Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
J Clin Psychiatry. Author manuscript; available in PMC 2011 June 7.
Published in final edited form as:
PMCID: PMC3109728

Incidence of Tardive Dyskinesia with Atypical and Conventional Antipsychotic Medications: Prospective Cohort Study



Most previous studies of the incidence of tardive dyskinesia with atypical compared to conventional antipsychotics have not had tardive dyskinesia as their primary focus. The current study aimed to compare the incidence of tardive dyskinesia with atypical vs. conventional antipsychotics using methods similar to those from a previous prospective cohort study at our site in the 1980s.


352 initially tardive dyskinesia-free psychiatric outpatients were examined for a new diagnosis of tardive dyskinesia every 6 months for up to 4 years at a community mental health center. At baseline, subjects were receiving conventional antipsychotics only (23%), atypicals only (64%), or both (14%). Only 26 subjects had never received conventional antipsychotics.


Compared with subjects treated with conventional antipsychotics alone since the previous visit, the adjusted tardive dyskinesia incidence rate-ratio for subjects treated with atypical antipsychotics alone was 0.68 (95% confidence interval 0.29 to 1.64). The incidence and prevalence of tardive dyskinesia was similar to previous findings at this site in the 1980s.


The incidence of tardive dyskinesia with recent exposure to atypical antipsychotics alone was more similar to that for conventional antipsychotics than in most previous studies. Despite high penetration of atypical antipsychotics into clinical practice, the incidence and prevalence of tardive dyskinesia appeared relatively unchanged since the 1980s. Clinicians should continue to monitor for tardive dyskinesia, and researchers should continue to pursue efforts to treat or prevent it.

When the atypical antipsychotics became available, it was hoped that they would be associated with a lower risk of tardive dyskinesia (TD) than the older conventional antipsychotics. A 2004 systematic review of early conventional-controlled and other studies indicated that the evidence seemed to support the idea that this hope had been realized.1 As noted in the review, however, few of the existing studies were designed to focus on TD and its accurate identification. It is possible that a limited focus on TD diagnosis could have introduced bias in favor of atypicals.2 The primary aim of the current study was to compare the incidence of TD among users of atypical and conventional antipsychotics. Methods were similar to those from a previous TD incidence study conducted at our site during the conventional antipsychotic era. 3, 4


Study Design

We conducted a cohort study of TD incidence in a population of outpatients maintained on antipsychotics at the Connecticut Mental Health Center in the United States. Baseline evaluations were conducted November 2000 through May 2003. Following baseline, subjects at risk for TD were followed prospectively with examinations every six months through February 2005.


The source population was the outpatient division at the Connecticut Mental Health Center. When the study began, the Connecticut Mental Health Center served a mostly urban catchment of 250,000 people and maintained an average daily census of roughly 2000 patients, of whom about 60% were maintained on antipsychotic medications. The racial/ethnic breakdown was 57% non-Latino white, 25% non-Latino African-American, and 18% Latino.

Inclusion criteria required subjects to have been maintained on antipsychotic medication for ≥3 months. The sole exclusion criterion was inability to examine subjects for TD due to primary neurological disease (such as Huntington’s). With institutional review board approval, we asked clinicians for permission to approach eligible patients for consent. Consenting subjects underwent baseline evaluation.


At each visit, we examined subjects for dyskinesia using the Abnormal Involuntary Movement Scale (AIMS).5 We gave the AIMS examination twice at each visit, at visit beginning and end, employing the Glazer-Morgenstern criteria for dyskinesia.6 These criteria require the total AIMS score be ≥3, with at least one body area rated ≥2 (mild), on both AIMS exams at that visit. Glazer-Morgenstern criteria are slightly more inclusive than the Schooler-Kane criteria,7 which require at least two body areas be ≥2 or one body area rated ≥3 (moderate). AIMS raters were blind to medication status. In subjects meeting Glazer-Morgenstern criteria, an investigator conducted a verification examination, when possible on the same day.

At study outset, the previous Yale Tardive Dyskinesia study principal investigator (WMG) and project coordinator conducted a full-day training session on use of the AIMS and the Glazer-Morgenstern criteria. Particular attention was paid to distinguishing dyskinesia from akathisia, tremor, dystonia, mannerisms, and tics. After initial training, reliability assessment exercises using videotaped examinations were conducted approximately quarterly. In 17 taped examinations with a median 5 raters per examination, the intraclass correlation for agreement among raters8 on the AIMS total scores was 0.93.

We considered subjects prevalent cases of persistent TD when Glazer-Morgenstern criteria were met at the first visit if there were a history of TD from medical record review. Subjects with no clinical history of TD were considered prevalent cases when Glazer-Morgenstern criteria were met at the first two consecutive visits.

At-risk cases were defined based on history and baseline examination. Prevalent cases and patients who previously received a Glazer-Morgenstern (GM) research diagnosis of persistent TD in the first Yale TD study (1985–1993)3, 4, 6 or in the CMHC TD Clinic (1978–1993),9 were defined as not at risk. Otherwise, subjects were considered at risk to develop TD. At-risk subjects included those with a positive clinical history when the baseline research examination was negative. The at-risk status of these subjects was considered an empirical question because the clinical diagnosis had been established by a means whose reliability had not been evaluated. Lastly, at-risk subjects also included those with a negative clinical history who met GM criteria at the initial but not at the second examination. These cases were considered instances of transient dyskinesia and therefore still at risk for developing persistent TD. All at-risk cases were scheduled for follow-up evaluation every six months.

Incident cases of persistent TD were those who, having first met at-risk criteria at baseline, subsequently met Glazer-Morgenstern criteria at two consecutive follow-up visits for both examinations at each visit and on verification exam when available.

