Patients and methods

Patients

Participants were cancer patients who had completed autologous or allogeneic BMT as part of their treatment at the H Lee Moffitt Cancer Center in Tampa, FL, USA between September 1997 and July 1999. To be eligible for the current study, these patients had to: (1) be between 18 and 65 years of age; (2) be able to speak and read English; (3) have completed at least 8 years of formal education; (4) have no clinical evidence of disease progression or recurrence at the most recent follow-up visit; and (5) be approximately 6 months post-discharge from BMT.

Of 100 BMT recipients meeting eligibility criteria, seven patients could not be contacted, 16 refused to participate, seven did not complete cognitive testing, but did complete self-report questionnaires, and five completed cognitive testing but did not complete self-report questionnaires. Complete cognitive and self-report data were obtained from 65 of 100 BMT recipients (65%). Analyses comparing participants who provided complete data (n=65) with eligible nonparticipants (n=35) indicated that there were no significant differences (P<0.05) between the groups in terms of gender, age, ethnicity, education, marital status, employment status, or type of transplant. However, there was a significant (P<0.05) group difference in cancer diagnosis, with participants more likely than nonparticipants to be diagnosed with breast cancer (67% vs 39%).

Study participants ranged in age from 23 to 63 years (M=47, s.d.=9). The majority of participants were female (78%), Caucasian (92%), currently married (62.5%) and were educated beyond high school (67%). Average education level was 14.1 years (s.d.=2.6, range 8–20 years). In all, 30% were currently employed, 41% were on leave, and the remaining 29% were not currently employed. In total, 81% of the sample underwent autologous transplantation and 19% underwent allogeneic transplantation. The majority of patients (67%) were diagnosed with breast cancer, followed by leukemia (13%), lymphoma (11%), and multiple myeloma (9%). Participants were approximately 6.8 months post-discharge from the BMT unit at the time of the follow-up assessment (s.d.=1.1, range=5.1–10.6 months). Premorbid intellectual function as estimated by the National Adult Reading Test (NART) averaged 106 (s.d.=9.4, range 79–122), which is within the average range of intelligence and is consistent with participants' mean education level.

Procedure

Approximately 5 months following discharge from the BMT unit, medical records were reviewed to determine which patients met eligibility criteria. A letter was sent to eligible patients informing them they would be contacted by telephone to discuss study participation. Approximately 1 week later, patients were contacted by telephone and details of the study were provided. Those patients who provided verbal informed consent during the telephone contact were given an appointment for an outpatient research visit in order to complete a cognitive battery and other clinician-administered measures. They were also sent a written informed consent form and self-report questionnaires to be completed before their outpatient appointment. Upon arriving for their outpatient appointment, patients returned the signed consent form and questionnaires and were then administered the cognitive battery and additional interview measures. All procedures and the consent form were approved by the University of South Florida Institutional Review Board.

Measures

Demographic and Clinical Data Form

Age, gender, ethnicity, marital status, employment status, and education level were obtained through self-report. Medical record review was performed prior to cognitive testing to record cancer diagnosis, type of BMT, dates of admission and discharge for BMT, and current cancer status.

Center for Epidemiological Studies Depression Scale (CES-D)¹⁰

A 20-item measure of depressive symptomatology with good reliability and validity with cancer patients.^{11, 12}

Medical Outcome SF-36 Health Survey (SF-36)¹³

A 36-item self-report measure assessing perceived health and functioning, measuring the extent to which health status impacts each of the following areas: (1) physical functioning, (2) role functioning – physical, (3) bodily pain, (4) general health, (5) vitality, (6) social functioning, (7) role functioning – emotional, and (8) mental health. The SF-36 yields two summary measures that measure global physical functioning and global mental health functioning¹⁴ and has acceptable reliability and validity with healthy and chronically ill individuals.¹³

Fatigue Symptom Inventory (FSI)¹⁵

A 14-item self-report measure assessing the frequency and severity of fatigue as well as perceived interference with quality of life, and has demonstrated reliability and validity with cancer patients.^{15, 16}

Multiple Abilities Questionnaire (MAQ)¹⁷

A 48-item self-report of subjective cognitive function in routine, daily activities across six domains: Attention, Language, Remote Memory, Verbal Memory, Visual–Spatial Memory, and Visual–Spatial Perception. Responses on the MAQ are in a five-point Likert format ranging from Almost Never (0) to Almost Always (4). Domain scores and a total MAQ score are calculated. Discriminant validity has been assessed by comparing normal controls to neurological subjects and concurrent validity assessed by comparing self-report to objective test performance with good results.¹⁷ In this study, complaints were defined as those items participants rated as Sometimes (2), Frequently (3), or Almost Always (4).

