Researchers examining cognitive side effects of chemotherapy have emphasized the importance of using longitudinal designs that include pretreatment neuropsychological assessments. However, in order to interpret longitudinal change, it is critical to understand whether cognitive performance is lower than expected prior to the initiation of treatment for a subgroup of breast cancer patients. Data from the current study suggest that when examining mean pretreatment performance across multiple domains of cognitive functioning, Stage 0 patients do not differ from healthy controls, and Stage 1–3 patients differ significantly from healthy controls only in the Reaction Time Domain. Importantly, despite the statistical significance, performance on the tests that make up the Reaction Time Domain are within normal limits for the Stage 1–3 patients based on published norms. Therefore, based on this analysis, one would conclude that there are minor or no pretreatment differences in neuropsychological performance between patients with invasive or noninvasive breast cancer and healthy controls. Consistent with this interpretation, there were no group differences on the self-report measure of cognitive problems, the MASQ.
Examination of the percentage of patients who fell below the cutoff of lower than expected cognitive performance is important. However, when comparing the performance of patients to normative data, there is the problem of the unknown misclassification error rate. Inclusion of the healthy control group allowed us to empirically examine the misclassification error rate. This analysis was based on the number of domains examined and the intercorrelations among the domains, and we were able to modify the definition of lower than expected cognitive performance so that the error rate of classification as impaired would be ~5% for the healthy control group. Utilizing this definition, we found that patients with Stage 0 disease did not differ from healthy controls, but that patients with Stage 1–3 breast cancer were significantly more likely to be categorized as having lower than expected cognitive performance as compared to healthy controls.
Interestingly, patients categorized as having lower than expected cognitive performance did not self-report more cognitive problems, suggesting that they have not perceived a change in cognitive abilities since the diagnosis of cancer and/or that they have developed strategies for compensating for their deficits. Lack of a relationship between subjective reports of cognitive functioning and performance on neuropsychological testing is not uncommon [2
]; however, the most common pattern is that patients self-report higher levels of cognitive problems relative to healthy controls that are not identified on formal testing [20
]. In this study the opposite pattern was seen.
The significantly greater rate of lower than expected cognitive performance for the Stage 1–3 breast cancer patients is consistent with previous reports [20
]. However, the correction for misclassification error resulted in a lower rate (22%) than has been previously reported in a similar population [35%, 20
]. On one level, it could be argued that the absolute percentage is less relevant than the relative difference between patients and controls. However, in order to examine factors that differentiate between patients who exhibit lower than expected cognitive performance and those who do not prior to adjuvant treatment, it is important to have an accurate classification strategy. We provide a method for controlling for misclassification error that resulted in a more stringent definition of lower than expected cognitive performance, hence the lower rate. However, it is important to note that we are not advocating that this is the “correct” definition, since the error rate will vary depending on the specific tests administered, their sensitivity and specificity for a given classification, the number of tests, and the intercorrelations among the tests. Furthermore, a cut-off based on high specificity (95%) may attenuate sensitivity to subtle deficits.
Importantly, patients classified as having lower than expected cognitive performance did not differ from those classified as normal on measures of depression, anxiety or fatigue. Additionally, the groups did not differ on any other measured variable, including age or education, menstrual status, length of general anesthesia or blood work.
The lack of a difference in the lower than expected cognitive performance rate between Stage 0 patients and healthy controls was a somewhat unexpected, but potentially important finding. The obvious hypothesis would be that patients with Stage 0 disease were less distressed than patients with Stage 1–3 cancer; however, the data demonstrated that both patient groups reported significantly higher levels of depression, anxiety and fatigue as compared to the healthy control group and that the cancer groups did not differ from each other on any of these variables.
The pattern of results raises the question of why a larger subgroup of patients with invasive breast cancer might perform in the lower than expected range prior to treatment as compared to healthy controls and patients with noninvasive cancer. Paraneoplastic syndromes, which include cognitive symptoms, have been reported in patients diagnosed with breast cancer [36
]. Although these syndromes are relatively rare, they may partially explain the pattern of results seen in this study. Alternatively, cancer and development of cognitive problems may share common risk factors. A study of twins discordant for cancer found evidence that cancer is a risk factor for the development of cognitive problems after the age of 65 [38
]; however, there are methodological limitations to this study [39
]. Although the factors that confer risk to both the development of cancer and cognitive problems have not been identified, several potential candidates have been suggested [40
]. Stimulation of proinflammatory cytokines have been linked to the development of cancer [41
] and neurocognitive disorders [43
]. Similarly, DNA damage and deficiencies in DNA repair mechanisms have been related to increased risk for the development of cancer [46
] and neurodegenerative disorders [48
Limitations of the study include the relatively small sample size for the Stage 0 group; therefore, the difference in the rate of lower than expected cognitive performance seen between the invasive and noninvasive cancer groups requires replication with a larger sample. Additionally, we recognize that the sample size for the study was low for conducting a formal factor analysis. However, the intent was not to determine the ideal factor/domain structure for the combination of tests administered, but to guide the neuropsychologists when deciding in which domain to include a specific test based on shared variance.
In conclusion, patients with invasive breast cancer were more likely to be categorized as exhibiting lower than expected cognitive performance based on neuropsychological testing, even after controlling for misclassification error, as compared to patients with noninvasive cancer and healthy controls. Interestingly, patients categorized as having lower than expected cognitive performance did not self-report more cognitive problems and factors such as depression, anxiety, fatigue, type of surgery, length of anesthesia, and biological markers were not associated with lower than expected cognitive performance. Future research is warranted to test whether other variables noted above, such as elevated levels of proinflammatory cytokines and/or DNA damage are related to cognitive performance in women with invasive breast cancer prior to adjuvant treatment.