Results based on this study were presented in a poster session at the Annual Meeting of the American Academy of Clinical Neuropsychology, Boston, Massachusetts, June 19-21, 2008.
The growing numbers of survivors of innovative cancer treatments, such as hematopoietic stem cell transplantation (HSCT), often report subsequent cognitive difficulties. The objective of this study was to evaluate and compare neurocognitive changes in patients with chronic myelogenous leukemia (CML) or primary myelodysplastic syndrome (MDS) after allogeneic HSCT or other therapies.
In this prospective cohort study, serial evaluations of attention, concentration, memory, mood, and quality of life were used in a consecutive sample of 106 eligible patients who had CML (n = 91) or MDS (n = 15) at enrollment and then 12 months and 18 months after HSCT or other therapy.
The 3 evaluations at enrollment, 12 months, and 18 months were completed by 98%, 95%, and 89% of surviving participants, respectively. Among all patients, there was significant improvement in memory over 18 months. For example, the 45 patients who underwent HSCT (42 patients with CML and 3 patients with MDS) compared favorably with the patients who received other treatment on most measures of neuropsychological function, except they had improved mental health (P = .034), worse physical function (P = .049), and more difficulty with coordination and fine motor speed bilaterally (dominant hand, P = .005; nondominant hand, P = .0019). Patients with CML overall had improved phonemic fluency (P = .014).
Allogeneic hematopoietic stem cell transplantation (HSCT) is an established treatment for hematologic malignancies and other immunohematopoietic disorders.1, 2 Most studies of long-term adjustment after HSCT demonstrate that recipients have good to excellent quality of life despite deficits in the physical, psychological, social, and cognitive functioning domains.3-5
Elucidation of the memory and attention concerns expressed by HSCT survivors is an area of active inquiry. Some investigators have described cognitive impairment in patients who were evaluated only either before or after the procedure.6, 7 Others, using baseline and follow-up assessment, generally have reported impairment before transplantation, decline in some areas shortly after transplantation, and improvement without a complete return to baseline levels of cognitive function 12 months after HSCT.
The effect of allogeneic HSCT on cognition also has been difficult to judge, because previous large studies have included patients who had a mix of diagnoses or who received a mix of treatments, which differentially may have affected the central nervous system. A large number of those who are tested before transplantation may not have been participants after transplantation. Two recent studies with initial sample sizes >100 are illustrative. For example, 1 study randomized 476 patients to 1 of 3 testing schedules, and only 1 group was tested before HSCT and then 6 months and 12 months later; the other 2 groups were tested only after HSCT. The participants in that study underwent either autologous (79%) or allogeneic (21%) transplantation and had a wide variety of diseases, including solid malignancies (multiple myeloma, 63%; non-Hodgkin lymphoma, 10%; acute myelogenous leukemia, 7%; breast cancer, 6%; Hodgkin lymphoma, 3%; other, 11%). Although those researchers observed an improvement in cognitive function over 12 months, there was 70% attrition in the group that had pre-HSCT evaluations.8 Another study that had an initial sample size of 142 adult recipients of allogeneic HSCT included patients who received total body irradiation (63%) and those who did not; patients who received previous cranial irradiation or intrathecal chemotherapy (12%); and also pooled results for patients who had a variety of hematologic malignancies (chronic-phase chronic myelogenous leukemia [CML], 42%; accelerated CML, 8%; acute myelogenous leukemia [AML] in remission, 14%; new or recurrent AML, 7%; myelodysplasias, 13%; lymphoma, 11%; multiple myeloma, 4%; chronic lymphocytic leukemia, 2%). Fifty-four of 142 participants (38%) were tested at all time points.9 The majority of the other published studies that have sought to complete serial evaluation of cognitive changes have had smaller initial sample sizes (range, 25-71 patients), and follow-up evaluations were obtained at time points ranging from 1 day to 3 days after HSCT and up to 1 year later.10-16
Thus, more remains to be learned concerning cognitive changes after allogeneic HSCT. Data regarding allogeneic HSCT are confounded by samples that include autologous recipients, patients with breast cancer, or patients who received previous cranial radiation or intrathecal treatment. Autologous HSCT recipients may have fewer central nervous system complications than recipients of allogeneic HSCT.17 The neuropsychological consequences of breast cancer differ from the consequences of hematologic disorders.14 Previous cranial radiation or intrathecal treatment may exert an adverse impact on subsequent cognitive performance, thus obscuring the true consequences of HSCT. A high attrition rate in participants who are tested before and after HSCT may not be random; for example, perhaps only those with better function return to studies.
