Approximately 13 million adult cancer survivors live in the United States; about 2 million of them were diagnosed over 20 years ago (National Cancer Institute, 2012). Current 5-year survival rates are close to 70% and rising, so the number of people living with the physical and mental health consequences of cancer and cancer treatment will increase. One of these consequences is cognitive decline, popularly known as chemo-brain. The papers in this symposium reflect some of the approaches used to understand the nature and course of cognitive decline in women with breast cancer.

Oncology research has long recognized that cognition may be affected by cancer. This was attributed to distress until 1978, when cognitive symptoms in breast cancer patients were attributed to “organic brain syndrome” (Levine, Silverfarb, & Lipowski, Reference Levine, Silberfarb and Lipowski1978). The first reports documenting greater cognitive deficits in cancer patients treated with chemotherapy than in patients who did not receive chemotherapy were published in the early 1980s (Greer and Silberfarb, Reference Greer and Silberfarb1982; Oxman and Silberfarb, Reference Oxman and Silberfarb1980; Silberfarb, Reference Silberfarb1983). Following a lull, Wieneke and Dienst (Reference Wieneke and Dienst1995) showed that breast cancer survivors treated with chemotherapy performed more than two standard deviations below test norms on tests of memory, mental flexibility, processing speed, attention, visuospatial ability, and/or motor function. Performance was correlated with length of treatment but not depression or time since treatment. Since then, studies examining cancer, chemotherapy, and cognition have increased exponentially.

However, the field is relatively new and there is a great deal we do not understand. Although many studies using self-reported concerns distinguish people who received chemotherapy from those who have not (Pullens, De Vries, & Roukema, Reference Pullens, De Vries and Roukema2010), cohort studies using standardized neuropsychological measures often reveal “average” abilities relative to population norms (Anderson-Hanley, Sherman, Riggs, Agocha, Compas, Reference Anderson-Hanley, Sherman, Riggs, Agocha and Compas2003; Correa & Ahles, Reference Correa and Ahles2008; Stewart, Bielajew, Collins, Parkinson, & Tomiak, Reference Stewart, Bielajew, Collins, Parkinson and Tomiak2006; Wefel & Schagen, Reference Wefel and Schagen2012). Discrepancies between patients’ complaints and objective test performance (Bender et al., Reference Bender, Pacella, Sereika, Brufsky, Vogel, Rastogi, Casillo and Ryan2008; Castellon et al., Reference Castellon, Ganz, Bower, Petersen, Abraham and Greendale2004; Cimprich, So, Ronis, & Trask, Reference Cimprich, So, Ronis and Trask2005; Hermelink et al., Reference Hermelink, Untch, Lux, Kreienberg, Beck, Bauerfeind and Münzel2007) are frustrating for cancer survivors who express concerns about concentration, memory, processing speed, word-finding, decision making, and problem solving (Pullens et al., Reference Pullens, De Vries and Roukema2010). These changes hinder people from returning to work, school, or household obligations (Oberst, Bradley, Gardiner, Schenk, & Given, Reference Oberst, Bradley, Gardiner, Schenk and Given2010), and affect psychological well-being (Boykoff, Moieni, & Subramanian, Reference Boykoff, Moieni and Subramanian2009) as well as relationships with the medical team, family and friends (Munir, Burrows, Yarker, Kalawsky, & Bains, Reference Munir, Burrows, Yarker, Kalawsky and Bains2010). In young adults with cancer, 27% fail to return to school or work 15–35 months after diagnosis and 30% report memory, attention and processing speed problems (Parsons et al., Reference Parsons, Harlan, Lynch, Hamilton, Wu, Kato and Keegan2012). Although cognitive decline may be one factor contributing to these outcomes, other reasons may underlie the discrepancy between subjective concerns and performance. Wefel and Schagen (this issue) provide compelling evidence that quiets suspicions about motivation or secondary gain (e.g., perhaps due to return to work, disability support, or other issues). Their analysis of large samples of breast cancer patients showed no evidence of noncredible performance on performance validity testing.

