Understanding the impact of cognitive problems in everyday life of cancer survivors: The added value of proxy ratings.
Hanneke A. M. Lettinga
Master thesis Research Master Psychology University of Amsterdam
Netherlands Cancer Institute, dept. Psychosocial Research and Epidemiology Supervisor: Prof. Dr. Sanne Schagen
Second supervisor: Prof. Dr. Richard Ridderinkhof Word count (excl. abstract): 4.333
Abstract
Objective. Earlier research has shown a dissociation between self-perceived cognitive
complaints of cancer survivors and tested neuropsychological performance. The current goal was to investigate the added value of proxy ratings in predicting neuropsychological profile of breast cancer patients who underwent systemic therapy.
Methods. In a cross-sectional study, breast cancer patients treated with chemotherapy
(N=240) and healthy controls (N=66) underwent seven neuropsychological tests. Patient and proxy ratings of cognitive complaints in everyday life of the patient were assessed using the Cognitive Failures Questionnaire (CFQ). Depression and anxiety were measured with the Hopkins Symptom Checklist (HSCL), fatigue was measured with the Multidimensional Fatigue Inventory (MFI).
Results. The patient ratings revealed more cognitive complaints compared to healthy control
ratings (p = .044), though ratings of both groups fell within average range of cognitive failures in everyday life. Proxy ratings were lower than average (p < .001). The correlation between patient ratings and their cognitive profiles was low (r = .028). The correlation between proxy ratings and patient’s cognitive profiles was also low (r = .12). Depression and anxiety, and fatigue had a sizeable correlation with patient ratings (r = .51, r = .43
respectively) and no correlation with tested cognitive functioning (r = -.07, r = .02, respectively).
Conclusions. Proxy ratings on cognitive complaints in everyday life of breast cancer patients
have no added value in predicting the patient’s neuropsychological test profile. For now, in clinical practice, it remains important to assess both objective neuropsychological capacity and subjective patient reports.
1. Introduction.
Breast cancer survival rates are steadily increasing (Miller et al., 2016). Between 46-60% of breast cancer survivors report self-perceived cognitive complaints (hereafter referred to as cognitive complaints) months to years after treatment (Janelsins, Kesler, Ahles, & Morrow, 2014), while the prevalence rates of tested cancer treatment-related cognitive impairment varies between 20-60% - heavily depending on the definition of cognitive impairment (Janelsins et al., 2014). Cancer treatment-related cognitive impairment is characterized by impairment in memory, attention, and executive functioning (Raffa et al., 2006; Sleight, 2016). Studies typically find mild to moderate cognitive impairment related to cancer
treatment (Wefel, Kesler, Noll, & Schagen, 2015). However, even mild and subtle changes in cognitive functioning can have a substantial impact on quality of life.
At present, there is no agreed-upon method to determine everyday cognitive abilities and no single marker applies equally well to all patients when trying to assess the “real world” impact of cognitive impairment. Previous research has shown that there is a weak – if not a lack of – relationship between cognitive complaints and objective cognitive impairment as assessed with neuropsychological tests in cancer patients (Collins, Paquet, Dominelli, White, & MacKenzie, 2017; Paquet et al., 2017; Pullens, De Vries, & Roukema, 2010). This finding is also common in other populations associated with neurological diseases, such as the general elderly, post-stroke, patients with multiple sclerosis and in patients with
Parkinson’s disease (Bol, Duits, Hupperts, Verlinden, & Verhey, 2010; Koerts et al., 2012; Mendonça, Alves, & Bugalho, 2016; van Rijsbergen, Mark, de Kort, & Sitskoorn, 2014). In contrast, affective states – such as depressed mood and anxiety – are consistently found to have important and sizable contributions to cognitive complaints, while their influence on neuropsychological test performance is generally far less pronounced (Jenkins et al., 2006;
Pullens et al., 2010; Reid & MacLullich, 2006; Scholtissen-In de Braek, Hurks, van Boxtel, Dijkstra, & Jolles, 2011; Shilling & Jenkins, 2007). This means that for now it is important to assess both objective neuropsychological capacity and subjective patient reports to
distinguish between abnormal biological function of the brain due to injury to the brain or normal biological function of the brain in the presence of psychological distress, in order to contribute to the differential diagnosis procedure and the intervention plan that arises from this process (McClintock, Husain, Greer, & Cullum, 2010; Rohling, Green, Allen, & Iverson, 2002).
Another potential valuable source to learn more about the daily life functioning of the patient comes from clinical practice. In clinical practice, it is common to include an interview with a secondary source close to the patient whom provides information about the cognitive functioning of the patient in daily life (e.g. a hetero-anamnesis). The rationale behind this, is that the secondary source may provide a more objective view about the functioning of the patient. In scientific research this is sometimes implemented in the form of a proxy rating, where somebody close to the patient (e.g. significant other, relative, or nurse) fills in a questionnaire about the patient. (Parveen, 2016; Sneeuw et al., 1997; Sneeuw, Sprangers, & Aaronson, 2002) Proxy assessments are mainly implemented in patient groups that are (or will become) unable to fill in these questionnaires themselves. Therefore, proxy assessment research has mainly focused on the agreement rate between patient and proxy ratings (Sneeuw et al., 2002). A study in glioma patients showed that proxy ratings were more in concordance with the neuropsychological deficits observed on testing than the rating of the patients themselves, who underreported cognitive complaints compared to their test results (Taphoorn et al., 1992).