Patients now are more concerned about the threat of cancer treatments to cognition. They may have heard about “chemobrain,” which has been characterized as cognitive slowing and difficulties with concentration, memory, and multitasking. Although some previous neurocognitive findings related to HSCT were compared with normative data, most did not compare results related to other treatment options. A comparison of cognitive changes in patients undergoing allogeneic HSCT with changes in patients receiving other treatment options is timely and can address the fears that may influence treatment decisions.18
The objective of this prospective cohort study was to evaluate and compare the neurocognitive changes in patients with either CML or primary myelodysplastic syndrome (MDS) as they underwent treatment for their illness. The 2 diseases were chosen because neither typically involves the central nervous system. Their treatments are similar, and allogeneic HSCT has curative potential.19, 20 Participants were asked to complete 3 evaluations over the course of the study so that the effects of time, treatment (allogeneic HSCT or other), and disease (CML or MDS) on neurocognitive function could be measured. Participants were evaluated before HSCT and then 12 months and 18 months later to extend our knowledge regarding the longer term effects. We hypothesized that individuals undergoing allogeneic HSCT would experience more neuropsychological changes than individuals receiving other treatments, but the duration and extent of such changes would be revealed over the course of the project.
MATERIALS AND METHODS
A prospective cohort study of 108 individuals with CML or primary MDS was planned. Sample size was based on an estimated enrollment of 36 individuals per year for 3 years with a 1:1 ratio of patients who did undergo HSCT and did not undergo transplantation and an estimated attrition rate of 20%. Eligibility criteria included agreement to participate in 3 neuropsychological evaluations over the course of 18 months, reading and listening comprehension of English, and diagnosis within the past year or a new treatment plan that included HSCT within the next year. Exclusion criteria included history of significant head injury (resulting in loss of consciousness), stroke, epilepsy, or other central nervous system pathology requiring radiation, surgery, or past/current intrathecal medication, and current alcohol or substance abuse or dependence, all factors that could influence performance on neuropsychological tests.
Participants were recruited from Dana Farber/Brigham and Women's Cancer Care in Boston, Massachusetts (84.3%), and other participants came from the Massachusetts General Hospital and other practices (6.5%). The remaining responded to web-based and other study advertisements (9.2%). All were to receive either allogeneic transplantation or other treatment, as directed by their physicians.
Participants completed a baseline assessment interview, which consisted of 1) a patient profile to include education, usual occupation, past medical and psychiatric history, and medication history, with usual occupation coded according to the 1989 General Social Survey prestige scales21; 2) the Shipley Institute of Living Scale, which was used to estimate a full-scale intelligence quotient (IQ)22; 3) the Medical Outcomes Study 36-Item Short Form (SF-36), which evaluated the physical and mental health of the participant23; and 4) the Brief Profile of Mood States (POMS), in which participants were asked to rate how each adjective from a list of 11 items that described how they felt in the past week with each item scored from zero (not at all) to 4 (extremely) for a total potential score that ranged from 0 to 44.24 For an estimate of “confusion” or “cognitive difficulties,” 2 items (bewildered and muddled) from the POMS were highlighted with scores ranging from zero (none) to 8 (most). To confirm current alcohol and drug diagnoses, participants completed the Alcohol and Drug Modules from the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders, fourth edition.25
