Abstract
There have been no systematic investigations of the effects of glucocorticoid treatment on memory in a clinical population despite experimental and clinical evidence that such treatment could cause memory disturbance.We conducted both cross-sectional and longitudinal studies. In Study 1, we administered tests of both hippocampal-dependent explicit memory and hippocampal-independent implicit memory to twenty-five prednisone-treated patients with systemic disease without CNS involvement and 25 matched clinical controls. All treated patients were taking doses of 5 to 40 mg of prednisone daily for at least 1 year. The glucocorticoid-treated group performed worse than the controls on tests of explicit memory, but the groups did not differ on the implicit memory task. Multiple regression analyses suggested that elderly patients are more susceptible to memory impairment with less protracted treatment. The results of Study 2, a prospective, longitudinal study of the effects of prednisone on memory across 3 months of therapy, suggest that even acute treatment can adversely affect memory. The observed alteration in memory was not secondary to inattention, affective disturbance, generalized global cognitive decline, or severity of disease. Results reported here, combined with previous clinical and experimental reports, indicate that the risk of memory impairment should be carefully considered before initiating treatment with glucocorticoids. Conversely, use of glucocorticoids should be considered in the differential diagnosis of memory loss. Finally, the potential benefit of anti-inflammatory treatment in Alzheimer's disease might be counterbalanced by possible iatrogenic memory impairment, at least when synthetic glucocorticoids are considered.
NEUROLOGY 1996;47: 1396-1402
Despite abundant evidence that both low [1] and high [2,3] levels of glucocorticoids adversely affect hippocampal physiology, and that the hippocampal structures are essential to certain types of memory processes, [4] the effects of glucocorticoid therapy on memory have not been systematically investigated in a clinical population. There is well-documented evidence for the role of the medial temporal lobes [5] and the selectivity of the hippocampus itself in supporting explicit memory in humans or nonhuman primates. [6,7] Subsequent studies have led to a widely accepted distinction between hippocampal-dependent explicit memory and hippocampal-independent implicit memory. [8,9]
Prolonged exposure to high concentrations of glucocorticoids can site-specifically accelerate pathologic changes in the hippocampus [10-12] and may impair spatial memory in rats. [13,14] Even brief exposure to glucocorticoids can diminish feedback sensitivity to the hypothalamic-pituitary-adrenal axis, leading to higher circulating glucocorticoid concentrations and potential hippocampal damage. [15] These effects are probably not a direct consequence of glucocorticoid exposure but rather the result of a number of possible cellular insults, including inhibition of glucose transport into hippocampal neurons and glia, [16] the potentiation of excitatory amino acid killing of hippocampal neurons, [17] and, most pertinent to the effect on memory, the modulation of both hippocampal long term potentiation and primed burst potentiation by adrenal steroids. [18,19]
Because the hippocampus is one of the essential neuroanatomic substrates for memory, glucocorticoid-induced adverse affects on hippocampal structure and function may interfere with those memory processes dependent on hippocampal integrity. There is sufficient evidence from clinical studies to suggest that such a memory impairment may result from both endogenous and exogenous hypercortisolemia. Memory impairment is evident in patients receiving high-dose steroid therapy for various medical problems, [20] and in patients with Cushing's syndrome, [21] presumably a result of spontaneous over-production of corticosteroids. Additionally, there is an association between elevated cortisol levels, memory dysfunction, and reduced hippocampal volume (as assessed with MRI) in patients with Cushing's syndrome. [22] A number of experimental studies support these clinical findings. Wolkowitz et al. [23] first described transient memory deficits in the form of increased intrusion errors on a word-list recall task in both normal and depressed adults after dexamethasone administration. Newcomer et al. [24] reported decreased performance on tasks of explicit memory in normal controls after low-dose dexamethasone for 4 consecutive days, and a replication study yielded evidence for a biphasic dose-response curve. [25] Increases in cortisol secretion over years is associated with declining verbal [26] and nonverbal [27] memory in the elderly as well as a stronger cortisol response and declining memory after a psychological stressor. [28]
No previous studies, however, have investigated the effects on both hippocampal-dependent and -independent memory of long-term treatment with high-dose glucocorticoids in a non-CNS involved clinical population. Accordingly, we conducted the following studies to determine if either long-term or acute prednisone treatment of patients with systemic disease without CNS involvement was associated with impaired memory, and to what degree age, dose, or duration of treatment contributed to any impairment noted.