The goal of the current paper is to assess the added value of proxy ratings in
understanding the impact of cognitive problems in everyday life of breast cancer survivors. Cognitive functioning of the patient in daily life was rated by both the patient and proxy, and neuropsychological test data of the patients was collected. We investigated 1) the relationship between the patient and proxy rating of patients’ cognitive complaints, 2) the relationship between the proxy rating and the tested cognitive functioning of the patient, 3) the added value of the proxy rating compared to the patient rating in predicting the outcome of tested cognitive functioning of the patient. Multivariate normative comparisons were used to assess tested cognitive functioning, as this method approaches the evaluation of neuropsychological assessment in clinical practice.
2. Methods 2.1. Participants and procedure.
Data was collected from September 1998 to January 2002. Patients that underwent chemotherapy were recruited from seven different hospitals. Chemotherapy regimens included standard-dose FEC, CMF and high-dose CTC (see appendix A for detailed information on the chemotherapy regimens). The study was approved by the ethics committees of the different hospitals. Exclusion criteria were a) presence of metastatic disease or relapse, b) previous neurologic or psychiatric disorder (according to Diagnostic and Statistical Manual – IV criteria), c) substance addiction, d) use of medication believed to affect current cognitive functioning (e.g. antidepressants, anxiolytics, stimulants). An
exception from latter exclusion criteria is made for systemic therapy, because this was received by all patients as part of the trial. Healthy controls were recruited via patients: patients were asked to invite a female friend or relative of approximately similar age to serve as a control. All participants signed informed consent and were asked to provide a proxy respondent consisting of the significant other, a family member or close companion. In- and exclusion criteria for the control group were the same as for the patients with breast cancer,
with additional exclusion criterion of a medical history of malignancies. Seven of the 98 breast cancer patients that were treated with standard-dose FEC chemotherapy (7.1%) and 8 of the 94 breast cancer patients that received high-dose chemotherapy (8.5%) who met the in- and exclusion criteria, declined to participate; for most patients, the reason not to participate in the neuropsychological study was that they did not wanted to be reminded of their illness. All patients that did participate received the planned courses of chemotherapy. Assessments of the patients took place at three moments in time: before the start of the therapy (t1), 6 months (t2) and 12 months post-therapy (t3). To include as much data as possible, a cross-sectional design was implemented. All first measurement moments of patients after receiving treatment were included (see appendix A for detailed information about the dataset). The final sample included 240 patients and 66 healthy controls, see table 1 for demographic variables. The two groups were different regarding age (BC group, mean = 46.52, SD = 6.42, control group, mean = 48.58, SD = 6.29, p = .021) and estimated IQ (BC group, mean = 98.65, SD = 15.27, control group, mean = 104.52, SD = 14.73, p = .006) and similar regarding educational level, coded following the Verhage system (Bouma, Mulder,
Lindeboom, & Schmand, 2012) (BC group, mean = 4.0, SD = 1.80, control group, mean = 4.42, SD = 1.86, p = 0.09).
2.2. Measures.
2.2.1. Demographic measures
Date of birth was extracted from the medical record. Data on educational level was self-reported.
2.2.2. Self-perceived cognitive complaints
Cognitive complaints were measured with the cognitive failure questionnaire [CFQ] (Broadbent, Cooper, FitzGerald, & Parkes, 1982). The CFQ consists of 25 items about frequency of everyday cognitive ‘slips’ that are scored on a 5-point scale (0 = ‘never’; 4 = ‘very often). Total score is the sum score of all items, where a higher score indicates more cognitive failures. A slightly modified version is used for the proxy rating, in which the questions need to be answered for somebody else (i.e. the patient). The EORTC QLQ-30 (Scott et al., 2008) was also administered to the patients and controls to assess quality of life. Based on the content of the current study the cognitive subscale is reported.
2.2.3. Psychological factors.
Anxiety and depression symptoms were measured with the Hopkins Symptom Checklist [HSCL] (Derogatis, Lipman, Rickels, Uhlenhuth, & Covi, 1974). The HSCL consists of 25 items that are scored on a 4-point scale (1 = ‘not at all’, 4 = ‘extremely’). Total score is the average of all 25 items. Fatigue was measured using the multidimensional fatigue inventory [MFI] (Smets, Garssen, Bonke, & De Haes, 1995; Smets, Garssen, Cull, & De Haes, 1996). The MFI consists of 20 items that are scored on a 5-point Likert scale. (from 1 = ‘that is correct’, to 5 = ‘that is incorrect’). The MFI covers five domains: 1) general fatigue, 2) physical fatigue, 3) reduction in activity, 4) reduction in motivation and 5) cognitive fatigue. Based on the content of the current study general, physical and mental fatigue are reported.
2.2.4. Assessment of premorbid IQ.
Premorbid IQ was measured with the Dutch Adult Reading List (Schmand, Bakker, Saan, & Louman, 1991) (NLV). Participants have to read aloud 50 irregularly spelled Dutch words. The raw score is corrected for age and gender which provides an estimated premorbid IQ.