Participants also were administered a battery of neuropsychological tests. Test selection was based on previous research on patients with hematologic malignancies and incorporated measures of attention, executive function, memory, and motor speed.26, 27 Attention and concentration were evaluated with the Trail Making Test and the Verbal Fluency Test.28, 29 The Trail Making Test has 2 parts, A and B, and measures visuomotor speed, attention, and mental flexibility. Part A involves connecting lines from 1 to 25 that are printed randomly on a sheet of paper. It primarily assesses attention and motor speed. Part B incorporates letters and numbers, and the individual alternates between numbers and letters, in order. The Verbal Fluency Test measures activation retrieval that is phonemic (letters) and semantic (animals). It is a measure of rapid word retrieval under restricted search conditions. Individuals are asked to generate as many words as they can in 1-minute trials. Anterograde learning and memory were assessed with the 6-trial Buschke Selective Reminding Test. A list of 12 words is read to the individual, who is instructed to remember as many words as possible. On subsequent trials of recollection, the individual is reminded selectively of omitted words. Two measures from the Buschke Selective Reminding Test were emphasized in the current study: total recall and long-term storage.30 The Grooved Pegboard test was used to test coordinated and fine motor speed by asking participants to place pegs in a board with their dominant and nondominant hands as quickly as possible.31
Participants completed repeat assessments 12 months and 18 months after the initial assessment. Each repeat assessment included the SF-36, Brief POMS, and the Grooved Pegboard test. To mitigate possible learning effects on the neuropsychological testing, individuals were given randomly alternate forms of the Verbal Fluency Test, Trail Making Test, and Buschke Selective Reminding Test. The reliability and validity of these tests are well documented.32, 33
Neuropsychological tests were scored according to established procedures.34-37 Raw scores were converted to Z-scores from age, sex, and/or education corrected normative data. A Z-score between −1.0 and +1.0 is between the 16th and 84th percentiles for the general population. The SF-36 items and scales were scored as required, and higher scores indicated a better state of health.38 Both the Physical Component Summary (PCS) and Mental Component Summary (MCS) on the SF-36 have a mean score of 50 and a standard deviation (SD) of 10.
Participants provided written informed consent. They received an honorarium of $50 for each assessment. This study was reviewed and approved by the Partners Institutional Review Board, which is responsible for the review and approval of all human subject research conducted by the staff of Brigham and Women's Hospital and other Partners-affiliated hospitals.
Data were analyzed using the SAS statistical package (version 9.1; SAS Institute, Inc., Cary, NC). Descriptive results are reported as counts with percentages or means with SDs. Fisher exact tests or chi-square tests of significance were used to compare categorical baseline participant demographic and clinical characteristics as appropriate between disease and treatment subgroups. In addition, t tests were used to compare age, occupational prestige, and estimated total IQ.
The primary purposes of this study were to quantify neuropsychological and mental status at the start of therapy and again after 12 months and 18 months and to determine whether there were any significant changes over time. Our analyses were based on the 77 participants who had completed at least 2 evaluations, and we used a longitudinal repeated measures linear regression to look for improvement or worsening over time. The SAS Mixed procedure was used to adjust for correlation between serial measurements on patients, and an unstructured correlation matrix was used to allow the correlation to diminish with time. One of our primary concerns, however, was the possibility that the course of neurocognitive changes might depend on the type of disease (CML or MDS) or the treatment regimen (HSCT or other treatment). Therefore, for each outcome variable, we ran 2 preliminary regression models. The first model included a time-by-disease interaction to check whether neurocognitive changes were different in CML patients than in MDS patients. The second model included a time-by-treatment interaction to check whether neurocognitive changes were different in HSCT recipients compared with patients who received other treatments. The 2 interaction terms were not included in the same preliminary model to avoid collinearity. If either (or both) interaction term was found to be statistically significant, then it was included in a final model that had our primary predictor, time (as a categorical variable), as well as covariates for disease type and treatment regimen. We emphasize, however, that our purpose was not to compare outcomes between patient subgroups or to compare outcomes between treatments in a nonrandomized setting. We simply wanted to ensure that, if the time course was different in a particular subgroup of patients, then we would be able to show the results separately for each relevant subgroup.