National Adult Reading Test¹⁸

Assesses pre-morbid intellect. Inter-rater reliability is high (0.96–0.98). The scale has been compared with the general factor of intelligence 'g' from the Wechsler scales, and was found to have a crossloading of 0.85.¹⁹

Controlled Oral Word Association (COWA) from the Multilingual Aphasia Examination²⁰

Assesses the speed and ease of word production and is a measure of executive function (eg initiation, perseveration, and effortful output). The test has strong psychometric properties.²¹

Trailmaking Test (TMT)²²

Assesses sequencing (an executive function) and is timed. Extensive reliability and validity data are available with Trails B more sensitive to cognitive impairment.²³

Hopkins Verbal Learning Test (HVLT)²⁴

Assesses verbal learning and memory consisting of a 12-item list of words belonging to three semantic categories presented over three learning/recall trials. The HVLT has strong psychometric properties and has been shown to discriminate between neurologically impaired and normal patients.²¹

WAIS-R Digit Symbol²⁵

Assesses sustained attention, psychomotor speed, and motor persistence. It has strong psychometric properties and is noted to be the most sensitive Wechsler subtest to brain dysfunction.²⁶

WAIS-R Digit Span²⁵

Assesses immediate verbal memory and auditory attention. Strong psychometric properties are reported.²⁶

Grooved Pegboard²⁷

Assesses speeded manual dexterity and is sensitive to general psychomotor slowing caused by medication and disease effects.²⁶

WMS-III Logical Memory²⁸

Assesses immediate and delayed memory for prose. Internal consistency assessed by split-half reliability ranges between 0.67 and 0.80 and the interscorer reliability coefficient is 0.99.²⁸

WMS-III Visual Reproduction²⁸

Assesses immediate and delayed nonverbal memory for designs. Reliability coefficients for immediate recall and delayed recall are 0.59 and 0.46. Interscorer reliability is reported at 0.97.²⁸

Stroop Neuropsychological Screening Test²⁹

Assesses concentration effectiveness, cognitive interference and the ability to inhibit responding.^{21, 26} Validity data indicate discrimination of cognitively impaired groups from normal groups.²⁹

Wisconsin Card Sorting Test (WCST)³⁰

Assesses executive function, specifically abstract concept formation and maintenance and shifting of cognitive sets. Extensive research indicates that it is sensitive to cognitive problems in multiple patient populations. Extensive reliability and validity data are available.³¹

Statistical analyses

Raw cognitive test scores were converted to z-scores according to published normative data.^{25, 28, 29, 31} Tests were combined into cognitive domains based on the cognitive function they were purported to measure as follows: Attention (Digit Span, Trails A), Verbal Learning (HVLT Total Recall and Recognition), Verbal Memory (Logical Memory immediate and delayed), Visual Memory (Visual Reproduction immediate and delayed), Simple Executive Function (Trails B, COWA, and Digit Symbol), Complex Executive Function (Stroop, WCST percent Perseverative Errors, and Percent Conceptual Level of Responding), and Psychomotor Speed (Grooved Pegboard dominant and nondominant hand).

Participants' total neuropsychological performance (TNP) scores were calculated by averaging all test z-scores in the battery. Cognitive domain z-scores were likewise computed by averaging the z-scores of tests in each respective domain. Cognitive domain scores and individual test scores below a z-score of -1.0 (more than 1 s.d. below the published normative mean) were defined as deficient.³² Z-scores between -1 and -1.5 were categorized as mildly deficient and z-scores equal to or lower than -1.5 were categorized as moderately to severely deficient. Z-scores greater than -1 were categorized as within the normal range.

