Response

We hope that our recent article on cancer-related cognitive impairment inspires discussion; therefore, we are very happy about the correspondence by Dr Sepehry and colleagues, who suggest considering microbleeds caused by chemotherapy as a possible mechanism behind cognitive dysfunction in cancer patients. We thank them for their contribution, albeit our results point in a completely different direction: Cancer-related cognitive impairment may be largely caused by psychological factors; moreover, its extent may have been overestimated in many studies because of methodological shortcomings.

Cognitive impairment has been observed independently of receipt of chemotherapy (1) in adult cancer patients. Having non–central nervous system cancer alone seems to be associated with some risk of cognitive impairment. From the largest study in this field to date, Vardy and colleagues very recently reported dramatically elevated rates of cognitive impairment in patients with colorectal cancer, irrespective of chemotherapy administration (2,3). We investigated whether cancer-related cognitive impairment is attributable to the potentially traumatic stress of being diagnosed with cancer. Stress and cognitive functioning are closely linked, and post-traumatic stress has been found to affect not only brain functioning but also brain structure (4). Our hypothesis was confirmed; however, we are even more intrigued by the fact that the observed differences of cognitive performance between the newly diagnosed breast cancer patients and the noncancer controls were only minimal.

Presumably, methodological issues are at the bottom of these contrasting findings. Most importantly, the choice of an appropriate comparison group is obviously crucial for the results.

It is well known that healthy individuals are more prone to participate in studies (5). The healthy volunteer effect certainly pertains to cancer patients, too, but may even be more pronounced in noncancer control subjects and normative samples predominantly self-selected from a very large group of eligible individuals. These volunteers may be particularly healthy in many social, mental, and emotional respects. In contrast, for cancer patients who are personally invited by their physicians, the threshold for participation may be low enough to be crossed also by those who are less open-minded, interested, and confident. Thus, cancer patients and their comparison groups, although apparently well matched, may systematically differ in inconspicuous features.

To minimize selection bias in our study, only a limited number of individuals were eligible for the control group (briefly, women who had undergone routine breast imaging with negative result at participating institutions), and they were enrolled by the same procedures as the breast cancer patients. Had we instead relied on data from normative samples, we would have reported completely different findings: The cancer patients performed worse than published norms on a number of tests; however, the control subjects also did. Both groups showed almost exactly the same deviations from the test norms.

We are therefore concerned that part of the evidence of cancer-related cognitive impairment, including previous work by our group (6), is compromised by methodological artifacts. In our view, it is time to ask whether any cancer-specific cognitive impairment—beyond the consequences of stress and life disruption concomitant with cancer—exists at all.