Of the 336 individuals who were screened for study enrollment, 106 (31.5%) were eligible, gave informed consent, and are described in Table 1. One hundred four of the 106 eligible participants (98%) completed the baseline assessment, and 2 participants with CML were unable to finish and were excluded. The baseline assessment was completed at a median of 5.6 months after the participants' diagnosis date or 6.1 months for the 91 patients who were diagnosed with CML and 3.8 months for the 15 patients who were diagnosed with MDS. With respect to HSCT, the baseline assessment was completed an average of 67 days (SD, 106 days) before the procedure. And 77 participants completed the second assessment, which occurred an average of 400 (SD, 77 days) days after the baseline assessment. Most loss at this time was because of death (23 of 104 cohort participants died), resulting in a follow-up rate of 95% of survivors (or 73% of those who gave initial consent). These 77 participants who had at least 2 serial measures were included in all of our analyses of longitudinal change. Between 12 months and 18 months, 2 participants died, and 8 were lost to follow-up. Sixty-seven of the 75 surviving participants (89%) completed the final, 18-month evaluation (average 185 days [SD, 59 days] after the 12-month evaluation). The 67 participants represent 63% of those who gave initial consent. The study began in January 2002 and ended in January 2007. Figure 1 summarizes the overall study flow.
The individuals with CML (n = 91) and MDS (n = 15) were similar in terms of sex, racial and ethnic background, education, occupational prestige (eg, 48.72, management-related occupation; 44.47, actuary),21 and total estimated IQ at enrollment. Individuals with MDS were significantly older (P < .0001), but this age difference was accounted for in the analyses by the use of Z-scores, which were corrected for age, sex, and education. Recipients of allogeneic HSCT were significantly younger (mean ± SD, 40.6 ± 10.6 years vs 53.5 ± 12.9 years [P < .0001]). None of the participants satisfied diagnostic criteria for current alcohol or drug abuse or dependence diagnoses.
There was a trend for more individuals who had CML to undergo allogeneic HSCT (47%; n = 42) compared with individuals who had MDS (20%; n = 3 [P = .053]). Graft-versus-host disease prophylaxis consisted of tacrolimus and methotrexate, which were administered using the standard techniques used for patients receiving HSCT.39 Approximately half of the HSCT recipients (53%) used stem cells from a donor who was related to the patient. Only 3 of 45 allogeneic HSCTs were nonmyeloablative; 94% of the HSCT recipients received total body irradiation (total dose, 14 grays in 7 fractions). Among the 48 individuals with CML who received other treatment, the majority received imatinib mesylate (85.4%), and the rest received either hydroxyurea (10.4%) or interferon (4.2%). Treatment for the 12 individuals with MDS who did not undergo allogeneic HSCT was as follows: hydroxyurea (n = 2 patients; 17%), supportive treatment (n = 3 patients; 25%), erythropoietin (n = 4 patients; 33%), and azactidine (n = 5 patients; 42%), with some patients receiving more than 1 of the listed treatments simultaneously. Table 1 summarizes patient characteristics at enrollment.
Time, Disease, and Treatment Effects on Outcome Measures
Outcome measures were affected differentially by time, disease, and treatment. In general, when significant time effects were identified as present, they were in the direction of improvement relative to baseline. Table 2 summarizes the least square means with standard errors (SEs) for estimating the time effects on the outcome measures.
SE indicates standard error of the mean; SF-36, Medical Outcomes Study 36-Item Short Form; MCS, Mental Component Scale; HSCT, hematopoietic stem cell transplantation; CML, chronic myelogenous leukemia; MDS, myelodysplastic syndrome.
The differences between the 18-month and 12-month values were not statistically significant (P > .05).
P values for total recall and long-term storage are for time effects across all patients. P values for other outcomes are for the time-by-treatment or time-by-diagnosis interaction.
Significant time effects on measures
−0.53 ± 0.15
−0.18 ± 0.15
−0.077 ± 0.16
P compared with baseline
−0.41 ± 0.15
−0.22 ± 0.15
.0018 ± 0.15
P compared with baseline
Significant time effects by treatment
43.62 ± 2.32
49.40 ± 1.87
50.26 ± 1.78
P compared with baseline
48.16 ± 1.47
48.57 ± 1.61
49.64 ± 1.53
P compared with baseline
Significant time effects by diagnosis
−0.13 ± 0.11
.10 ± 0.13
.065 ± 0.12
P compared with baseline
−0.039 ± 0.29
−0.47 ± 0.20
−0.033 ± 0.26
P compared with baseline
−0.72 ± 0.20
−0.55 ± 0.16
−0.46 ± 0.18
P compared with baseline
−0.1.79 ± 0.50
−1.41 ± 0.54
−2.32 ± 0.58
P compared with baseline
Time Effects on Neuropsychological Measures
Overall, participants' total recall measured by the Buschke Selective Reminding Test improved over time, with significant gains observed from baseline to the 12-month evaluation (P = .0055) and from baseline to the 18-month evaluation (P = .0016). Similarly, there was improvement in long-term storage, with the greatest improvement observed from baseline to the 18-month evaluation (P = .0028). There were neither treatment effects nor diagnosis effects on either of these 2 measures of memory (see Table 2).