Study 1.
Methods.
Plan of design.
The study was a dose-controlled cross-sectional design comparing performance of prednisone-treated patients without CNS involvement on tests of memory with that of closely matched medical controls.
Subjects.
Both experimental and control subjects were recruited by referral from the rheumatology and neurology clinics of two major hospitals in the metropolitan Detroit area. Research assistants, with the assistance of the primary attending physician, attended the clinics to screen scheduled clinic patients for possible inclusion. Inclusion criteria for the experimental group consisted of long-term glucocorticoid use (>or=to5 mg qd) for at least 1 year for medical conditions without CNS involvement. Control subjects were matched to the experimental group on age, sex, IQ, and, as closely as possible, on diagnosis. Exclusion criteria, targeting confounds to the interpretation of results, included respiratory conditions causing documented or clinically recognized hypoxia; active treatment with any compound having documented effects on cognitive performance; [29,30] pregnancy or high-dose estrogens; Cushing's syndrome; body weight <80% ideal body weight; hepatic enzyme induction; uncontrolled diabetes mellitus; major physical illness other than the primary illness under study, trauma, fever, dehydration, or nausea within the past 2 weeks; temporal lobe epilepsy; narcotic therapy, active sedative-hypnotic withdrawal, or high-dose benzodiazepine therapy; Addison's disease, hypopituitarism, other endocrine disease; spironolactone therapy; [31] excessive caffeine use; [32] carbamazepine therapy; [33] alcohol or drug abuse; [34] psychosis, anxiety disorder, major affective disorder as assessed with the Structured Clinical Interview for DSM III-R, [35] or any condition likely to affect brain function.
Of the 108 experimental and control patients who expressed interest in the study, 51 were excluded after either a review of their records or an initial telephone contact. Fifty-seven subjects were seen for initial testing. Of those, two patients met criteria for current depressive episode and were referred for psychiatric treatment, three had visual disturbances secondary to diabetes or optic neuritis, one was taking high-dosage estrogen, and one was taking too low a prednisone dosage (5 mg qod). A total of 25 experimental and 25 control subjects were enrolled and completed the study protocol. The experimental group consisted of 10 patients with rheumatoid arthritis, 7 with scleroderma, 6 with myasthenia gravis, 1 with polymyositis, and 1 with chronic autoimmune hepatitis. The control group consisted of 7 patients with rheumatoid arthritis, 15 with scleroderma, 1 with chronic pain, 1 with Raynaud's phenomenon, and 1 with CREST syndrome. All subjects gave written informed consent and were paid $15.00 as compensation for participation.
Materials.
Rationale for choice of neuropsychological battery. List learning and paragraph recall tasks were included in the study because clinical observation has frequently revealed discrepancies in performance on these two tasks. Both are tests of explicit memory and are sensitive to hippocampal damage. The Word Stem Priming test, a test of implicit memory, assesses memory abilities not dependent on the integrity of the hippocampus and was necessary to identify any dissociation between explicit and implicit memory. The Vigilance Test of sustained auditory attention and the WAIS-R Digit Span test were used to assess for underlying attention deficits that could account for any observed memory impairment. Two tests of general cognitive abilities, the North American Adult Reading Test and the Vocabulary subtest of the WAIS-R, were included to obtain estimates of IQ and for the purpose of matching. The Structured Clinical Interview for DSM III-R was administered to determine the presence of any Axis I psychopathology that would serve as exclusion criteria. Since corticosteroids can induce affective symptoms, measures of depression and mania were administered to determine if any impairment in memory was secondary to differences between groups on these measures.
A number of tasks were used as time fillers during the memory testing delay period and did not address the primary hypothesis. These supernumerary tests included Line Orientation, [36] a visual matching test of angled lines; FAS, [37] a test of verbal fluency; Spatial Recall, [38] a nonverbal memory task; and Raven's Progressive Matrices, [39] a nonverbal test of general intellectual ability. There were no significant group differences on any of these variables.