Antipsychotic exposure history was determined for at-risk subjects primarily by review of available medical records, including records sent from other facilities. Prescribed dose and duration were recorded for each lifetime episode of treatment with each antipsychotic and antiparkinsonian agent. We utilized chart information exclusively if there were no missing periods of exposure. When periods were missing, we supplemented chart information for duration of exposure using subject reports that coincided with gaps in the medical record. Subjects generally could not remember specific doses, so subject reports were not used to supplement missing dose information. Staff conducting medical record reviews and interviews of subjects about medications were blind to results of AIMS examinations. We converted all antipsychotic doses to chlorpromazine equivalents, using published equivalencies for oral conventional10 and atypical11 antipsychotics. We converted depot doses to oral doses using the manufacturers’ recommended equivalents: haloperidol (15 mg/4 weeks per 1 mg/d), fluphenazine (12.5 mg/3 weeks per 10 mg/d), and risperidone (25 mg/2 weeks per 2 mg/d); these are supported by empirical studies.1214 Drug exposure variables derived from these data included antipsychotic type and years of exposure to, and average dose of, antipsychotic by type before and since the prior visit.

Antipsychotic exposure during follow-up intervals was characterized as conventional only since the prior visit, atypical only since the prior visit, or both conventional and atypical at some point since the prior visit. We explored heterogeneity in the pattern of overlap and non-overlap within this last group. In approximately half of these intervals (78.7 patient-years of exposure), atypical and conventional medications were prescribed simultaneously for all but 30 days since the prior visit. In the remaining half (87.3 patient-years), patients were prescribed atypical and conventional medications during the intervals in a wide variety of simultaneous, sequential, and cross-tapering patterns. Crude TD incidence rates were similar for these two groups (0.102/year and 0.092/year, respectively), so these exposure intervals were considered together for purposes of analysis.

At baseline, psychiatric diagnoses were established using the Structured Clinical Interview for DSM-IV.15 We also assessed other reported risk factors for incident TD (parkinsonian and akathisia symptoms, psychosis positive and negative symptom severity, premorbid adjustment, educational attainment, handedness, cognitive impairment, obstetrical complications, smoking, diabetes, and alcohol and substance use)4, 1631 to treat as potential confounders or modifiers of the atypical/conventional drug effect.

Statistical Methods

Analyses focused on the relative incidence rate of TD, comparing recent users of atypical antipsychotics only (in the past 6 months) or recent users of both types with recent users of conventional antipsychotics only, controlling for potential confounders (i.e., TD risk factors associated with type of antipsychotic exposure). Proportional hazards analysis was used to estimate drug-type effects (rate-ratios, RR; and their 95% confidence intervals, CI), control for confounders, and assess possible interactions between antipsychotic drug type and other TD predictors by adding product terms to the model. Certain predictors, including type of antipsychotic exposure since the prior visit, were treated as time-dependent variables. Following estimation of crude (unadjusted) recent drug effects, we adjusted for core model variables from our prior report: age at baseline, race, years of conventional antipsychotic exposure, and recent antipsychotic CPZ-equivalent dose.4 The other reported risk factors were then added to the core model one at a time to determine if they had confounded or modified the estimated effect of antipsychotic drug type. Schoenfeld residuals analysis indicated that the proportionality assumption held satisfactorily. In addition to these primary analyses of recent antipsychotic use, we also estimated effects of lifetime use (available on request from SWW).

Literature Review Methods

We searched Pub Med for studies with the words “tardive,” “clozapine,” “risperidone,” “olanzapine,” quetiapine,” “aripiprazole,” or “ziprasidone” in the title, as well as bibliographies and subsequent citations (in Web of Science) of the identified articles. Studies selected for inclusion were those that reported incidence of new onset cases of TD prospectively over time among adult subjects who were free of TD at baseline and that compared incidence on atypical antipsychotic to incidence on conventional antipsychotic. Geriatric and adolescent/child studies were not included. When multiple definitions of incident TD cases were reported, we selected the definition that corresponded most closely to incident persistent TD, e.g. present on two consecutive occasions. From each study and from each identified medication group, we abstracted the number of subjects at risk for TD, the number of incident cases, and follow-up time, or calculated these quantities from published data. Atypical/conventional rate-ratios (RRs) and their 95% confidence intervals were then calculated, and RRs were synthesized across study using a random effects Mantel-Haentzel model in Review Manager 5.32 In studies where several atypical arms were compared to a single conventional arm, the atypical arms were first weighted by person-years of follow-up and pooled.


Sample Description

Baseline evaluation was completed on 619 subjects. Of these, 195 met criteria for persistent TD at baseline (estimated prevalence 31.5%, 95% CI 27.9 to 35.3%) and were ineligible for the incidence analysis. In addition, 23 subjects with negative baseline TD examinations were also not eligible for the incidence sample because of previous Glazer-Morgenstern TD diagnoses. The remaining 401 subjects were free of TD at baseline, of whom 352 were re-examined at least once during follow-up (the study population). Demographic, diagnostic, and treatment characteristics of the study population are shown in Table 1. Female gender and longer histories of conventional exposure were more common among subjects receiving conventional antipsychotics at baseline, and histories of atypical exposure were less common in this group. Most of the 352 subjects qualified as at-risk by virtue of both negative clinical histories and negative initial research examinations (81%). The remainder had no clinical history but a positive research examination that was negative on repeat (3%) or a positive clinical history but negative baseline research examination (17%). The distribution of conventional medications at baseline is shown in Table 1.

Table 1
Demographic, diagnostic, and treatment findings by antipsychotic type at baseline.

At-risk individuals underwent 1344 follow-up examinations. There were 52 new persistent cases of TD detected during 783 person-years of follow-up, yielding an average incidence rate of 0.066/year. TD risk (cumulative incidence) after 3.9 years of follow-up was 19.7% (95% CI 15.2 to 25.1%). The average of the four total AIMS scores leading to TD diagnosis in the 52 incident cases was 4.8 (range 3.0 to 8.2).

Estimated Effects of Recent Antipsychotic Type on New Occurrence of TD

Crude analyses revealed that patients receiving conventional antipsychotic alone since the prior visit developed new-onset TD at a rate of 5.6 per 100 patient-years of exposure (8 cases per 141.8 patient-years, 0.056/year), patients receiving atypical antipsychotic alone developed TD at a rate of 0.059/year (28 cases per 475.2 patient-years), and patients receiving both types since the prior visit developed TD at a rate of 0.096/year (16 cases in 166.0 patient-years). Based on crude (unadjusted) analyses, subjects treated with atypical antipsychotics alone since the prior visit developed TD at a similar rate as subjects treated with conventionals alone (crude RR=1.04; 95% CI 0.50 to 2.22). Subjects treated with both types of antipsychotic since the prior visit developed TD at a somewhat higher rate as subjects treated with conventionals alone (crude RR=1.71; 95% CI 0.77 to 3.82).