Time-by-diagnosis interactions were identified for 2 neurocognitive measures: phonemic fluency (P = .014) and Grooved Pegboard for the dominant hand (P = .04). For phonemic fluency, patients with CML had improved scores at 12 months (P = .009) and at 18 months (P = .033), both compared with baseline. In contrast, those with MDS had no improvement compared with baseline at either 12 months or 18 months. For the Grooved Pegboard, dominant hand, those with CML improved from baseline to 18 months (P = .094), but those with MDS declined during the same period (P = .092).
Disease Effects on Neuropsychological Measures
Across all time points, participants with CML had better performance than those with MDS on several neurocognitive measures. Specifically, those with CML had better performance on the Grooved Pegboard test of motor function for the nondominant hand (effect estimate, 1.47; SE, 0.50 [P = .0042]). Participants with CML also scored approximately 1 SD better than those with MDS for the dominant-hand Grooved Pegboard test, but the magnitude of difference varied with time, as shown in Table 2. Similarly, there was a significant difference by diagnosis on performance for both Part A and Part B of the Trail Making Test. Overall, individuals who had CML had better scores on Part A (effect estimate, 0.69; SE, 0.32 [P = .034]) and Part B (effect estimate, 2.09; SE, 0.50 [P = .0002]) compared with individuals who had MDS. Table 3 summarizes these selected neuropsychological results.
HSCT recipients had worse performance on tests of motor function (Grooved Pegboard) for both the dominant hand (effect estimate, −1.13; SE, 0.31 [P = .0005]) and the nondominant hand (effect estimate, −1.08; SE, 0.33 [P = .0019]). These results are summarized in Table 3.
Other Neuropsychological Measures
There were no time, treatment, or disease effects on semantic fluency, which generally appeared to stay within normal limits. The Z-scores for this measure are listed by disease and treatment in Table 3.
Medical Outcomes Study 36-Item Short Form Mental and Physical Component Summary Scores
Significant time by treatment effects for the MCS of the SF-36 were identified (P = .034). HSCT recipients had low initial MCS scores (least square mean = 46.6) but demonstrated statistically significant gains at the 12-month (∼+5.8; P = .013) and 18-month (∼+6.6; P = .01) follow-up assessments compared with the baseline assessment. In contrast, patients who received treatments other than HSCT had no significant changes in their mean MCS scores from the time of initial evaluation to the 18-month follow-up, and all of their scores were <50 (see Table 2).
HSCT recipients had poorer average PCS scores from the SF-36 by nearly 4 points (P = .049). The deficit was present at the pretreatment baseline evaluation and through follow-up and was not attributed to HSCT. Figure 2 illustrates the PCS scores for participants by disease and treatment.
Profile of Mood States
Table 4 lists the mean and SD scores for the POMS and the 2 “confused” items. The results for the POMS did not differ by disease (effect estimate, 2.56; SE, 2.65 [P = .34]) or treatment (effect estimate, −0.34; SE, 1.80 [P = .85]). Overall, participants did not endorse many items that were consistent with depression. Similarly, there were no differences on the 2 “confused” items by disease (effect estimate, 0.18; SE, 0.50 [P = .72]) or treatment (effect estimate, −0.23; SE, 0.34 [P = .51]).
