Do neuroleptic drugs hasten cognitive decline in dementia? Prospective study with necropsy follow up

BMJ 1997; 314 doi: (Published 25 January 1997) Cite this as: BMJ 1997;314:266
  1. Rupert McShane, clinical lecturera,
  2. Janet Keene, research assistanta,
  3. Kathy Gedling, research assistanta,
  4. Christopher Fairburn, professorb,
  5. Robin Jacoby, clinical readera,
  6. Tony Hope, readera
  1. a Section of Old Age Psychiatry, University Department of Psychiatry, Warneford Hospital, Oxford OX3 7JX
  2. b University Department of Psychiatry, Warneford Hospital, Oxford, OX3 7JX
  1. Correspondence to: Dr McShane
  • Accepted 31 October 1996


Objective: To investigate the contribution of neuroleptic drugs to cognitive decline in dementia.

Design: Two year prospective, longitudinal study consisting of interviews every four months, with necropsy follow up.

Setting: Community settings in Oxfordshire.

Subjects: 71 subjects with dementia, initially living at home with informant.

Main outcome measures: Cognitive function (score from expanded minimental state examination); behavioural problems (physical aggression, hallucinations, persecutory ideas, and disturbance of diurnal rhythm); and postmortem neuropathological assessment (cortical Lewy body pathology).

Results: The mean (SE) decline in cognitive score in the 16 patients who took neuroleptics was twice that in the patients who did not (20.7 (2.9) v 9.3 (1.3), P=0.002). An increased rate of decline was also associated with aggression, disturbed diurnal rhythm, and persecutory ideas. However, only use of neuroleptics and severity of persecutory ideas were independently associated with more rapid cognitive decline when all other variables were adjusted for. The start of neuroleptic treatment coincided with more rapid cognitive decline: median rate of decline was 5 (interquartile range 8.5) points per year before treatment and 11 (12) points per year after treatment (P=0.02). Cortical Lewy body pathology did not account for association between neuroleptic use and more rapid decline.

Conclusions: Neuroleptic drugs that are sometimes used to treat behavioural complications of dementia may worsen already poor cognitive function. Randomised controlled trials are needed to confirm a causal relation.

Key messages

  • Neuroleptic drugs are used to treat behavioural problems in patients with dementia, but they may cause more rapid decline in cognitive function

  • In our longitudinal study of patients with dementia we found that the rate of cognitive decline in patients taking neuroleptics was twice that in those not taking neuroleptics

  • Furthermore, the start of neuroleptic treatment was associated with an increase in the rate of cognitive decline

  • Cortical Lewy body pathology at necropsy did not account for the association between neuroleptic use and more rapid decline

  • Although our study does not prove a causal relation, we suggest that there should be regular review of the need for patients to continue taking neuroleptic drugs


Neuroleptic drugs can be of modest benefit in the treatment of behavioural problems in dementia.1 2 No other class of drugs is of superior efficacy1–indeed, only one adequate trial has shown a beneficial effect with a non-neuroleptic drug.3 4 5 The frequency of use of neuroleptics has varied widely,6 7 suggesting that differences in institutional practice contribute substantially to prescribing decisions. In the United States, anxiety about the overuse of neuroleptics in nursing homes has led to federal legislation prohibiting their use for behaviour such as wandering, restlessness, insomnia, or unspecified agitation in the absence of other justifying reasons.8 A recent survey of nursing homes in Glasgow found that 88% of patients taking neuroleptics might be receiving them inappropriately according to the American guidelines.9

The side effects of neuroleptics include the worsening of behavioural disturbance of patients with dementia,10 11 12 falls,13 and fractures.14 The possibility that the cognitive function of patients with dementia may be made worse by neuroleptics has received little attention. Cross sectional surveys have shown that patients with dementia who take neuroleptics have worse cognitive function than those who do not.15 However, this might be because neuroleptics are often prescribed for behavioural problems that are more common in patients with severe dementia. Results from studies comparing cognitive function before and after the use of neuroleptics have been conflicting. Some studies have not shown any effect on global assessments,10 16 17 others have found improvements in memory or orientation,18 19 20 21 22 and some have shown a worsening of cognitive function.11 23

None of these studies examined the long term effects of neuroleptics in dementia or examined the extent to which the effects on cognition were independent of other factors that might affect cognitive function. A more rapid decline in cognitive function has been associated with psychosis,24 25 26 27 hallucinations,28 29 30 31 sleep disturbance, and aggression,31 but no study has adequately controlled for the possibility that this decline may have been due to the neuroleptics used to treat such behaviours.

Subjects and methods


Through general practitioners and community psychiatric nurses, we recruited 104 subjects with dementia (Diagnostic and Statistical Manual of Mental Disorders, third edition, revised) who were living at home with carers who were able to give a good account of their symptoms. They were assessed every four months. Six subjects dropped out after two to five interviews and have been excluded from the analyses. Another 27 patients died within 20 months of entry to the study. The remaining 71 patients were assessed on at least six occasions and form the sample for the first multiple regression analysis reported below. Of these 71 patients, 42 underwent postmortem neuropathological assessment and were the sample for the second regression analysis.