Using the SAS 6.0 statistical analysis software, the frequency of cognitive deficiency per domain was calculated. Spearman and Pearson correlations were performed between variables and regression analysis was performed to determine the relative contribution of variables to cognitive function and cognitive complaints.

Results

Prevalence of cognitive deficits

Means, s.d.'s, and ranges of individual tests are presented in Table 1. Rates of deficient performance on individual tests are presented in Table 2. Based on the above deficit classification criteria, 37% (n=24) of participants had at least mild deficits (-1 s.d.) in psychomotor speed, 15% (n=10) had deficits in complex executive function, 6% (n=4) had deficits in verbal learning, and 3% (n=2) had deficits in verbal memory. Deficits in attention, visual memory, and simple executive function were each experienced by 1.5% (n=1).

Table 1 - Cognitive tests means, s.d.'s, and range (z=scores).

Full table

Table 2 - Cognitive test performance level – rate of deficiency.

Full table

Examining only those participants with moderate or severe deficits (-1.5 s.d.), 14% (n=9) had deficits with psychomotor speed, 6% (n=4) had deficits with complex executive function, and 3% (n=2) had deficits with verbal learning, verbal memory, and visual memory. No participants had moderate-to-severe deficits in attention and simple executive function.

Considering all deficits observed, 32 participants (49%) demonstrated no deficits across the cognitive domains, 16 (25%) were deficient in one domain, eight (12%) demonstrated deficits in two domains, and nine (14%) demonstrated deficits in three or more domains.

Considering only moderate or severe deficits, 47 participants (72%) demonstrated no deficits across cognitive domains, 12 (18%) were deficient in one domain, four (6%) were deficient in two domains, and two (3%) were deficient in three or more domains.

Frequency of cognitive complaints

The most common complaints were in the domains of Remote Memory, Attention/Concentration, and Language (see Table 3). The number of complaints per participant ranged from 1 to 42, with a mean of 14.8 (s.d.=10.4). In all, 50% of participants reported 11 complaints or less.

Table 3 - Correlation of TNP and total subjective complaints (MAQ) with demographic, clinical, and psychosocial variables, and quality of life.

Full table

Relationship of demographic, clinical, and quality of life variables to TNP

Correlational analyses were conducted to examine the association between TNP scores and demographic, clinical, and quality of life variables (see Table 4). Given the large number of variables correlated and the exploratory nature of this study, a conservatively stringent alpha level of <0.01 was chosen to designate significance. TNP was chosen over individual test scores and domain scores because it was determined to be a more meaningful variable, as it is a composite score of all cognitive domains assessed. With regard to demographic variables, findings indicated that males, older participants, and participants with lower estimated IQ scores were more likely to have lower TNP scores (P's<0.01). There was a trend (P=0.03) for participants with less education to have lower TNP. With regard to clinical variables, TNP scores were not associated with time since discharge, length of hospital stay, type of transplant, and type of cancer diagnosis (P's>0.01), and no trends were observed.

Table 4 - Cognitive complaints endorsed by 40% or more of participants – MAQ items.

Full table

With regard to quality of life variables, TNP scores were significantly negatively associated with depressive symptomatology and interference from fatigue (P's<0.01). There was a trend toward participants' average daily rating of fatigue (P=05). TNP scores were not associated with overall physical quality of life (P=0.12); however, there was a trend for higher TNP scores to be associated with better mental quality of life (P=0.08).

Relationship of cognitive complaints to TNP

Correlational analyses were conducted to examine the association between cognitive complaints (as measured by MAQ scores) and objective cognitive performance (as measured by TNP scores). TNP scores were not significantly associated with total subjective cognitive complaints (r=0.13, P>0.01) or with any of the subjective cognitive domains (P>0.01), although there was a trend for lower TNP scores related to more complaints in the Visuospatial Perception domain (r=0.28, P=0.02). Correlational analyses were also conducted to examine the association between objective cognitive domain scores and subjective (MAQ) domain scores. Performance in the Motor domain was significantly correlated (P<0.01) with three MAQ domains, specifically Visual–Spatial Perception, Visual–Spatial Memory, and Remote Memory. Performance in the Visual Memory domain was significantly correlated with the Visual–Spatial Perception MAQ domain (P<0.01) and trends were observed with Visual–Spatial Memory and Verbal Memory (P=0.02 and 0.06, respectively). There were no other significant correlations between objective cognitive domains and subjective cognitive domains.