Neuropsychological battery.
Tests of hippocampal-dependent explicit memory: Wechsler Memory Scale Paragraph Recall [40]; California Verbal Learning Test (CVLT). [41]
Test of hippocampal-independent implicit memory: Word Stem Completion Priming Task. [42]
Test of auditory attention: The Vigilance Test. [43]
Tests of general cognitive ability: North American Adult Reading Test [44]; WAIS-R Vocabulary. [45]
Affective measures: Hamilton Psychiatric Rating Scale for Depression [46]; Young Manic Rating Scale [47]; Structured Clinical Interview for DSM III-R (SCID). [35]
Disability measures: The Disability Inventory Questionnaire is a Likert scale developed in our clinic to assess objective and subjective aspects of illness. All items were subjected to item-total analyses. The final questionnaire included items assessing activities of daily living (i.e., driving, bathing, shopping), medication use, changes in activities (socializing, traveling, working), and medical information (use of cane, walker, hospitalizations).
Statistical analyses.
Differences between the experimental and control groups on two affective and six cognitive variables were assessed using two-tailed Student's t tests. Responses for the Disability Inventory Questionnaire were analyzed using Mann-Whitney U tests for nonparametric data. Pearson product moment correlations were used to determine the association between degree of mania or depression and scores on the cognitive tests. To determine if any drug effect on paragraph recall was influenced by patient age, drug dose, or duration of treatment, hierarchical multiple regressions were conducted. After the effect of premorbid IQ was partialed out, age, drug dosage (mg), treatment duration (yr), and their higher-order interactions were entered into two regression equations.
Results.
Subjects in the experimental group (24 women, 1 man) ranged in age from 22 to 73 years with a mean age of 51 years (standard deviation = 13.0). The control group (24 women, 1 man) had a mean age of 50 (standard deviation = 12.6). Prednisone dosage ranged from 5 mg qd to 40 mg qd with a mean of 16.4 mg/day. The groups did not differ significantly on Hamilton Depression Rating Scale scores (t = -0.91, p = 0.37). Although the experimental group had higher scores on the Young Manic Rating Scale (t = 2.31, p = 0.03), individual scores were low and were not clinically significant. Young Manic Rating scores did not correlate significantly with any of the variables (p > 0.05).
With the exception of more doctor visits on the part of the control group, Mann-Whitney U tests for nonparametric distributions yielded no significant group differences on either the total scores or individual items of the Disability Inventory Questionnaire. This particular variable was not considered to be a significant difference in severity of illness between groups because certain members of the control group were taking weekly injections of gold sodium thiomalate.
Memory variables.
Statistical analyses of memory variables yielded significant group differences on those tests that are sensitive to hippocampal dysfunction. The means and standard deviations for each of the variables are presented in Table 1. The glucocorticoid-treated group recalled fewer bits of information on the Long Delay condition of the Paragraph Recall Test (t = -2.79, p = 0.008) and a strong nonsignificant trend was noted for Long Delay List Learning (t = -1.89, p = 0.066). List Learning Discriminability scores, an index of recognition memory, were significantly lower for the experimental group (t = -2.17, p = 0.035). The groups did not differ on the implicit, or nonhippocampal-dependent, Word Stem Completion Priming Task (t = 1.11, p = 0.272), the Vigilance Test (t = -1.00, p = 0.326) or Digit Span Forward (t = -1.57, p = 0.12).
Table 1. Means, standard deviations, and results of t tests for cognitive measures for the entire sample
To determine if the drug effect on paragraph recall was influenced by patient age, dose, or duration of treatment, hierarchical multiple regressions were conducted. After the effect of premorbid IQ was partialed out, age, drug dosage (mg), treatment duration (yr), and their higher-order interactions were entered into two separate regression equations (Table 2).
