantial added financial costs associated with longer hospitalizations among the polytherapy cases as well as the higher drug acquisition costs involved. However, younger age at onset in the polytherapy group may represent a population that is more severely ill with a longer duration of illness and decreased medication responsiveness that may tend to encourage use of multiple antipsychotics.

It is somewhat reassuring that the clinical burden of adverse effects associated with antipsychotic polytherapy did not include severe or life-threatening adverse events, probably reflecting the relative safety of modern antipsychotic agents, as well as the limited exposure involved in this study of 140 patients during relatively brief hospitalizations (about 3,700 person-days). Nevertheless, some adverse effects associated with modern and traditional antipsychotic agents are not clinically trivial (e.g., acute and later extrapyramidal reactions, hypotension with falls and injuries or cardiovascular events, and weight gain with potentially increased risk of diabetes and other long-term cardiovascular risks). Some are even life-threatening (particularly the neuroleptic malignant syndrome and its apparent variants associated even with modern drugs [11, 26, 27]). While this study did not demonstrate a difference in serious adverse event rates, the number of participants was too low to detect a difference in such uncommon events. Accordingly, the safety of polytherapy remains to be proven.

The present findings accord very closely with those of at least one previous retrospective study (3), in which longer hospital stays and no clear evidence of differential clinical benefit were found in inpatients treated with multiple versus single psychotropic agents. Both investigations provide no evidence for substantial gains in benefit (such as superior improvements on standard clinical rating scales), and both indicate a pronounced association of psychotropic polytherapy with longer hospitalization. In addition, we report a nearly significant association of polytherapy with incidence of adverse effects and a correlation of adverse effects with extended length of stay. Yet, caution is required, since an appropriate interpretation of the findings exceeds the limits of the study design, particularly regarding the critical question of how to account for the longer hospitalizations associated with use of antipsychotic polytherapy. Nevertheless, since both adverse events and length of stay were closely correlated with the total daily doses of antipsychotics involved with polytherapy, it may be that extra days in the hospital, in part, are occasioned by management of iatrogenic side effects of the more complex treatment.

However, another plausible interpretation is that selection of antipsychotic polytherapy, and the associated longer hospitalization, may both represent responses to patients with less treatment-responsive conditions. Although subjects were closely matched on several baseline demographic and clinical variables, such factors would not necessarily predict individual responsiveness to treatment. The design of the present study does not allow determination of reasons for electing use of antipsychotic polytherapy nor does it avoid the issue through random assignment to treatment options. Also, clinical decisions about the degree of improvement required for discharge were uncontrolled, and the study method did not permit monitoring rates of improvement per time. These are all limitations inherent in the retrospective-naturalistic character of this investigation. Moreover, the study size is limited, even though much larger than several earlier case series involving combinations of antipsychotic drugs (4–7).

Future studies aimed at examining outcomes associated with multiple versus single medication treatment ideally should employ random assignment to fixed and comparable alternative regimens, with efforts to control additional adjunctive pharmacological and psychosocial components of treatment. Such requirements obviously are not easily attained, and such studies are not likely to receive support from the pharmaceutical industry, since multiple proprietary products are involved. A potentially informative, if scientifically less satisfactory, alternative to randomized studies would be large, prospective, naturalistic studies that systematically acquire data on dosing, adverse events, and clinical outcomes during the elective use of regimens of varied complexity.

In conclusion, there is currently no empirical foundation for the evidently increasingly common clinical practice of combining antipsychotic drugs in the short- or long-term treatment of severely mentally ill patients. The consequences of this practice, even in short-term use, are uncertain, although the present findings do not indicate a favorable cost/benefit relationship for hospitalized and mainly psychotic or manic patients. The unknown potential for adverse long-term iatrogenic effects of antipsychotic polytherapy, and of psychiatric polytherapy in general, is of even greater concern. The practice appears to arise from the unwarranted, and potentially harmful, presumption that greater benefits should be achieved with more aggressive treatment at minimal or no extra risk or cost.

Received Aug. 29, 2002; revision received Aug. 11, 2003; accepted Sept. 15, 2003. From the Department of Psychiatry and Neuroscience Program, Harvard Medical School, Boston; the Bipolar and Psychotic Disorders Clinic, McLean Hospital; and Northeastern University School of Pharmacy, Boston. Address reprint requests to Dr. Centorrino, Bipolar and Psychotic Disorders Clinic, North Belknap III, McLean Hospital, Belmont, MA 02478–9106; [email protected] (e-mail). Supported in part by an award from the Bruce J. Anderson Foundation and the McLean Private Donors Neuropsychopharmacology Research Fund (Dr. Baldessarini).