Based on adjusted results from our core model, subjects treated with atypical antipsychotics alone since the prior visit developed TD at approximately two-thirds the rate as subjects treated with conventionals alone (adjusted RR=0.68; 95% CI 0.29 to 1.64). Subjects treated with both types of antipsychotic since the prior visit developed TD at nearly double the rate as subjects treated with conventionals alone (adjusted RR=1.85; 95% CI 0.72 to 4.75).

Logistic regression models on new TD present at any time during follow-up were also fitted by baseline medication status, employing as a covariate only years of total prior antipsychotic exposure or years of total prior conventional exposure, each expressed as a continuous measure. These analyses produced findings comparable to those of the unadjusted cox regressions: a similar proportion of subjects treated with atypical antipsychotics alone at baseline developed TD as subjects treated with conventionals alone (adjusted RR=1.00, 95% CI 0.48 to 2.08 and 0.94, 95% CI 0.44 to 2.03) and a somewhat higher proportion of subjects treated with both types of antipsychotic at baseline developed TD as subjects treated with conventionals alone (adjusted RR=1.36, 95% CI 0.52 to 3.52 and 1.31, 95% CI 0.50 to 3.41).

AIMS total scores were slightly lower among incident cases appearing after recent atypical only exposure than recent conventional only exposure (mean difference −0.5, 95% CI -1.7 to 0.8). Analyses employing single-visit Glazer-Morgenstern criteria or consecutive-visit Schooler-Kane criteria as alternate definitions of incident caseness produced adjusted RR estimates similar to those for the primary analysis (0.69, 95% CI 0.35 to 1.36 and 0.56, 95% CI 0.21 to 1.48, respectively).

These analyses included clozapine-treated cases in the atypical antipsychotic group, our original intention. Crude analyses showed that 7 incident TD cases occurred among 55 at-risk subjects receiving clozapine who were followed an average of 17.8 months (81.6 person-years, crude TD rate 0.086/year). These results include 5 incident TD cases occurred among 23 at-risk subjects receiving clozapine as their sole antipsychotic followed an average of 23.2 months (44.4 person-years, crude TD rate 0.111/year). Because crude TD rates among clozapine-treated at-risk cases were unexpectedly high, clozapine-treated cases were removed into a separate category. These analyses are shown in Table 2.

Table 2
Crude (unadjusted) and adjusted estimated effects (rate-ratio, RR; and 95% CI) of antipsychotic type and other covariates in the core model on the incidence rate of persistent tardive dyskinesia.

Based on adjusted results from the core model, subjects treated with atypical antipsychotics alone (excluding clozapine) since the prior visit developed TD at slightly over half the rate as subjects treated with conventionals alone (adjusted RR=0.55, 95% CI 0.23 to 1.36). The adjusted RR for combined AT and CV antipsychotics (excluding clozapine) was 2.21 (95% CI 0.85 to 5.80).

Based on adjusted results, the TD incidence rate was also associated with age and years of previous conventional antipsychotic use (Table 2). Being African American was only weakly associated with TD, and the association with recent antipsychotic dose was not monotonic. None of the remaining planned covariates appear to have appreciably confounded individual antipsychotic effects.

The adjusted RR for atypicals vs conventionals (0.55) was lower than the crude RR (0.94) in Table 2 due to apparent confounding by years of conventional antipsychotic exposure. Subjects receiving atypicals had shorter durations of prior (TD-free) conventional exposure than subjects receiving conventionals (Table 1), and shorter durations of prior TD-free conventional exposure were associated with a higher rate of TD (Table 2), as in our prior report.4 Thus adjusting for this confounder lowered the RR for atypicals. When this variable was omitted from the core model, the otherwise-adjusted RR for atypical antipsychotic was very similar to the unadjusted RR. The relative effect of atypicals vs conventionals did not appear confounded by years of previous lifetime atypical exposure when this variable was added to the model. None of the remaining planned covariates, including gender, appreciably changed antipsychotic effects on TD incidence after adjustment for core model variables.

The overall modest advantage of atypical antipsychotics (excluding clozapine) since the prior visit on TD incidence was stronger among affective disorder subjects (RR=0.15, 95% CI 0.03 to 0.71) than among schizophrenia subjects (RR=0.97, 95% CI 0.31 to 3.04, p for interaction term 0.050). The stronger advantage among affective patients appeared partly due to lower risk among affective patients exposed to atypicals (crude rate 0.028/year, 4 cases in 142.5 person-years) and partly due to higher risk among affective patients exposed to conventionals (crude rate 0.100/year, 3 cases in 29.9 person-years). None of the remaining variables appreciably modified the estimated atypical antipsychotic effect on TD incidence.

Results for individual atypical antipsychotics are shown in Table 3. Estimates were imprecise in this analysis, since fewer subjects had received one atypical only for the entire time since the prior visit.

Table 3
Estimated crude and adjusted rate-ratio (RR, 95% CI) for tardive dyskinesia comparing each atypical antipsychotic with conventional antipsychotic since the prior visit.

Estimated Effects of Duration of Antipsychotic Use by Type

Only 26 subjects were naïve to conventional exposure over their lifetimes. Among these, two incident cases of persistent TD appeared during 51.3 years of atypical exposure (crude rate 0.039/year). Comparing that rate to the crude rate for conventionals only yields a crude rate-ratio of 0.69 (95% CI 0.15 to 3.25).

Models focusing on lifetime duration of antipsychotic use at the current visit by drug type revealed estimated antipsychotic effects similar to those from the recent exposure analyses (available on request from SWW).

Loss to Follow-Up

Among 401 TD-free subjects enrolled, 49 (12%) were never re-examined and 133 (33%) withdrew sometime later during follow-up before developing TD. Analyses omitting the partial data for dropouts produced results that were not appreciably different from the primary analysis.