The main finding of this prospective cohort study is that time effects on measures of neurocognitive function are important. There was general improvement from the time of treatment initiation to 18 months later for measures of memory (total recall, P = .005; long-term storage, P = .01). Hence, complaints regarding “chemobrain” after HSCT merit further study, because deficits appear to predate the initiation of that treatment and subsequently improve. Indeed, participants did not strongly endorse the 2 POMS items that were suggestive of confusion or difficulties with mental acuity (feeling bewildered or muddled); in addition, they did not endorse many items that were consistent with depressed mood. This is important, because earlier, influential studies suggested that cancer patients who reported concentration and memory problems, in fact, were more likely to be clinically depressed or anxious than to have such deficits confirmed on objective testing.40 Strengths of the study included a low attrition rate, focus on allogeneic HSCT for 2 hematologic diseases, and exclusion criteria that eliminated prior central nervous system trauma and treatment, all of which could have obscured results. Several other findings are highlighted.
First, individuals who underwent HSCT compared favorably with those who received other treatment with regard to their performance on most measures of neuropsychological function and mood. Compared with individuals who received other types of treatment, those who underwent HSCT had worse physical function (P = .049) and had more difficulty with coordination and fine motor speed in both the dominant hand (P = .005) and the nondominant hand (P = .0019). Possible explanations for the lower PCS scores among HSCT recipients include greater severity or acuity of hematologic malignancy. Greater difficulty in coordination and fine motor speed in HSCT recipients may be accounted for by the late effects of total body irradiation and chemotherapy, by graft-versus-host disease, or by the medications for its treatment.
Second, diagnosis or disease may be an important factor when assessing neurocognitive and other changes, because there may be disease-specific features. For example, patients with CML had significantly more improvement over time on phonemic fluency and motor tasks than patients with MDS.
Several potential limitations to the generalizability of findings should be noted. Because patients were not randomized to treatment, any differences between treatment groups easily could be ascribed to the patient-selection process because of his/her disease or because of the treatment itself. In the current study, the number of participants with MDS was small relative to the number of those with CML. The IQ of the sample was average to high average; therefore, it is possible that individuals who had higher cognitive reserve participated. Participants were given modest honoraria, which may have influenced rates of participation in either direction. HSCT was compared naturalistically with other treatment, and imatinib mesylate was the most common. This medication is a molecularly targeted therapy for the treatment of cancer and may be associated with fewer side effects than conventional chemotherapy. The use of 2 items from the POMS to capture self-assessment of confusion is a novel application and needs further study. Finally, it was impossible to evaluate the neurocognitive function of participants before their diagnosis. Future, ideal studies would address some of these limitations, if possible.
These considerations notwithstanding, the results from this study support findings from other research and extend our knowledge regarding the trajectory of neurocognitive recovery. Similar to other investigations, a general improvement in the 12 months after HSCT was observed with some impairments pre-HSCT.8-16 Moreover, there appeared to be further improvement at 18 months. Because allogeneic HSCT is a therapy for a wide variety of diseases, these results suggest that it also may be important to evaluate neurocognitive changes in the context of baseline disease.
Early detection and improved treatments of cancer, such as HSCT, have resulted in growing numbers of cancer survivors, now >10 million.41 This large group has entered a new phase of medical care that will need to take into account the long-term physical and psychological effects of their cancer treatment.42 Although there appeared to be some differences in neurocognitive function after HSCT compared with function after other treatments, particularly with regard to fine motor function, there was no apparent increase in adverse effects for the other aspects of executive function that were evaluated. The PCS and MCS scores for the study sample were lower than the norms established for a healthy US population with no chronic conditions (median PCS, 56.86; median MCS, 54.26) but were comparable to those for individuals with cancer (except skin cancer) in the United States (median PCS, 41.49; median MCS, 51.32).36 Thus, the results from the current study may reassure prospective HSCT recipients that this particular treatment compares favorably with other treatments when mental status side effects are considered. Indeed, the overall pattern of measures of neurocognitive function indicate that there is improvement over time.
We thank Alyson Lavigne-Dolan and Christina Breigleb, lead research assistants.
Conflict of Interest Disclosures
Supported in part by Grant RSG-01-246-01-PBP from the American Cancer Society and by Grant K24 AA 000,289 from the National Institute on Alcohol Abuse and Alcoholism (both to Grace Chang).