2.2.5. Neuropsychological Assessment
Tested cognitive functioning was assessed with a test battery consisting of seven neuropsychological tests. The battery covered seven domains 1) verbal memory, 2) visual memory, 3) verbal function, 4) motor function, 5) attention/concentration, 6) processing speed, 7) mental flexibility. Verbal memory was measured with the verbal learning and memory test (Delis, Kramer, Kaplan, & Ober, 1987). Visual memory was measured with the Wechsler memory scale (David Wechsler, 1945). Verbal functioning was measured with word fluency (Luteijn & Barelds, 2004). Motor functioning was measured with the tapping task of the FePsy (Alpherts & Aldenkamp, 1994). Attention and concentration was measured with the Eriksen Flanker task and Stroop task card I and card II (Eriksen, 1995; Hammes, 1971). Processing speed was measured with digit symbol substitution and Trailmaking test A (Reitan, 1958; D Wechsler, 2005). Mental flexibility was measured with the Stroop task card III and Trailmaking test B (Hammes, 1971; Reitan, 1958).
2.3. Statistical analysis
Data analysis took place using Statistical Package for Social Sciences (SPSS) 24.0 and the statistical programming language R (Team, 2016). To assess the relationship between patient and proxy rating and proxy rating and tested cognitive functioning, Pearson
correlations were calculated. Differences between patient and proxy ratings were evaluated using multilevel linear model accounting for coupled data. Differences between patients and healthy control ratings were evaluated using independent samples T-tests. For tested
cognitive functioning, raw scores were standardized to Z-scores based upon data from the healthy control group. Second, data was demographically standardized (that is, corrected for age and estimated IQ) using a simple regression. This was done by calculating the predicted test score for every participant – given their age and NLV score. The final test score consisted of the difference between the observed score and predicted score (i.e. the residual). With these final test scores, a general aggregated profile and an aggregated profile per domain of the individual test-scores was calculated. This was calculated by means of multivariate normative comparisons (MNC).
MNC is a method that compares an individual’s test scores against the distribution of the same scores in the control sample by creating one profile of multiple test scores. It can be seen as the multivariate version of the Student’s t-Test for one sample to compare scores of each single study participant against the distributions of the same scores of a control sample (Huizenga, Smeding, Grasman, & Schmand, 2007; Su et al., 2015). Because of this
aggregation of neuropsychological test-data, Type I errors were avoided that could otherwise arise due to multiple testing when making comparisons based on multiple test-outcomes (Huizenga et al., 2007). The MNC method provides both a dichotomous result (deviant profile or no deviant profile) and a continuous Hotelling’s T2 statistic for every patient that reflects their amount of ‘deviation from the norm’. To evaluate if proxy ratings provided added predictive value besides patient ratings to predict patients tested cognitive functioning, a two-stage hierarchical regression analysis was implemented. The dependent variable ‘cognitive functioning’ was expressed by the Hotelling’s T2 statistic. Sequential regression models were created by adding patient-CFQ score in step 1 and proxy-CFQ score in step 2. To determine whether newly added proxy-CFQ showed a significant improvement we looked at significant increase in R2 (the proportion of explained variance in tested cognitive
3. Results 3.1. Self-perceived cognitive complaints
Table 1 displays the demographic variables, cognitive complaints, fatigue, anxiety and depression for the patients, healthy controls and their proxies. For clarity, patient ratings of cognitive complaints and proxy ratings of cognitive complaints of the patient are hereafter referred to as patient ratings and proxy ratings respectively.
According to the norms of the CFQ, both the patients and the healthy controls rating fell within average range of cognitive failures in daily life. The proxy ratings of both the patients and the healthy controls fell from average to low amount (1 SD below the mean). A multilevel linear model, which accounted for nested data (that is, patient and proxy), showed that patient ratings were significantly higher (36.76 ± 13.06) compared to their proxy ratings (21.58, ± 13.44), F(1, 226.91) = 229.22, p < 0.001. For the healthy controls, the multilevel linear model showed that healthy control ratings were significantly higher (34.09 ± 8.20) compared to their proxy ratings (23.14 ± 10.02), F(1, 65.85) = 54.88, p < 0.001. The patient
Table 1
Mean and SD (in brackets) of Demographic Variables, Self-Perceived Cognitive Complaints and Psychological Factors for the Breast Cancer Patients, Healthy Controls and Proxies of Both Groups
BC (N = 240) HC (N = 66) p- value Proxy BC (N = 219) Proxy HC (N = 58) p-value Age 46.52 [6.42] 48.58 [6.29] .021 46.86 [10.54] 48.16 [11.95] .436 Educational level 3.99 [1.80] 4.42 [1.86] .089 4.76 [2.49] 4.47 [2.00] .431 Estimated IQ 98.65 [15.27] 104.52 [14.73] .006 Self-perceived cognitive complaints CFQ Total score 36.76 [13.06] 34.09 [8.20] .044 21.58 [13.44] 23.14 [10.02] .331 Absent mindedness 9.23 [4.15] 8.09 [3.04] .015 6.12 [4.26] 6.24 [3.24] .811 Social interactions 6.23 [2.32] 6.17 [1.76] .822 4.11 [2.72] 4.16 [2.15] .902 Names & Words 6.15 [2.26] 5.46 [1.75] .008 3.18 [2.29] 3.50 [1.58] .216 Orientation 3.63 [2.12] 3.53 [1.79] .722 2.07 [2.24] 2.52 [1.80] .160 Psychological Factors EORTC Cognitive functioning 77.34 [20.53] 91.92 [12.14] <.001 MFI <.001 General fatigue 10.61 [4.74] 7.26 [3.64] <.001 Physical fatigue 9.58 [4.54] 6.44 [3.20] <.001 Mental fatigue 10.58 [4.72] 7.23 [3.35] <.001 HSCL Total 15.58 [12.30] 9.02 [7.65] <.001 Anxiety 15.06 [13.26] 8.74 [7.62] <.001 Depression 15.94 [13.03] 9.21 [8.92] <.001
Note: Educational coded following the Verhage system (Bouma et al., 2012), estimated IQ based on Dutch Adult Reading List. Abbreviations: BC, breast cancer; HC, healthy control; CFQ, Cognitive Failures Questionnaire; EORTC, European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire; MFI, Multidimensional Fatique Inventory; HSCL, Hopkins Symptom Checklist.