Literature Review

We identified nine previous studies in adult patients that reported TD incidence with atypical compared to conventional antipsychotics.3341 Table 4 shows the mean age, design, TD acquisition methods, baseline TD prevalence, average antipsychotic doses, incident cases, mean follow-up time, patient-years of exposure, annual incidences, and the atypical/conventional RRs and 95% CIs for each study, as well as for the nine studies taken together and the present study. The 7 randomized and 2 cohort studies together reported 123 incident cases of TD among 2287 patient years of conventional antipsychotic exposure (0.085/year) and 247 incident cases of TD among 12,018 patient years of atypical antipsychotic exposure (0.031/year, RR 0.24, 95% CI 0.12 to 0.48). About three-quarters of the exposure time was accounted for by the two large cohort studies.37, 38 The present study reports on more atypical antipsychotic exposure time than any of the previous studies except the two previous cohort studies and one of the seven randomized studies.36 Average length of follow-up in the previous studies was about one-third that for the present study.

Table 4
Previous studies comparing newly identified tardive dyskinesia in atypical- and conventional-treated groups, compared to current study.


The major finding of this study is that the incidence of TD with recent exposure to atypical antipsychotics alone at our CMHC was more similar to that for conventional antipsychotics than in 8/9 previous studies. Taken together, the previous studies suggest that the risk of TD with atypicals is one-quarter that of conventionals (Table 4); our findings suggest the risk with atypicals is more than half that of conventionals (when clozapine patients are excluded, Table 2) or more than two-thirds the risk (when clozapine patients are included). Furthermore, our adjusted TD rate-ratio of 0.97 among schizophrenia patients suggests less of an advantage for atypicals than reported in any of the nine previous studies (all of schizophrenia patients) in Table 4.

Strengths and Limitations

Methodologic strengths of this study include the prospective cohort design with multiple years of follow-up, careful screening for previous and current TD symptoms at baseline, systematic identification of new TD cases periodically during follow-up, careful compilation of medication histories, and appropriate multivariable analysis that controlled for multiple potential confounders and treated type and dose of antipsychotic medications as time-dependent covariates.

The major limitation of our study is that nearly all of our CMHC subjects had lifetime histories of conventional antipsychotic exposure, often extensive and most of it occurring before baseline examination. It is possible that prior conventional antipsychotic use could sensitize patients subsequently receiving atypicals to be at higher risk than if they had been conventional-naïve. In addition very few patients in our study were exposed to only one antipsychotic over their lifetime, which also complicates interpretation in attributing newly emergent TD to current medication vs possible lingering effects of previous treatment. These limitations are not unique to our study but are characteristic of most other modern attempts to estimate differential risks of TD with conventional and atypical antipsychotic drugs42 including at least 5/7 of the nine previous comparative TD incidence studies found by out review.33, 34, 3638 The two recently reported first episode risperidone vs haloperidol analyses are exceptions.39, 41

A second important limitation of the present study is our use of a cohort design. This design carries the advantage of not artificially requiring treatment change at the start of the study but can lead to imbalances in the treated groups, such as the markedly shorter observed median exposure to conventional antipsychotics among persons currently treated with atypical agents (3.6 years) vs. those currently treated with conventionals (12.9 years, Table 1) and the far shorter prior exposure to atypical agents among those currently given conventional antipsychotics (0.1 vs. 3.0 years, Table 1). Our analyses, however, adjusted for lifetime duration of conventional antipsychotic use, as well as other measured potentially confounding variables such as gender, anticholinergic use, and negative symptoms. One might speculate that our conventional-treated cohort had been selected by prescribers to remain on conventional antipsychotic based on some unmeasured protective factor for which we cannot adjust. For such selection to account for our findings, however, we would expect a low crude TD incidence rate among our conventional-treated patients, and the observed rate of 0.056/year (Table 2) was not unexpectedly low.4, 43

Another limitation of our study is that we lost 45% of our initial cohort during the four-year follow-up. Examination of sample size and follow-up length data in Table 4 suggests, however, that the differences between our findings and those of most previous studies are unlikely to be explained by differences in follow-up time. Measured differences between conventional-treated dropouts and atypical-treated dropouts are unlikely to have biased our findings since we adjusted for these variables.

Lastly, the relatively high rate of emerging TD we observed among clozapine-treated patients was surprising given our expectations that clozapine would be associated with minimal risk of TD. Important caveats are that we estimated the RR for clozapine alone very imprecisely (7 cases among 55 patients exposed to clozapine with or without concomitant other antipsychotic, Table 2, and and55 cases among 23 exposed to clozapine alone). Because of the surprising findings, an investigator (JRS or SWW) thoroughly re-reviewed all available medical records for the 5 incident cases appearing in patients treated with clozapine alone. Previous history of TD despite negative baseline research examination did not appear to account for the high rate. Three of these cases had previously participated in the earlier Yale TD Incidence Study (no TD throughout). Two of five incident cases with clozapine alone did have a clinical history of TD on at least one examination, but in one case only on one examination of four recorded lifetime before clozapine and in the other only on 1/17 clinical examinations before clozapine. Neither of these positive clinical examinations was the last one before beginning clozapine. Still, it is possible that previous clinical or research examinations could have overlooked previous TD or that records reporting previous TD could have existed but been unavailable for our review.

Table 5
Comparison between tardive dyskinesia at CMHC in 1980s and 2000s.

It is worth mentioning that the expected minimal risk of TD with clozapine is supported by a surprisingly small direct incidence database. We are aware of only three studies, none of which are impressively larger than ours, only one of which unequivocally found very low risk.44 In this study, single study, two of 28 patients developed TD during an average 7.7 years of clozapine treatment, yielding a rate roughly one-tenth of ours in Table 2. Two other small studies, however, have reported clozapine findings similar to ours.17, 45 In one of these, 7 cases of dyskinesia emerged during roughly 4 year follow-up of 25 clozapine-treated patients (approximately 0.070/year).45 In the other study, a possibly increased crude TD risk with clozapine (among only 13 patients, however) was reduced when the model adjusted for response to treatment of the first episode.17 Unfortunately, we did not collect data permitting us to adjust for first episode treatment response. Thus it is not clear whether clozapine increased the risk for dyskinesia in our cases or whether our clozapine-treated cases were at greater risk for illness-related dyskinesia. Current use of clozapine could potentially also have been confounded by indication if it were prescribed because of earlier, and unmeasured, intolerable adverse neurological effects that could themselves have conferred an increased risk of TD.