The patients were interviewed every four months and their behaviour assessed with the present behavioural examination.32 Physical aggression, hallucinations, and persecutory ideas were rated on a seven point frequency scale, relating to the number of days over the four weeks immediately before assessment on which the symptom had occurred. Disturbance of diurnal rhythm that influenced behaviour was rated as present or absent. Non-cognitive symptoms that were clearly related to temporary physical illness were rated as “missing” throughout the study. Cognitive function was assessed with an expanded version of the minimental state examination.33

Current medication was recorded at each interview. All doses of neuroleptics were converted to a “chlorpromazine equivalent.”34 In view of the possible adverse effect of low doses in a subgroup of patients35 we set a low threshold for the definition of substantial neuroleptic use: a mean of 10 mg of chlorpromazine daily over the first 20 months. Our study was conducted before substituted benzamides such as sulpiride were widely used in Britain for treating elderly people. Four drugs accounted for nearly all the neuroleptics prescribed–thioridazine, promazine, haloperidol, and chlorpromazine.

Statistical analysis

The dependent variable in the two multiple regression analyses was the rate of cognitive decline over the first six interviews (that is, the coefficient of regression of minimental state examination score by interview number). In the first analysis, of 71 subjects, the independent variables age, sex, duration of dementia, and cognitive score at study entry were entered simultaneously in a first block. Neuroleptic use and the mean scores over the six interviews for hallucinations, persecutory ideas, physical aggression, and disturbance in diurnal rhythm were then entered stepwise. We repeated the analysis for the group of 42 patients for whom necropsy data were available, with the addition to the first block of a dichotomous variable indicating the presence or absence of cortical Lewy body pathology.

For the purposes of examining the temporal relation between the start of neuroleptic use and the rate of cognitive decline, we performed a third analysis. We identified 20 patients from the total sample of 98 who were not taking neuroleptics at entry to the study but who started them during the study and continued to take them for at least four months and who had not reached the lowest levels of the expanded minimental state examination when they started taking neuroleptics. We compared the decline in their scores over the year before they started taking neuroleptics with the decline in the year after.

We also compared the 20 subjects' decline in the year after starting neuroleptics with that of 20 matched control patients, drawn from the same sample, who never took neuroleptics. We identified each control on the basis that he or she had the same cognitive score (within four points) as that of the matching subject at the interview immediately before starting neuroleptic treatment. In order to reduce the risk of biased selection of those controls who had a larger number of interviews because they survived longer (a factor which was likely to be associated with a slower rate of cognitive decline), the controls were required to have an equivalent cognitive score at the same point after entry to the study as their matching case. We assessed the significance of these comparisons with unpaired tests (Wilcoxon rank sum) because there were some cases who were not followed for the whole two year period–either because they died within a year of starting neuroleptic treatment or because they started treatment less than a year after entering the study.



Of the 71 patients in our first analysis, 37 were men. The mean (SE) age at entry to the was 72.6 (0.7) years. The age of onset of dementia was less than 65 years in 11 patients, and the mean duration of dementia at study entry was 5.7 (0.5) years. The mean minimental state examination score at study entry was 15.5 (0.9), while the mean score for the expanded version of the examination was 33 (1.6). Of the 42 patients who underwent necropsy, seven had cortical Lewy body pathology and 10 took neuroleptics at some point.

Rate of cognitive decline

The decline in the mean (SE) score for the expanded examination in the patients who took neuroleptics was twice that in those patients who did not (20.7 (2.9) v 9.3 (1.3), P=0.002). An increased rate of decline was also associated with aggression, disturbance of diurnal rhythm, and persecutory ideas (table 1). Furthermore, the patients who were prescribed neuroleptics were more aggressive and had more severe disturbance of diurnal rhythm (table 2). Table 3 shows the extent of correlation between the severity of behavioural changes and other factors that might have predicted a more rapid rate of cognitive decline. There was no correlation between the rate of cognitive decline and the dose of neuroleptic (r =0.19, P=0.4).

Table 1

Relation of individual behavioural symptoms to rate of cognitive decline in patients with dementia

View this table:
Table 2

Mean (SE) values of possible determinants for use of neuroleptic drugs in 71 patients with dementia

View this table:
Table 3

Correlation matrix of factors potentially influencing rate of cognitive decline in 71 patients with dementia (values are Pearson r coefficients)

View this table:

None of the variables that were entered at the first step of the multiple regression analysis (age, sex, duration of dementia, and initial cognitive score), either taken together or separately, made a significant contribution to cognitive decline. Similarly, hallucinations, disturbance of diurnal rhythm, and aggression were excluded from the final model. The final model shows that only neuroleptic treatment and the severity of persecutory ideas were associated with increased rate of cognitive decline once the effect of all the other variables was taken into account (table 4). The significance of these variables as independent predictors was unaffected by square root transformation. The amount of the explained variance in cognitive decline that was attributable to the use of neuroleptics was 19.9%.