ratings were also significantly higher (36.76 ± 13.06) compared to the healthy control ratings (34.09 ± 8.20), t(166.85) = 2.03, p = 0.044. There was no significant difference in scores between proxy ratings of the patients (21.58 ± 13.44) and proxy ratings of the healthy controls (23.14 ± 10.02), t(117.271) = -.98, p = .331. Pearson correlations were calculated to assess the relationship between patient and proxy ratings, see table 2. For overall cognitive functioning (CFQ total score) the correlation is weak to moderate. The correlation for the subscales ranged from weak to moderate.
3.2. Neuropsychological test performance (dichotomous)
Overall cognitive functioning, after adjusting for age and estimated IQ, was calculated by means of multivariate normative comparisons. This method yields a dichotomous measure (e.g. deviant profile yes/no) and a continuous hotelling’s T2 statistic (hereafter referred to as
‘cognitive profile’). The dichotomous measure revealed that 59 patients (24.5%) deviated significantly in cognitive profiles from the healthy controls. The same procedure was carried out for the healthy controls. With the ‘leave-one-out’ method each participant was once drawn from the norm group, after which their cognitive profile was also calculated by means of multivariate normative comparisons. The dichotomous measure revealed that no one in the healthy control group deviated significantly in cognitive profiles. This difference between groups was significant, t(304) = 4.62, p <.001.
3.3. Self-perceived cognitive complaints and neuropsychological test performance (continuous).
A Pearson correlation was calculated to evaluate the relationship between patient ratings and their overall cognitive profile, as well as the relationship between proxy ratings and the patients’ overall cognitive profiles. There was a nonsignificant relationship between the patient ratings and their overall cognitive profile (r = .028, n = 236, p = .34). There was a nonsignificant relationship between proxy ratings and patients’ overall cognitive profiles (r = .12, n = 219, p = .078). To determine whether proxy ratings add predictive value to patients’ cognitive profiles, a two-stage hierarchical regression was conducted with patients’ cognitive profile as dependent variable. Table 3 shows a summary of the two models. Patient ratings were added in step 1 and proxy ratings were added in step 2. Table 3 shows a summary of the two models. The hierarchical multiple regression revealed that at stage one, patient ratings did not significantly contribute to the regression model F(1, 213) = .164, p = .686, and accounted for 0.1% of the variation in cognitive profiles. Introducing the proxy ratings explained only an additional 1.4% of the variance in cognitive profiles, F(1, 212) = 2.958, p = .087.
Table 2
Correlations between Patient and Proxy Ratings of Patients’ Cognitive Complaints [CFQ total score and subscales]
Patient Rating (N = 240)
Proxy Rating (N =220) 1 2 3 4 5
1. CFQ total score .36** .38** .22** .30** .21** 2. CFQ absent mindedness .37** .20** .27** .24**
3. CFQ social interactions .26** .17* .14*
4. CFQ names and words .32** .26**
5. CFQ orientation .20**
3.4. Self-perceived cognitive complaints and neuropsychological test performance (dichotomous and continuous) per domain.
The neuropsychological tests that were administered covered seven different cognitive domains. By means of multivariate normative comparisons, the cognitive profiles were also calculated for each cognitive domain, for patients and healthy controls. The absolute numbers of deviating cognitive profiles yielded by the dichotomous measure can be found in table 4. With the continuous cognitive profiles, the two-step hierarchical regression model as described above was repeated for every cognitive domain. None of the regression models explained significant variation in the cognitive profiles per cognitive domain (see appendix B for a summary of the models).