Comparison with Previous Studies Comparing Atypicals and Conventionals

Table 4 shows that the differing relative risk in our study vs previous studies is accounted for by the previous studies finding a somewhat higher TD incidence rate with conventionals than we did (0.085 vs 0.056/year) and a somewhat lower incidence rate with atypicals (0.031 vs 0.059/year). The incidence rate we observed with conventionals is similar to those from large studies from the conventional era.4, 43 Previous studies in Table 4 reporting prevalence found substantial lower proportions of TD at baseline than we did (8.7% vs 31.5%).

The limitation of our study that nearly all our subjects had lifetime histories of conventional antipsychotic exposure is unlikely to account for differences between our findings and others, since most subjects had extensive prior conventional exposure histories in many previous studies that did find lower rates of TD with atypicals.33, 34, 37, 38 Our use of a cohort design is also unlikely to explain differences between our findings and others, since previous cohort studies37, 38 agree with most previous randomized studies in reporting a stronger TD advantage for atypicals than we do (Table 4). Similarly, many previous studies appear to have experienced higher attrition rates than ours, and none adjusted for the possibility of TD risk differing between dropouts from atypical versus conventional antipsychotic.

Likely explanations for the difference between our findings and those obtained previously relate to study-design features. The previous studies all articulated broad efficacy and safety aims and therefore did not focus substantial attention on training raters to detect TD accurately (Table 4). None of the previous studies report more than initial training for TD ratings, and no previous study reports TD interrater reliability data. In the absence of careful training and ongoing monitoring, two types of errors have been previously reported. First, true TD can be missed fairly often.46 Second, cases of extrapyramidal syndrome (EPS) movement such as jaw tremor, hand tremor, or leg restlessness can be misidentified as TD.4752 These two types of error could explain the pattern of findings among previous studies in Table 4: missed true TD in the previous studies could explain the low baseline prevalence and low incidence in the atypical-treated patients. Patients assigned to conventional antipsychotic could experience new EPS, which could be sometimes misidentified as TD, leading to higher than expected rates of “TD” in the conventional group. Among atypical-treated patients, misidentification of EPS as TD would not inflate the TD incidence rate to a similar degree, because these patients would be expected to be less likely to experience EPS that could be misidentified. The propensity of a study to falsely detect “TD” in conventional-treated patients (despite missing it at baseline) would be particularly high if the design called for forced antipsychotic change or initiation at entry (all previous studies), change to or initiation of high-EPS conventionals at entry,3335, 3841 or proscription or discouragement of anticholinergic medication after entry.3335 This propensity would also be magnified if movements emerging in the first 3 months after antipsychotic initiation or change were permitted to qualify as TD.3336, 3941

Table 4 shows substantial variability among the nine previous studies. Annualized incidence for conventional antipsychotic varied from 0.019/year to 0.228/year. Annualized incidence for atypical antipsychotic varied from 0.000/year to 0.206/year. Some of the variability may be methodological. The previous study with the highest incidence rates36 was the only one to report rates based on a meeting criteria on one occasion across as many as 7 follow-up time points. The other studies either required criteria be met twice on consecutive occasions or one occasion but at a single specified time point38 or at one of two follow-up time points.37 The study requiring only one occasion but at a single specified time point38 (12 months) also reported rates based on meeting criteria at any of three time points (3, 6, and 12 months); these rates were 2–3 times higher.

Comparison with Our Previous Study during the Conventional Era

Another finding of the present study is that overall TD prevalence, incidence, and incident case severity in the current cohort differed little from estimates obtained from a similar cohort studied at our site with similar methods before the introduction of atypical antipsychotics (Table 5). Other researchers have also reported persistence of substantial TD prevalence despite widespread atypical antipsychotic use.5357 One group58 has recently published evidence of a decline in TD prevalence from 31% during the conventional era to 10–12% during the atypical era; the studies compared used the same rating and training methods but were not conducted at the same sites.

Implications for Combination Prescribing

The incidence of TD with atypical and conventional antipsychotics in combination was somewhat higher than for conventional antipsychotics alone (Table 2). The average daily CPZ equivalent dose was strikingly higher in patients receiving combination prescribing (Table 1), but the association between combination prescribing and risk of TD was unchanged after adjusting for dose. Although combination prescribing is common,5966 TD risk with combination prescribing has not previously been studied to our knowledge. TD risk associated with this practice should be balanced against the infrequently studied likelihood of benefit.67, 68

Implications of Psychiatric Diagnosis

Little TD advantage for atypicals was apparent in schizophrenia subjects, while a relatively strong advantage was estimated in affective disorder subjects. Since numerous interactions were examined, and power was low for detecting them in this study, caution is indicated in interpreting these findings. We are not aware of other TD incidence data with atypicals relative to conventionals in affective disorder subjects.

Implications for Specific Atypical Antipsychotics

Little data were available for ziprasidone or aripiprazole. Among other atypical antipsychotics, olanzapine showed the lowest relative TD rate (Table 3). Confidence intervals in the present study for specific medications were wide, however. These findings do agree with some previous studies (Table 4). For example, in the other two studies that compared multiple atypicals to conventionals in Table 4, olanzapine had the lowest rate-ratio in both.36, 38 Additional studies comparing TD risk among atypical antipsychotics are needed.

Overall Risk of TD with Atypical Antipsychotics

While our findings differ from most previous TD studies from the atypical era, ours is the first incidence study to focus primary investigative attention on the TD question, and previous studies may have consistently been susceptible to ascertainment bias. Our findings suggest that the incidence rate of TD with atypical antipsychotics, while modestly reduced, remains substantial, at least in patients with prior conventional antipsychotic exposure who currently constitute the large majority of patients at our facility. Risk appeared little different among the few patients who were conventional-naïve. Future studies should investigate TD incidence in large samples with no conventional exposure history. Comparison of findings from the current study with those from our site prior to the atypical era reveal little impact on TD from a decade of increasing atypical antipsychotic prescription.