Table 4

Results of stepwise multiple linear regression analysis: final model for predicting rate of cognitive decline in patients with dementia

View this table:

We repeated the analysis for the 42 subjects who underwent necropsy. Cortical Lewy body pathology did not make an independent contribution to more rapid decline. Neuroleptic treatment remained a significant independent predictor of cognitive decline (data not shown).

Rate of decline before and after taking neuroleptics

Figure 1 shows the pattern of cognitive decline in the 20 patients who started taking neuroleptics after study entry and the pattern of decline in their 20 matched controls. Among the 20 cases, the median (interquartile range) score for cognitive function at the interview before neuroleptic treatment was started was 28.5 (18.5). The median rate of decline in the score over the year before the start of neuroleptic treatment was 5 (8.5) points per year in the 10 cases for whom this information was available. In the 17 cases who were followed for at least a year after neuroleptics were started the median rate of decline was more than twice as fast at 11 (12) points per year. This difference was significant (W=95, z=-2.3, P=0.02). The median decline in scores in the matched controls over the equivalent period was 7.5 (9). This was also significantly lower than the rate of cognitive decline in those taking neuroleptics (W=377, z=-2.3, P=0.02).

Fig 1
Fig 1

Relation between start of neuroleptic treatment and cognitive function (mean score of expanded minimental state examination) in patients with dementia


The main finding in this study was that the use of neuroleptic drugs was associated with an increased rate of cognitive decline in dementia. This association was independent of the degree of dementia and of the behavioural symptoms for which the neuroleptics might have been prescribed. This finding does not, of itself, exclude the possibility that the neuroleptics seemed to cause more rapid decline because patients who were already on a steeper trajectory of cognitive decline were more likely to be prescribed them for reasons other than behavioural problems. However, we also found that the point when patients started treatment with neuroleptics coincided with an increase in their rate of cognitive decline. In addition, we reproduced the findings of others that the rate of cognitive decline is greater in patients who have persecutory ideas.24 26 27 30

Robustness of study

One of the main strengths of our study is that it was longitudinal in design and that valid, reliable instruments were used for the collection of behavioural and cognitive data. It was also possible to compare the rates of decline of patients starting neuroleptic treatment over a much longer period than in conventional drug trials. A third advantage was that neuropathological diagnosis was available in over half of our subjects.

Several methodological weaknesses need to be addressed. First, this study is not a randomised controlled trial of the effect of neuroleptics on cognitive decline. Second, subjects' extrapyramidal function was not assessed. The effect of neuroleptics may have been mediated in some way by impaired extrapyramidal function. Third, our sample may have been less homogeneous than others which were confined to patients with clinical Alzheimer's disease. However, most of our subjects had clinical Alzheimer's disease and most of those with clinical vascular dementia also had Alzheimer's disease at autopsy (data not shown), a finding that concurs with that of Galasko et al.36 Fourth, the number of subjects whose cognitive decline was examined before and after starting neuroleptic treatment was small. This militated against the application of statistical tests to multiple time points. The number of statistical tests applied in this analysis was therefore limited by examining only two time points, a year before and a year after the start of treatment, and conventional levels of significance were applied.

Mechanisms for cognitive decline

Our results do not support the idea that the effect of neuroleptics on cognitive function is confined to a subgroup of patients with cortical Lewy body pathology who are particularly sensitive to the drugs.35 There are several other possible mechanisms by which neuroleptics might contribute to cognitive decline. The most parsimonious explanation is that the anticholinergic effect of the neuroleptics resulted in reduced attention.12 37 A second possibility is that both the faster cognitive decline and the persecutory ideas were due to unidentified episodes of delirium. Changes in cognitive function might also have been mediated by changes associated with chronic use of neuroleptics, such as enlargement of the caudate nucleus,38 39 40 or an increased susceptibility to develop the neurofibrillary change of Alzheimer's disease41 or impairment of compensatory neurotransmitter responses to neuronal degeneration.42

Our results support the hypothesis that neuroleptics contribute to cognitive decline in people with dementia but do not prove it. Randomised controlled trials are necessary to confirm a causal link. We believe that an appropriate response at present would be to undertake regular review of the need for patients to continue taking neuroleptic drugs, pursuing trials without medication where possible. This study highlights the importance of understanding the neurological basis of behavioural changes in dementia43 so that less toxic drugs can be developed for their treatment.


We thank Dr Paul Griffiths for statistical advice and Sandra Cooper for administrative help.

Funding: This work was supported by a Medical Research Council project grant to TH and CF.

Conflict of interest: None.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
View Abstract