3.5. Self-perceived cognitive complaints and psychological factors: fatigue, anxiety and depression.
To evaluate the relationship between patient and proxy ratings and psychological factors such as anxiety and depression (HSCL) and fatigue (MFI; general scale), a Pearson correlation was calculated. There was a significant relationship between patient ratings and anxiety and depression (r = .51, n = 235, p < .001). There was also a significant relationship between proxy ratings and anxiety and depression, however the relationship was less strong (r =.28, n
Table 3
Summary of Hierarchical Regression Analysis for Patient and Proxy ratings on Cognitive Complaints predicting Patients’ Cognitive Profile
Variable b t sr2 R2 DR2 Step 1 .001 .001 Patient CFQ .028 .405 .028 Step 2 .015 .014 Patient CFQ -.017 -.233 -.016 Proxy CFQ .126 1.720 .117
Note: N = 219. *p < .05, **p < .01, ***p< .001. Abbreviations: CFQ, Cognitive Failures Questionnaire [total score]; b, standardized coefficient; t, t-statistics; sr2, semi partial correlation coefficient, R2, coefficient of determination; DR2, difference in R2 between two models.
Table 4
The Absolute Numbers and Percentages (in brackets) of Deviant Cognitive Profiles in General and Per Domain, for Breast Cancer Patients and Healthy Controls
BC (N = 240)
HC
(N = 66) p-value General cognitive profile 59 (24.5) 0 <.001 Cognitive domain Verbal memory 26 (10.8) 5 (7.5) .439 Visual memory 30 (12.5) 1 (1.5) <.001 Verbal function 7 (2.9) 4 (6.0) .322 Motor function 26 (10.8) 3 (4.5) .057 Attention/Concentration 43 (18.1) 2 (3.0) <.001 Processing speed 27 (11.3) 3 (4.5) .041 Mental flexibility 33 (13.9) 2 (3.0) .001 Abbreviations: BC, breast cancer; HC, healthy controls.
= 217, p < .001). There was a significant relationship between patient ratings and fatigue (r = 0.43, n = 236, p < .001). There was also a significant relationship between proxy ratings and fatigue, however the relationship was less strong (r = .15, n = 218, p = .028).
3.6. Psychological factors: fatigue, anxiety and depression, and neuropsychological test performance (continuous).
A Pearson correlation was calculated to evaluate the relationship between psychological factors such as anxiety and depression (HSCL) and fatigue (MFI; general scale), and the patients overall cognitive profile. There was a nonsignificant negative relationship between anxiety and depression and patient’s cognitive profiles (r = -.07, n = 238, p = .15). There was also a nonsignificant relationship between fatigue and the patient’s cognitive profiles (r = 0.02, n = 239, p = .39).
4. Discussion
The overall aim of the current study was to assess the added value of proxy ratings in understanding the impact of cognitive problems in everyday life of cancer survivors. By means of multivariate normative comparisons, a recent sensitive method that more closely resembles clinical practice, almost one fourth of the patients showed a cognitive deviating profile. Replicating previous research, breast cancer patients reported more self-perceived cognitive complaints as measured with the cognitive failures questionnaire, than healthy controls (Paquet et al., 2017; Pullens et al., 2010) – though the reports of both groups fell within the average range of cognitive failures in everyday life. Proxies (from both patients and healthy controls) however, reported a lesser – and lower than average – amount of cognitive slips in everyday life by the patients, hence the relationship between patients self-perceived cognitive complaints and those from the proxy turned out to be weak to moderate. There was no relationship between patient and proxy rating of patients’ cognitive complaints with the cognitive profiles of the patients. This finding holds for the overall cognitive profile, as well as for the different cognitive domains. Replicating previous research, psychological factors had a sizeable contribution in patients self-perceived cognitive complaints and no contribution to tested cognitive functioning.
In the current study, self-perceived cognitive complaints – irrespective of whether they are reported by the patient or proxy– are independent from tested cognitive functioning. Our hypothesis, based on brain tumor studies, that the proxy’s view could provide a more “objective: (i.e. more related to cognitive performance as assessed with neuropsychological tests) did not hold. When evaluating the proxy ratings, the roles in the current study seemed to be turned around compared to previous research. That is, usually proxies report a higher rate of cognitive problems in patients with severely impaired disease insight (Taphoorn et al., 1992). However, in the current study proxies reported a substantial lesser amount of cognitive complaints, both in the patient group as in the control group. An alternative explanation could be that some of the symptoms are not ‘visible’ enough for the proxy, and are therefore hard to make a judgement about. This is somewhat reflected by the correlations between patient and proxy ratings, where the subscales absent-mindedness and names and words seem to have the highest correlations. To investigate this further, we exploratory created another subscale, by adding up these two subscales and repeat the regression analysis. The outcomes did not differ from our main analysis and thus it seems that based on this analysis, that this is not a
sufficient explanation. For now, it remains unclear why proxies in this study seem to underreport in both the patient and the healthy control group.