Despite the feeling among some clinicians that TD is much less of a problem now in the atypical era, such a conclusion may unfortunately be premature. In the 1960s and 70s there was some well-intentioned resistance and skepticism about conventionals being associated with risk of TD,69 and now during the atypical era we are perhaps not immune to some of the same forces. Until we are certain we have developed antipsychotics that carry minimal risk, we should continue to inform patients prescribed antipsychotics about TD and continue monitoring for it. Research efforts should continue to discover novel antipsychotics that are free of TD risk as well as to discover new treatments that can help patients who already have TD.


Supported by United States Public Health Service grant R01 MH61008 (SWW) from the National Institute of Mental Health. The Veteran’s Affairs New England Mental Illness Research Education and Clinical Center provided some assistance with data management. Presented at the Institute for Psychiatric Services meeting, San Diego CA, October 2005 and at International Congress on Schizophrenia Research, Colorado Springs CO, March 2007. Ramy Mahmoud, M.D., Jonathon Rabinowitz, Ph.D., and Diederick Tenback, M.D.gave permission to acknowledge personal communications about previous study methods.


Declaration of Interest

Dr. Woods acknowledges receipt of investigator-initiated grant support from Bristol Myers-Squibb, Eli Lilly, Janssen, and UCB Pharma and advisory board honoraria from Otsuka and Schering-Plough. Dr. Glazer acknowledges receipt of consultant fees from AstraZeneca, Eli Lilly, and Pfizer and speakers’ bureau honoraria from Eli Lilly.