Relationship of demographic, clinical, and psychosocial variables to cognitive complaints and cognitive performance

Correlational analyses were conducted to examine the association of demographic, clinical, and quality of life variables with both cognitive complaints and objective cognitive performance (see Table 4). With regard to cognitive complaints, there was a trend between age and total cognitive complaints (MAQ) (P=0.04), with younger participants reporting more cognitive complaints. There were no other demographic variables that approached significance. Among quality of life variables, more cognitive complaints were significantly associated with greater depressive symptomatology, more severe average daily fatigue, greater interference from fatigue, and poorer overall physical quality of life and mental quality of life (all P's<0.001).

Regression analysis was conducted to assess the relative contribution of demographic variables to the variability of cognitive performance as assessed by the TNP score. Using stepwise regression, estimated pre-morbid IQ was significant (F=18.30, P<0.01) and accounted for 23% of the variability. Age was also significant (F=7.54, P<0.01) and accounted for an additional 8% of the variability. Gender neared significance (F=5.50, P<0.05) and accounted for 6% of the variability. Education also neared significance (F=3.34, P=0.07) and accounted for an additional 3% of the variability.

Discussion

Relative to published norms, 51% of cancer patients in this study experienced at least mild deficiency in at least one cognitive domain when assessed 6 months following BMT. For reference, -1 s.d., the cutoff used to label mild deficiency, corresponds to 16% of the normal population. In all, 28% of participants experienced moderate or severe deficiency in at least one domain. For reference, -1.5 s.d.'s, the cutoff used to label moderate to severe deficiency, corresponds to 7% of the normal population. Among patients with deficits, the majority experienced deficits in psychomotor speed, a finding that echoes the results of the study by Harder et al.⁸ A notable percentage of patients, however, experienced deficits in simple and complex executive functioning tasks.

Previous studies found rates of impairment that differed from those reported here. van Dam et al⁵ identified 32% of their sample as impaired at 2 years post-BMT, Harder et al⁸ found 60% to be impaired an average of 45 months post-BMT, and Syrjala et al⁹ found 94% to be impaired at 80 days post-BMT and 74% to be impaired at 1 year post-BMT. Differences in the impairment rate across studies may be partly be due to differing criteria for impairment. The present study identified impaired patients as those who averaged at least 1 s.d. below published norms on tests combined to form cognitive domain scores. van Dam et al⁵ identified impaired patients as those scoring at least 2 s.d.'s below the mean of a healthy control group on at least three individual tests. Harder et al⁸ defined impairment as scoring at least 2 s.d.'s below published norms on at least one test, and Syrjala et al⁹ identified impairment as scoring at least 1 s.d. below published norms on at least one test. As different studies used different criteria for impairment, and because they measured patients at different times post-BMT, cross study comparisons may be problematic. All studies agree, however, that cognitive impairment following BMT affects a significant proportion of patients.

Cognitive impairment in cancer patients is not unique to BMT and has been observed, to some degree, in patients with CNS malignancy, breast cancer, and lymphoma treated with standard chemotherapy, and more recently in some cancer patients treated with hormone suppressive therapy. The use of published normative data to determine cognitive deficiency was employed as an efficient means to identify the prevalence of cognitive problems in this patient sample, but future research would benefit from the addition of a matched control group of noncancer patients.

Cognitive complaints reported by this study's participants largely failed to reflect their cognitive performance. That is, the TNP score was unrelated to total cognitive complaints (MAQ), and performance in specific cognitive domains was mostly unrelated to complaints in corresponding subjective domains. This finding corresponds with similar results reported by van Dam et al⁵ in BMT patients, as well as other studies³³ that have examined cognitive complaints in samples of non-BMT cancer patients. The exception in this study was that patients' complaints about their visual–spatial abilities corresponded with their objective performance in visual memory.