A possible methodological issue, is that cognitive complaints throughout this article are based on scores on the cognitive failures questionnaire. However, based on this questionnaire, patients report an average amount of cognitive complaints in daily life, which is not
according to expectation and according to previous research (Collins et al., 2017). At the same time, the cognitive scale of the questionnaire for quality of life did show the expected result – that is, the breast cancer patients reported worse cognitive functioning than the controls. Following this result, one must consider the possibility that the cognitive failures questionnaire might not be sensitive enough for measuring self-perceived cognitive complaints. This idea is supported by the finding that results on the cognitive failures questionnaire tend to be influenced by non-cognitive factors, such as personality traits, psychological factors and genes (Carrigan & Barkus, 2016; Ponds & Jolles, 1996). Also, one would expect an effect on scores of the cognitive failures questionnaire due to age-related cognitive decline, yet this is not supported by previous research (Carrigan & Barkus, 2016; Ponds, Van Boxtel, & Jolles, 2006). Some authors recommend to use the CFQ as indicator of where to focus on during treatment, that is, if a patient reports above average cognitive failures without a deviating cognitive profile, the treatment should focus on the perception of the patient. (Ponds et al., 2006). We recommend to use other measurement tools for self-perceived cognitive complaints in the future. For example, the M.D. Anderson Symptom Inventory [MDASI] could be included, where symptoms are rated on a 11-point scale that indicates presence of a symptom for the last 24 hours (Cleeland et al., 2000). An adapted version has been used in previous research in multiple myeloma patients, that showed excellent correlations with cognitive subscales of other questionnaires (Jones et al., 2013).
To conclude, the current study showed that almost one fourth of patients that received adjuvant chemotherapy had deviating cognitive profile. Self-perceived cognitive complaints, nor from the patient themselves nor from their proxies, appear to be a relevant predictor of tested cognitive functioning of cancer survivors. This finding suggests we should treat self-perceived cognitive complaints and tested cognitive functioning as two different entities. The fact that measures of self-reported and tested cognitive functioning are weakly correlated does not (dis)qualify one over the other, or make one more “true” than the other in all situations. It shows that these measures reflect distinct constructs, thus cannot be considered as proxies, and should not be applied as such. For now, in clinical practice, it remains
important to assess both objective neuropsychological capacity and subjective patient reports to provide the best care as possible.
References
Alpherts, W., & Aldenkamp, A. (1994). FePsy: the iron psyche. Heemstede: Instituut voor epilepsiebestrijding.
Bol, Y., Duits, A. A., Hupperts, R. M., Verlinden, I., & Verhey, F. R. (2010). The impact of fatigue on cognitive functioning in patients with multiple sclerosis. Clinical
rehabilitation, 24(9), 854-862.
Bouma, A., Mulder, J., Lindeboom, J., & Schmand, B. (2012). Handboek neuropsychologische diagnostiek.-2e herz. dr.
Broadbent, D. E., Cooper, P. F., FitzGerald, P., & Parkes, K. R. (1982). The cognitive failures questionnaire (CFQ) and its correlates. British journal of clinical psychology, 21(1), 1-16.
Carrigan, N., & Barkus, E. (2016). A systematic review of cognitive failures in daily life: Healthy populations. Neuroscience & Biobehavioral Reviews, 63, 29-42.
Cleeland, C. S., Mendoza, T. R., Wang, X. S., Chou, C., Harle, M. T., Morrissey, M., & Engstrom, M. C. (2000). Assessing symptom distress in cancer patients. Cancer, 89(7), 1634-1646.
Collins, B., Paquet, L., Dominelli, R., White, A., & MacKenzie, J. (2017). Metamemory function in chemotherapy-treated patients with breast cancer: an explanation for the
dissociation between subjective and objective memory measures? Psycho-Oncology, 26(1), 109-117.
Delis, D., Kramer, J., Kaplan, E., & Ober, B. (1987). California verbal learning test (CVLT). San Antonio: The Psychological CorporationÕ. In: Harcourt Brace & Company. Derogatis, L. R., Lipman, R. S., Rickels, K., Uhlenhuth, E. H., & Covi, L. (1974). The
Hopkins Symptom Checklist (HSCL): A self-report symptom inventory. Systems Research and Behavioral Science, 19(1), 1-15.
Eriksen, C. W. (1995). The flankers task and response competition: A useful tool for investigating a variety of cognitive problems. Visual Cognition, 2(2-3), 101-118. Hammes, J. (1971). De Stroop kleur-woord test: handleiding [The Stroop color-word test:
manual]. 1971. Netherlands: Lisse: Swets & Zeitlinger Google Scholar. Huizenga, H. M., Smeding, H., Grasman, R. P., & Schmand, B. (2007). Multivariate
normative comparisons. Neuropsychologia, 45(11), 2534-2542.
Janelsins, M. C., Kesler, S. R., Ahles, T. A., & Morrow, G. R. (2014). Prevalence, mechanisms, and management of cancer-related cognitive impairment. Int Rev Psychiatry, 26(1), 102-113. doi:10.3109/09540261.2013.864260
Jenkins, V., Shilling, V., Deutsch, G., Bloomfield, D., Morris, R., Allan, S., . . . Sadler, G. (2006). A 3-year prospective study of the effects of adjuvant treatments on cognition in women with early stage breast cancer. British journal of cancer, 94(6), 828.
Jones, D., Vichaya, E. G., Wang, X. S., Sailors, M. H., Cleeland, C. S., & Wefel, J. S. (2013). Acute cognitive impairment in patients with multiple myeloma undergoing
autologous hematopoietic stem cell transplant. Cancer, 119(23), 4188-4195.
Koerts, J., Van Beilen, M., Leenders, K. L., Brouwer, W. H., Tucha, L., & Tucha, O. (2012). Complaints about impairments in executive functions in Parkinson's disease: The association with neuropsychological assessment. Parkinsonism & related disorders, 18(2), 194-197.