1. Correll CU, Leucht S, Kane JM. Lower risk for tardive dyskinesia associated with second-generation antipsychotics: A systematic review of 1-year studies. Am J Psychiatry. 2004;161:414–425. [PubMed]
2. Woods SW. Olanzapine and tardive dyskinesia. Br J Psychiatry. 1999;175:391–392. [PubMed]
3. Morgenstern H, Glazer WM, Gibowski LD, et al. Predictors of tardive dyskinesia: Results of a cross sectional study in an outpatient population. J Chron Dis. 1987;40:319–327. [PubMed]
4. Morgenstern H, Glazer WM. Identifying risk factors for tardive dyskinesia among long-term outpatients maintained with neuroleptic medications. Results of the Yale Tardive Dyskinesia Study. Arch Gen Psychiatry. 1993;50:723–733. [PubMed]
5. Branch PR. Abnormal involuntary movement scale (AIMS) Early Clin Drug Eval Unit Intercom. 1975;4:3–6.
6. Glazer WM, Morgenstern H, Doucette JT. Predicting the long-term risk of tardive dyskinesia in outpatients maintained on neuroleptic medications. J Clin Psychiatry. 1993;54:133–139. [PubMed]
7. Schooler NR, Kane JM. Research diagnoses for tardive dyskinesia [letter] Arch Gen Psychiatry. 1982;39:486–487. [PubMed]
8. Cicchetti DV, Showalter D. A computer program for determining the reliability of dimensionally scaled data when the numbers and specific sets of examiners may vary at each assessment. Ed Psychol Meas. 1988;48:717–720.
9. Glazer WM, Moore DC. A tardive dyskinesia clinic in a mental health center. Hosp Community Psychiatry. 1981;32:572–574. [PubMed]
10. Am Psychiatric Association. Practice Guideline for the Treatment of Patients with Schizophrenia. Am J Psychiatry. 1997;154:1–63. [PubMed]
11. Woods SW. Chlorpromazine equivalent doses for the newer atypical antipsychotics. J Clin Psychiatry. 2003;64:663–667. [PubMed]
12. Nayak RK, Doose DR, Nair NP. The bioavailability and pharmacokinetics of oral and depot intramuscular haloperidol in schizophrenic patients. J Clin Pharmacol. 1987;27:144–150. [PubMed]
13. Schooler NR, Levine J. Initiation of long-term pharmacotherapy in schizophrenia -Dosage and side-effect comparisons between oral and depot fluphenazine. Pharmakopsychiatrie Neuro-Psychopharmakologie. 1976;9:159–169. [PubMed]
14. Chue P, Eerdekens M, Augustyns I, et al. Comparative efficacy and safety of long-acting risperidone and risperidone oral tablets. Eur Neuropsychopharmacol. 2005;15:111–117. [PubMed]
15. First MB, Spitzer RL, Gibbon M, et al. Structured Clinical Interview for DSM-IV Axis I Disorders, Patient Edition, January 1995 FINAL, (SCID-I/P Version 2.0) New York: Biometrics Research Department, New York State Psychiatric Institute; 1995.
16. Kane JM, Woerner M, Weinhold P, et al. Incidence of tardive dyskinesia - 5-year data from a prospective study. Psychopharmacol Bull. 1984;20:387–389. [PubMed]
17. Chakos MH, Alvir JM, Woerner MG, et al. Incidence and correlates of tardive dyskinesia in first episode of schizophrenia. Arch Gen Psychiatry. 1996;53:313–319. [PubMed]
18. Morgenstern H, Glazer WM, Doucette JT. Handedness and the risk of tardive dyskinesia. Biol Psychiatry. 1996;40:35–42. [PubMed]
19. Jeste DV, Caligiuri MP, Paulsen JS, et al. Risk of tardive dyskinesia in older patients: A prospective longitudinal study of 266 outpatients. Arch Gen Psychiatry. 1995;52:756–765. [PubMed]
20. Crane GE. Pseudoparkinsonism and tardive dyskinesia. Arch Neurology. 1972;27:426–430. [PubMed]
21. Chouinard G, Annable L, Rosschouinard A, et al. A 5-year prospective longitudinal study of tardive dyskinesia - Factors predicting appearance of new cases. J Clin Psychopharmacol. 1988;8:S21–S26. [PubMed]
22. Woerner MG, Alvir JMJ, Saltz BL, et al. Prospective study of tardive dyskinesia in the elderly: Rates and risk factors. Am J Psychiatry. 1998;155:1521–1528. [PubMed]
23. Sachdev P. Early extrapyramidal side-effects as risk factors for later tardive dyskinesia: a prospective study. Austral N Z J Psychiatry. 2004;38:445–449. [PubMed]
24. van Os J, Walsh E, van Horn E, et al. Changes in negative symptoms and the risk of tardive dyskinesia: a longitudinal study. UK700 Group Acta Psychiatr Scand. 2000;101:300–306. [PubMed]
25. Caligiuri MP, Lacro JP, Rockwell E, et al. Incidence and risk factors for severe tardive dyskinesia in older patients. Br J Psychiatry. 1997;171:148–153. [PubMed]
26. Campbell M, Armenteros JL, Malone RP, et al. Neuroleptic-related dyskinesias in autistic children: a prospective, longitudinal study. J Am Academy Child Adol Psychiatry. 1997;36:835–843. [PubMed]
27. Wegner JT, Kane JM, Weinhold P, et al. Cognitive Impairment in Tardive-Dyskinesia. Psychiatry Res. 1985;16:331–337. [PubMed]
28. Emsley R, Turner HJ, Oosthuizen PP, et al. Neurological abnormalities in first-episode schizophrenia: temporal stability and clinical and outcome correlates. Schizophr Res. 2005;75:35–44. [PubMed]
29. Pappadopulos E, Goldman R, Bates J, et al. Neuropsychological abnormalities precede the onset of tardive dyskinesia (abstract) J Int Neuropsychol Soc. 1997;3:3–4.
30. van Harten PN, Hoek HW, Matroos GE, et al. Evidence that lithium protects against tardive dyskinesia: the Curacao Extrapyramidal syndromes study VI. Eur Neuropsychopharmacol. 2008;18:152–155. [PubMed]
31. Diehl A, Reinhard I, Schmitt A, et al. Does the degree of smoking effect the severity of tardive dyskinesia? A longitudinal clinical trial. Eur Psychiatry. 2008 (epub ahead of print) [PubMed]
32. The Nordic Cochrane Centre. Review Manager (RevMan) [Computer program]. Version 5.0. Copenhagen: The Cochrane Collaboration; 2008.
33. Beasley CM, Dellva MA, Tamura RN, et al. Randomised double-blind comparison of the incidence of tardive dyskinesia in patients with schizophrenia during long-term treatment with olanzapine or haloperidol. Br J Psychiatry. 1999;174:23–30. [PubMed]
34. Csernansky JG, Mahmoud R, Brenner R, et al. A comparison of risperidone and haloperidol for the prevention of relapse in patients with schizophrenia. NEJM. 2002;346:16–22. [PubMed]
35. Dossenbach MRK, Folnegovic-Smalc V, Hotujac L, et al. Double-blind, randomized comparison of olanzapine versus fluphenazine in the long-term treatment of schizophrenia. Prog Neuropsychopharmacol Biol Psychiatry. 2004;28:311–318. [PubMed]
36. Lieberman JA, Stroup TS, McEvoy JP, et al. Effectiveness of antipsychotic drugs in patients with chronic schizophrenia. NEJM. 2005;353:1209–1223. [PubMed]
37. Tenback DE, van Harten PN, Slooff CJ, et al. Effects of antipsychotic treatment on tardive dyskinesia: a 6-month evaluation of patients from the Eur Schizophrneia Outpatient Health Outcomes (SOHO) study. J Clin Psychiatry. 2005;66:1130–1133. [PubMed]
38. Dossenbach M, Arango-Davila C, Ibara HS, et al. Response and relapse in patients with schizophrenia treated with olanzapine, quetiapine, risperidone, or haloperidol: 12-month follow-up of the Intercontinental Schizophrenia Outpatient Health Outcomes (IC-SOHO) Study. J Clin Psychiatry. 2005;66:1021–1030. [PubMed]
39. Gharabawi GM, Bossie CA, Zhu Y. New onset tardive dyskinesia in patients with first episode psychosis receiving risperidone or haloperidol. Am J Psychiatry. 2006;163:938–939. [PubMed]