Table 2. Regression equations (Paragraph Recall-long delay-dependent variable)
While the drug effect was not dose- or treatment-duration-dependent, patient age did account for a significant proportion of the variance in paragraph recall scores. Furthermore, a significant age by duration interaction effect indicated that the effect of duration of treatment on paragraph recall scores varied as a function of patient age. To interpret the direction of this interaction, Pearson correlations were calculated between patient age and paragraph recall scores for two treatment duration groups. The short-duration group consisted of 15 patients between the ages of 22 and 72 (median age = 55) who had been taking prednisone for 1 to 3 years. The long-duration group of 10 patients (aged 34 to 73, median = 45) had been taking prednisone between 4 and 15 years. Correlation analyses indicated that increasing age was associated with greater memory impairment in those patients taking the drug for less than 3 years (r = -0.65; p = 0.008) but not for those taking the drug between 4 and 15 years (r = -0.55; p = 0.125).
Study 2.
Methods.
Plan of design.
Study 2, a prospective, longitudinal design examining the acute effect of therapeutic dosages of prednisone on memory, was conducted to minimize the limitations inherent in the cross-sectional design of Study 1.
Subjects.
Exclusion and inclusion criteria and recruitment procedures were the same as in Study 1. Seven experimental and seven closely matched control subjects were enrolled in the study. Prednisone dosage ranged from 5 to 80 mg. Four women and three men ranging in age from 29 to 54 years participated as experimental subjects. None of the subjects had undergone corticosteroid treatment in the previous 6 months.
Procedures.
The same battery of neuropsychological tests was administered, except that alternate forms with well-documented reliability were used for the repeated-measures design. All experimental subjects were administered the test battery within 1 week prior to the initiation of prednisone, as well as 1 week and 12 weeks after the drug administration. The matched control subjects were tested at the same intervals with each pair of subjects receiving a different order of alternate test forms to avoid the possibility of order effects.
Statistical analyses.
Based on the results of Study 1, Delayed Paragraph Recall and Word Stem Completion Priming were the only variables analyzed because of the limited number of subjects. A series of analyses of covariance (ANCOVA) was conducted to investigate differences between groups on these variables for the two post-treatment testings. Although there were no significant baseline group differences for either variable, baseline scores were used as covariates. This method of analysis was chosen because baseline paragraph recall was a significant covariate for paragraph recall after 12 weeks of treatment (t[1,10] = 2.63, p = 0.025). Treatment status was the between-groups factor and time of testing was the within-subjects repeated measures factor.
Results.
(Table 3) lists the adjusted group means for the explicit (Paragraph Recall) and implicit (Word Stem Priming) memory measures for the three testings. Results of the ANCOVA with baseline scores as a covariate are presented in Table 4.
Table 3. Means and standard deviations for explicit (Paragraph Recall) and implicit (Word Stem Priming) memory tests
Table 4. Analysis of covariance for Paragraph Recall* and Word Stem Priming* (between-group effect)
Paragraph Recall scores for the treated group were significantly lower than those of the control group after 12 weeks of treatment (F[1,10] = 7.77, p = 0.019) with a strong trend toward significance after 1 week of treatment (F[1,10] = 4.44, p = 0.061). Figure 1 depicts the increasingly improved performance on the Paragraph Recall Task for the control group, which demonstrated the expected practice effect. There was no such expected effect for the treated groups, but rather, there was a consistent, albeit small, decrease in performance across time. There were no significant effects for the Word Stem Priming Task.
Figure 1. Mean scores for Paragraph Recall and Word Stem Priming for treated and control groups at baseline and 1 and 12 weeks.
Discussion.
Results of Study 1 indicate that patients receiving glucocorticoid treatment for at least 1 year perform significantly worse than closely matched medical controls on tests of hippocampal-dependent memory. Furthermore, the selective impairment on tasks of verbal explicit memory was not secondary to inattention, affective disturbance, generalized global cognitive decline, or disease severity. The drug effect on memory was not influenced by dose or duration of treatment, but, as expected, patient age did account for a significant amount of the variability in scores on paragraph recall. Furthermore, increasing age was associated with greater memory impairment when the duration of treatment was less than 3 years, suggesting that the aging brain is more susceptible to the effects of less-protracted treatment with prednisone.