Luteijn, F., & Barelds, D. P. H. (2004). GIT2: Groninger Intelligentie Test 2: Harcourt Test Publishers.
McClintock, S. M., Husain, M. M., Greer, T. L., & Cullum, C. M. (2010). Association between depression severity and neurocognitive function in major depressive disorder: a review and synthesis. Neuropsychology, 24(1), 9.
Mendonça, M. D., Alves, L., & Bugalho, P. (2016). From subjective cognitive complaints to dementia: who is at risk?: a systematic review. American Journal of Alzheimer's Disease & Other Dementias®, 31(2), 105-114.
Miller, K. D., Siegel, R. L., Lin, C. C., Mariotto, A. B., Kramer, J. L., Rowland, J. H., . . . Jemal, A. (2016). Cancer treatment and survivorship statistics, 2016. CA: a cancer journal for clinicians, 66(4), 271-289.
Paquet, L., Verma, S., Collins, B., Chinneck, A., Bedard, M., & Song, X. (2017). Testing a novel account of the dissociation between self-reported memory problems and memory performance in chemotherapy-treated breast cancer survivors. Psycho-Oncology.
Parveen, S. (2016). Comparison of self and proxy ratings for motor performance of individuals with Parkinson disease. Brain and cognition, 103, 62-69.
Ponds, R., & Jolles, J. (1996). Memory complaints in elderly people: The role of memory abilities, metamemory, depression, and personality. Educational Gerontology: An International Quarterly, 22(4), 341-357.
Ponds, R., Van Boxtel, M., & Jolles, J. (2006). De Cognitive Failure Questionnaire als maat voor subjectief cognitief functioneren. Tijdschrift voor neuropsychologie, 1(2), 37-45. Pullens, M. J., De Vries, J., & Roukema, J. A. (2010). Subjective cognitive dysfunction in
Raffa, R., Duong, P., Finney, J., Garber, D., Lam, L., Mathew, S., . . . Jen Weng, H. F. (2006). Is ‘chemo-fog’/‘chemo-brain’caused by cancer chemotherapy? Journal of Clinical Pharmacy and Therapeutics, 31(2), 129-138.
Reid, L. M., & MacLullich, A. M. (2006). Subjective memory complaints and cognitive impairment in older people. Dementia and geriatric cognitive disorders, 22(5-6), 471-485.
Reitan, R. M. (1958). Validity of the Trail Making Test as an indicator of organic brain damage. Perceptual and motor skills, 8(3), 271-276.
Rohling, M. L., Green, P., Allen, L. M., & Iverson, G. L. (2002). Depressive symptoms and neurocognitive test scores in patients passing symptom validity tests. Archives of clinical neuropsychology, 17(3), 205-222.
Schmand, B. A., Bakker, D., Saan, R. J., & Louman, J. (1991). De Nederlandse Leestest voor Volwassenen: een maat voor het premorbide intelligentieniveau. Tijdschrift voor Gerontologie en Geriatrie.
Scholtissen-In de Braek, D. M., Hurks, P. P., van Boxtel, M. P., Dijkstra, J. B., & Jolles, J. (2011). The identification of attention complaints in the general population and their effect on quality of life. Journal of Attention Disorders, 15(1), 46-55.
Scott, N., Fayers, P., Aaronson, N., Bottomley, A., de Graeff, A., Groenvold, M., . . . Sprangers, M. (2008). EORTC QLQ-C30. Reference values. Brussels: EORTC. Shilling, V., & Jenkins, V. (2007). Self-reported cognitive problems in women receiving
adjuvant therapy for breast cancer. European Journal of Oncology Nursing, 11(1), 6-15.
Sleight, A. (2016). Coping with cancer-related cognitive dysfunction: a scoping review of the literature. Disability and rehabilitation, 38(4), 400-408.
Smets, E., Garssen, B., Bonke, B. d., & De Haes, J. (1995). The Multidimensional Fatigue Inventory (MFI) psychometric qualities of an instrument to assess fatigue. Journal of psychosomatic research, 39(3), 315-325.
Smets, E., Garssen, B., Cull, A., & De Haes, J. (1996). Application of the multidimensional fatigue inventory (MFI-20) in cancer patients receiving radiotherapy. British journal of cancer, 73(2), 241-245.
Sneeuw, K. C., Aaronson, N. K., Osoba, D., Muller, M. J., Hsu, M.-A., Yung, W. A., . . . Newlands, E. S. (1997). The use of significant others as proxy raters of the quality of life of patients with brain cancer. Medical care, 35(5), 490-506.
Sneeuw, K. C., Sprangers, M. A., & Aaronson, N. K. (2002). The role of health care providers and significant others in evaluating the quality of life of patients with chronic disease. Journal of clinical epidemiology, 55(11), 1130-1143.
Su, T., Schouten, J., Geurtsen, G. J., Wit, F. W., Stolte, I. G., Prins, M., . . . Majoie, C. B. (2015). Multivariate normative comparison, a novel method for more reliably detecting cognitive impairment in HIV infection. Aids, 29(5), 547-557.