40. Miller DD, Eudicone JM, Pikalov A, et al. Comparative assessment of the incidence and severity of tardive dyskinesia in patients receiving aripiprazole or haloperidol for the treatment of schizophrenia: a post hoc analysis. J Clin Psychiatry. 2007;68:1901–1906. [PubMed]
41. Gaebel W, Riesbeck M, Wolwer W, et al. Maintenance treatment with risperidone or low-dose haloperidol in first-episode schizophrenia: 1-Year results of a randomized controlled trial within the German research network on schizophrenia. J Clin Psychiatry. 2007;68:1763–1774. [PubMed]
42. Tarsy D, Baldessarini RJ. Epidemiology of tardive dyskinesia: Is risk declining with modern antipsychotics? Mov Disord. 2006;21:589–598. [PubMed]
43. Kane JM, Woerner M, Lieberman J. Tardive dyskinesia: prevalence, incidence, and risk factors. J Clin Psychopharmacol. 1988;8:52S–56S. [PubMed]
44. Kane JM, Woerner MG, Pollack S, et al. Does clozapine cause tardive dyskinesia? J Clin Psychiatry. 1993;54:327–330. [PubMed]
45. Bunker MT, Sommi RW, Stoner SC, et al. Longitudinal analysis of abnormal involuntary movements in long- term clozapine-treated patients. Psychopharmacol Bull. 1996;32:699–703. [PubMed]
46. Weiden PJ, Mann JJ, Haas G, et al. Clinical nonrecognition of neuroleptic-induced movement disorders: A cautionary study. Am J Psychiatry. 1987;144:1148–1153. [PubMed]
47. Hansen TE, Glazer WM, Moore DC. Tremor and tardive dyskinesia. J Clin Psychiatry. 1986;47:461–464. [PubMed]
48. Cummings JL, Wirshing WC. Recognition and differential diagnosis of tardive dyskinesia. Int J Psychiatry Med. 1989;19:133–144. [PubMed]
49. Munetz MR, Cornes CL. Distinguishing akathisia and tardive dyskinesia: a review of the literature. J Clin Psychopharmacol. 1983;3:343–350. [PubMed]
50. Sachdev P. Akathisia and Restless Legs. Cambridge, UK: Cambridge University Press; 1995.
51. Fahn S. Tardive dyskinesia may be only akathisia. NEJM. 1978;299:202–203. [PubMed]
52. Crane GE. Neuroleptics and their long-term side effects on the central nervous system. In: Bristol Devaugh-Geiss J., editor. Tardive Dyskinesia and Related Involuntary Movement Disorders. John Wright PSG, Inc; 1982. pp. 71–84.
53. Modestin J, Stephan PL, Erni T, et al. Prevalence of extrapyramidal syndromes in psychiatric inpatients and the relationship of clozapine treatment to tardive dyskinesia. Schizophr Res. 2000;42:223–230. [PubMed]
54. Halliday J, Farrington S, MacDonald S, et al. Nithsdale Schizophrenia Surveys 23: movement disorders. 20 year review. Br J Psychiatry. 2002;181:422–427. [PubMed]
55. Ross DE, Thomas M, Booth M, et al. Rate of tardive dyskinesia in hospitalized patients. Am J Psychiatry. 2005;162:816. [PubMed]
56. Marshall DL, Hazlet TK, Gardner JS, et al. Neuroleptic drug exposure and incidence of tardive dyskinesia: A records-based case-control study. J Manag Care Pharm. 2002;8:259–265. [PubMed]
57. de Leon J. The effect of atypical versus typical antipsychotics on tardive dyskinesia: a naturalistic study. Eur Arch Psychiatry Clin Neurosci. 2007;257:169–172. [PubMed]
58. Chouinard G, Chouinard VA. Atypical antipsychotics: CATIE study, drug-induced movement disorder and resulting iatrogenic psychiatric-like symptoms, supersensitivity rebound psychosis and withdrawal discontinuation syndromes. Psychother Psychosom. 2008;77:69–77. [PubMed]
59. Centorrino F, Goren JL, Hennen J, et al. Multiple versus single antipsychotic agents for hospitalized psychiatric patients: Case-control study of risks versus benefits. Am J Psychiatry. 2004;161:700–706. [PubMed]
60. Covell NH, Jackson CT, Evans AC, et al. Antipsychotic prescribing practices in Connecticut’s public mental health system: Rates of changing medications and prescribing styles. Schizophr Bull. 2002;28:17–29. [PubMed]
61. Ganguly R, Kotzan JA, Miller LS, et al. Prevalence, trends, and factors associated with antipsychotic polypharmacy among medicaid-eligible schizophrenia patients, 1998–2000. J Clin Psychiatry. 2004;65:1377–1388. [PubMed]
62. McCue RE, Waheed R, Urcuyo L. Polypharmacy in patients with schizophrenia. J Clin Psychiatry. 2003;64:984–989. [PubMed]
63. Stahl SM. Antipsychotic polypharmacy, Part 1: Therapeutic option or dirty little secret? J Clin Psychiatry. 1999;60:425–426. [PubMed]
64. Tapp A, Wood AE, Secrest L, et al. Combination antipsychotic therapy in clinical practice. Psychiatr Serv. 2003;54:55–59. [PubMed]
65. Waddington JL, Youssef HA, Kinsella A. Mortality in schizophrenia - Antipsychotic polypharmacy and absence of adjunctive anticholinergics over the course of a 10-year prospective study. Br J Psychiatry. 1998;173:325–329. [PubMed]
66. Wang PS, West JC, Tanielian T, et al. Recent patterns and predictors of antipsychotic medication regimens used to treat schizophrenia and other psychotic disorders. Schizophr Bull. 2000;26:451–457. [PubMed]
67. Stahl SM, Grady MM. A critical review of atypical antipsychotic utilization: Comparing monotherapy with polypharmacy and augmentation. Curr Med Chem. 2004;11:313–327. [PubMed]
68. Goff DC, Freudenreich O. Polypharmacy in schizophrenia: does anyone truly benefit? Int J Neuropsychopharmacol. 2004;7:109–111. [PubMed]
69. Kane JM. Tardive dyskinesia circa 2006. Am J Psychiatry. 2006;163:1316–1318. [PubMed]
70. Tollefson GD, Beasley CM, Tamura RN, et al. Blind, controlled, long-term study of the comparative incidence of treatment-emergent tardive dyskinesia with olanzapine or haloperidol. Am J Psychiatry. 1997;154:1248–1254. [PubMed]
71. Beasley CM. Author’s reply. Br J Psychiatry. 1999;175:392. [PubMed]
72. Mahmoud R. TD assessment procedures for Csernansky et al (NEJM 2002;346:16–22) personal communication. Jun 7, 2006.
73. Miller DD, McEvoy JP, Davis SM, et al. Clinical correlates of tardive dyskinesia in schizophrenia: Baseline data from the CATIE schizophrenia trial. Schizophr Res. 2005;80:33–43. [PubMed]
74. Haro JM, Edgell ET, Jones PB, et al. The Eur Schizophrenia Outpatient Health Outcomes (SOHO) study: Rationale, methods and recruitment. Acta Psychiatrica Scandinavica. 2003;107:222–232. [PubMed]
75. Haro JM, Edgell ET, Frewer P, et al. The Eur Schizophrenia Outpatient Health Outcomes Study: Baseline findings across country and treatment. Acta Psychiatrica Scandinavica. 2003;107:7–15. [PubMed]
76. Tenback DE. Proportion of patients with TD at 6 months who did not have TD at baseline. Personal communication re Tenback et al (J Clin Psychiatry 2005;66:1130–1133) to Woods, S.W. Oct 16, 2005.
77. Dossenbach M, Erol A, Kessaci ME, et al. Effectiveness of antipsychotic treatments for schizophrenia: Interim 6-month analysis from a prospective observational study (IC-SOHO) comparing olanzapine, quetiapine, risperidone, and haloperidol. J Clin Psychiatry. 2004;65:312–321. [PubMed]
78. Kasper S, Lowry AJ, Hodge A, et al. Tardive dyskinesia: Analysis of outpatients with schizophrenia from Africa and the Middle East, Asia, Central and Eastern Europe, and Latin America. Schizophr Res. 2006;81:139–143. [PubMed]
79. Schooler N, Rabinowitz J, Davidson M, et al. Risperidone and haloperidol in first-episode psychosis: A long-term randomized trial. Am J Psychiatry. 2005;162:947–953. [PubMed]
80. Rabinowitz J. TD rater training procedures for Schooler et al (Am J Psychiatry 2005;162,947–953) personal communication to Woods, S.W. Jun 15, 2006.