The general disconnect found here between patients' cognitive performance and their cognitive complaints may reflect the insensitivity of the cognitive tests used to detect subtle changes in patient's cognitive performance. That is, some patients may have experienced declines that failed to reach the level required to identify the presence of impairment. Longitudinal research tracking patients' cognitive performance both before and after BMT may be able to determine whether complaints reflect reductions in patients' cognitive abilities even when criteria for impairment are not met.

Patients in this study were significantly (P<0.01) or marginally (P<0.05) more likely to experience impairment in cognitive performance if they were older, male, less educated, and had lower estimated IQ scores. Younger patients were more likely to express cognitive complaints, but gender, education, and other demographic characteristics were unrelated to such complaints. These findings raise the possibility that patients at most risk for cognitive impairment following BMT may not be aware of or may not be outspoken about their deficits.

Both objective cognitive performance and subjective cognitive complaints were related to depressive symptomatology and average fatigue, as well as interference of fatigue with a patient's daily activities. Physical and mental quality of life were also related to cognitive complaints, but not to objective cognitive performance. These results confirm and extend results reported by van Dam et al,⁵ who identified psychological distress, including anxiety and depression, as related to subjective cognitive complaints following BMT. The present findings also indicate that fatigue is a strong correlate of cognitive complaints in these patients.

Certain methodological features of this study limit the conclusions that may be drawn. Most importantly, the cross-sectional nature of this study does not allow us to rule out the possibility that observed cognitive deficits were present prior to BMT. Recently, researchers have undertaken longitudinal studies to determine the course of cognitive difficulties following BMT,⁹ an important first effort. Additionally, the present study does not include neuroimaging data, thus it is unclear whether cognitive deficits observed here reflect specific brain abnormalities. Previous imaging studies have found diffuse white matter damage related to cyclosporin A,^{34, 35} a treatment for post-BMT graft-versus-host disease (GVHD); however, patients who receive autologous BMT are typically unaffected by GVHD and transplant type was not associated with cognitive deficits in this sample. Future studies should collect both cognitive and neuroimaging data in order to determine how post-BMT cognitive deficits may reflect underlying neurological problems. Additionally, the BMT patient sample is diverse in diagnosis and there were significant differences in treatment regimens before and during BMT, which may limit interpretation of our findings. On the other hand, it should be noted that the results of this study generalize primarily to breast cancer patients treated with autologous transplants. Given the more intensive medical regimen and different cancer diagnoses associated with allogeneic transplant, it is possible that patients who undergo allogeneic transplant would show a different pattern of results than those found in this study. Finally, this study compared patients' cognitive performance with published norms rather than a matched comparison group. As demonstrated by van Dam et al,⁵ matched comparison groups could be made up of healthy controls or of cancer patients who undergo less toxic treatments than BMT.

In conclusion, the results of this study indicate that there is a subset of BMT patients who experience cognitive deficits 6 months following BMT, with impairment most prevalent in the domains of psychomotor speed and executive function. As the study is cross-sectional, it is unclear whether the deficits observed here are long lasting. Longitudinal assessment of BMT patients' cognitive abilities would help determine possible recovery patterns of these deficits. Notably, nearly half of the participants demonstrated fully intact cognitive function in all domains assessed.

The results of this study suggest that patients who complain about cognitive deficits following BMT may not be the same as those who experience greater cognitive declines. Clinicians should be aware of the demographic risk factors found here to be significantly or marginally associated with greater cognitive impairment after BMT, including male gender, less education, and lower IQ. Since these patients may not be as forthcoming or, in some cases, as insightful about their cognitive declines following BMT, clinicians should consider more careful screening of BMT-related cognitive decline in these patients. Likewise, patients' complaints regarding their cognitive performance should be carefully considered. As cognitive complaints are highly related to patient distress levels, BMT clinicians should identify patients with complaints or concerns about their cognitive abilities in order to provide appropriate directions for intervention.

References

References

Acknowledgements

This work was supported by the American Cancer Society – Institutional Research Grant entitled 'Prevalence and Impact of Cognitive Sequelae of BMT', Margaret Booth-Jones, principal investigator. We are grateful to Christine Marsella and Lori Deitz for editorial support. We thank Karen K Fields, MD, former Program Leader of the Moffitt Cancer Center Blood and Marrow Transplant Program, for her support of psychosocial studies with her patients.