Taphoorn, M., Heimans, J., Snoek, F., Lindeboom, J., Oosterink, B., Wolbers, J., & Karim, A. (1992). Assessment of quality of life in patients treated for low-grade glioma: a preliminary report. Journal of Neurology, Neurosurgery & Psychiatry, 55(5), 372-376.
Team, R. C. (2016). R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2014. R Foundation for Statistical Computing. In.
van Rijsbergen, M. W., Mark, R. E., de Kort, P. L., & Sitskoorn, M. M. (2014). Subjective cognitive complaints after stroke: a systematic review. Journal of Stroke and Cerebrovascular Diseases, 23(3), 408-420.
Wechsler, D. (2005). WAIS-III nl. Wechsler adult intelligence scale WAIS-III. Dutch version. Manual. In: Amsterdam: Harcourt Test Publishers.
Wefel, J. S., Kesler, S. R., Noll, K. R., & Schagen, S. B. (2015). Clinical characteristics, pathophysiology, and management of noncentral nervous system cancer-related cognitive impairment in adults. CA Cancer J Clin, 65(2), 123-138.
Appendix A
Information regarding chemotherapy regimens and dataset
Chemotherapy regimens: Chemotherapy regimens included standard-dose FEC (5 courses of:
5-fluorouracil, 500 mg/m2 intravenously; epidoxorubicin, 90-120 mg/m2 intravenously;
cyclophosphamide, 500 mg/m2 intravenously, repeated every three weeks), CMF (6 courses of
cyclophosphamide, 100 mg/m2 orally on days 1 to 14: methotrexate 40 mg/m2 intravenously on days 1
and 8; 5-fluorouracil 600 mg/m2 intravenously on days 1 and 8, repeated every four weeks) and
high-dose CTC (cyclophosphamide, 6 g/m2 intravenously; thiotepa, 480 mg/m2 intravenously; carboplatin,
1.6 g/m2 intravenously, divided over four days, given after 4 courses of FEC chemotherapy).
In total, 240 participants are included in the current study, that received either FEC (n=91), CTC (n=86) or CMF (n=63), and 66 healthy controls.
Dataset. Assessments of the patients took place at three moments in time: before the start of the therapy (t1), 6 months (t2) and 12 months post-therapy (t3). However, not all patients were tested at t1. Most patients were included after t1 and some patients were included after t2. This led to three different groups: patients that were first tested at t1, patients that were first tested at t2 (and subsequently were also tested at t3) and patients that were first tested at t3. The healthy controls were tested twice with an interval of 6 months (t1 and t2). In the current study, we implemented a cross-sectional design. To include as much data as possible and keep groups as equal as possible, we only use data of all first measurement moments of patients, after receiving some form of treatment (see figure 1). Therefore, all patients that were tested at t1 were excluded. Evidently, the data from the healthy control group at t1 will be included, since this group did not receive any treatment. According to these criteria, we included 306 participants, of which 240 patients and 66 healthy controls.
Figure A1
Flow chart that shows the construction of the final dataset. In the top column are the total numbers of patients that were in the initial dataset. The middle three columns represent the three measurement moments. The numbers represent the patients that were tested for the first time on that particular measurement moment. The final column shows which data was included in the final dataset (patient data from t2 and t3, and healthy control data from t1)
Appendix B
Hierarchical Regression Analysis per Cognitive Domain
Table B1
Summary of Hierarchical Regression Analysis for Patient and Proxy Ratings predicting Patients’ Cognitive Profile per Domain.
Domain b t sr2 R2 DR2 Verbal memory Step 1 .002 .002 Patient CFQ .005 .656 .045 Step 2 .019 .017 Patient CFQ .000 -.059 -.004 Proxy CFQ .015 .138 .129 Visual memory Step 1 .000 .000 Patient CFQ .000 .024 .002 Step 2 .015 .015 Patient CFQ -.009 -.610 -.042 Proxy CFQ .025 1.774 .121 Verbal functioning Step 1 .008 .008 Patient CFQ .006 .1.351 .092 Step 2 .009 .000 Patient CFQ .006 1.159 .079 Proxy CFQ .001 .281 .019 Motor functioning Step 1 .002 .002 Patient CFQ .005 .694 .047 Step 2 .004 .002 Patient CFQ .007 .869 .060 Proxy CFQ -.005 -.621 -.043 Processing speed Step 1 .005 .005 Patient CFQ -.013 .-1.039 -.071 Step 2 .006 .001 Patient CFQ -.014 -1.107 -.076 Proxy CFQ .005 .388 .027 Attention/ Concentration Step 1 .014 .014 Patient CFQ .019 1.743 .119 Step 2 .150 .008 Patient CFQ .013 1.152 .078 Proxy CFQ .015 1.345 .091 Mental flexibility Step 1 .010 .010 Patient CFQ .029 .1.472 .100 Step 2 .024 .014 Patient CFQ .016 .757 .051 Proxy CFQ .036 1.751 .119
Note: N = 219. *p < .05, **p < .01, ***p< .001. Abbreviations: CFQ, Cognitive Failures Questionnaire [total score]; b, standardized coefficient; t, t-statistics; sr2, semi partial correlation coefficient, R2, coefficient of determination; DR2, difference in R2 between two models.