Results of Study 2 suggest that even short-term use of prednisone adversely affects performance on tests of explicit memory. The small, progressive decrement in performance noted across testings for the treated group is even more significant when the lack of an expected practice effect from task familiarity is taken into account. The practice effect was quite apparent in the closely matched medical control group. The effect of age could not be assessed because of the limited number of subjects. Although not a statistically significant difference, the baseline scores for Paragraph Recall were lower for the treated group despite careful matching of controls on age, sex, IQ estimates, diagnosis, and severity of disease. Accordingly, these scores were used as a covariate in the regression equation. Although these differences in baseline scores may reflect normal variability, effects of previous steroid use on the part of the treated subjects remains a possibility.
The dissociation between drug effect on tests of explicit and implicit memory strongly suggests an effect of exogenous glucocorticoids on hippocampal function. Such a dissociation also has been recorded in patients with damage to the medial temporal area as a result of viral encephalitis, posterior cerebral artery occlusion, Alzheimer's-type dementia, and hypoxic-ischemic hippocampal damage. [48] Impaired list learning discriminability scores on the part of the treated group in Study 1 supports an encoding rather than retrieval deficit. This is further evidence for medial temporal lobe involvement.
The age by treatment duration effect on memory contributes to the growing literature on age-related dysfunction of the hypothalamic-pituitary-adrenal axis. In the rat, hippocampal glucocorticoid receptor content declines with age [49-51] and glucocorticoid neurotoxicity is exacerbated. Daily corticosterone injections into young adult rats over 12 weeks mimicked the pyramidal neuron loss seen in aging, [3] while adrenalectomy in midlife protects against hippocampal neuronal loss in adulthood. [52] In elderly humans, there is a decreased threshold for feedback resistance to dexamethasone, [53-55] as well as increased dexamethasone resistance in aged people with depression compared with younger people with depression. [56] Dexamethasone resistance, increasing age, and declining cognitive function are highly correlated in both the healthy elderly [56] and patients with Alzheimer's disease (AD). [57] Basal hypercortisolemia occurs in patients 80 years and older. [58-60]
Age-related hypersensitivity to glucocorticoids is of particular importance considering the current discrepancy of thought regarding glucocorticoid treatment. Despite the evidence reviewed here that glucocorticoids cause hippocampal neuronal loss, and the acceptance of hippocampal neuronal loss as a current hypothesis of biologic aging, [61] a multicenter trial is nevertheless underway to assess the efficacy of prednisone in the treatment of early AD. The rationale for the study is based on pilot data that suggest that anti-inflammatory drugs suppress peripheral markers of the acute-phase response in AD, which may contribute to neuronal loss, [62] epidemiologic studies that support a lack of co-morbidity between AD and rheumatologic disease, [63] and a presumed protective effect of long-term, nonsteroidal anti-inflammatory drug use against AD. [64] There is one case report of a stroke patient with severe cognitive impairment associated with low plasma cortisol whose condition improved with prednisone treatment. [65]
Data from both studies reported here, along with clear experimental evidence of a detrimental effect of glucocorticoids on hippocampal function, suggest that the potential benefit of anti-inflammatory treatment in AD might be counterbalanced by possible iatrogenic memory impairment, at least when synthetic glucocorticoids are considered. They also indicate that the risk of memory impairment should be carefully considered before initiating treatment with glucocorticoids, and that the use of glucocorticoids must be considered in the differential diagnosis of memory loss.
Acknowledgments
We would like to express our sincere appreciation to the physicians and staff of the Department of Internal Medicine and the Department of Neurology, Wayne State University School of Medicine, and the Department of Neurology, Henry Ford Hospital, Valerie Evans for her technical assistance in preparing the manuscript, and Dr. Darren Fuerst for statistical consultation.
- Copyright 1996 by Advanstar Communications Inc.
REFERENCES
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.
- 8.↵
- 9.
- 10.↵
- 11.
- 12.
- 13.↵
- 14.
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.
- 51.
- 52.↵
- 53.↵
- 54.
- 55.
- 56.↵
- 57.↵
- 58.↵
- 59.
- 60.
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
Letters: Rapid online correspondence
REQUIREMENTS
You must ensure that your Disclosures have been updated within the previous six months. Please go to our Submission Site to add or update your Disclosure information.
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.