Another challenge is making sense of variable results across studies. The divergence between subjective and objective deficits indicates that they are not identical constructs; direct measurement of performance is important to characterize abilities rather than only relying on patient report. Yet across cross-sectional and longitudinal studies that use neuropsychological tools, either no impairment is found, or there is variability in which cognitive domains are impaired and the magnitude of impairment (see Vardy & Tannock, Reference Vardy and Tannock2007 for review).

These discrepancies may result from methodological differences, including but not limited to:

1. 1. Different tests evaluating the same domain. Some tests are more sensitive to practice than others, masking subtle changes in cognition in longitudinal designs (Jansen, Miaskowski, Dodd, & Dowling, Reference Jansen, Miaskowski, Dodd and Dowling2007).

2. 2. Different criteria and statistical methods determining impairment or reliable change. Discrepant findings may be related to data categorization (e.g., grouping domains to capture “global” scores) or analysis (e.g., impairment cutoff scores, such as one or two standard deviations below comparison). There are also differences in how reliable change is best defined (Bläsi et al., Reference Bläsi, Zehnder, Berres, Taylor, Spiegel and Monsch2009; Chelune, Naugle, Lüders, Sedlak, & Awad, Reference Chelune, Naugle, Lüders, Sedlak and Awad1993; Iverson, Reference Iverson2001; Jacobson & Truax, Reference Jacobson and Truax1991; McSweeny, Naugle, Chelune, & Lüders, Reference McSweeny, Naugle, Chelune and Lüders1993).

3. 3. Choice of comparison groups. Results may vary based on whether patients are compared to norms, healthy non-cancer controls, or cancer patients who do not receive chemotherapy (Castellon et al., Reference Castellon, Ganz, Bower, Petersen, Abraham and Greendale2004; Collins, Mackenzie, & Kyeremanteng, Reference Collins, Mackenzie and Kyeremanteng2013; Collins, this issue; Schagen et al., Reference Schagen, Muller, Boogerd, Rosenbrand, van Rhijn, Rodenhuis and van Dam2002; Vardy et al., Reference Vardy, Wong, Yi, Park, Maruff, Wagner and Tannock2006).

4. 4. The timing of assessments. Cancer without chemotherapy can affect cognition and brain function (e.g., Ahles et al., Reference Ahles, Saykin, McDonald, Furstenberg, Cole, Hanscom and Kaufman2008; Cimprich et al., Reference Cimprich, Reuter-Lorenz, Nelson, Clark, Therrien, Normolle and Welsh2010; Hermelink et al., Reference Hermelink, Untch, Lux, Kreienberg, Beck, Bauerfeind and Münzel2007; Wefel et al., Reference Wefel, Saleeba, Buzdar and Meyers2010). Therefore, pre-post treatment comparisons may differentiate chemotherapy effects from those of the disease. Also, variable intervals between diagnosis, treatment, and testing make it difficult to compare results across studies (Rugo & Ahles, Reference Rugo and Ahles2003), because of the time course of cognitive changes after treatment (e.g., see Collins, this issue).

In addition to differences in methodology, participant characteristics within and across studies should be taken into account because of potential effects on cognition. These characteristics include age, education, previous head injury, genetics, medical comorbidities, anxiety, depression, and fatigue (Wefel, Vardy, Ahles, & Schagen, Reference Wefel, Vardy, Ahles and Schagen2011). Hormonal status can also be affected by chemotherapy or non-chemotherapy treatment (e.g., cytotoxic chemotherapy agents, estrogen receptor antagonists) and may independently affect cognition (Bender et al., Reference Bender, Sereika, Brufsky, Ryan, Vogel, Rastogi and Berga2007; Paganini-Hill & Clark, Reference Paganini-Hill and Clark2000; Schilder et al., Reference Schilder, Seynaeve, Beex, Boogerd, Linn, Gundy and Schagen2010). Cancer-specific variables that contribute to the variance within and across studies include disease staging, surgery, anesthesia, chemotherapy regimens (i.e., drug type, dose, and number of cycles), and newer targeted therapies such as human epidermal growth factor receptor 2 antagonist, trastuzumab (Herceptin), or vascular endothelial growth factor receptor inhibitor, bevacizumab (Avastin). These factors increase the between-subject variance, making it hard to detect differences between people who do and do not receive chemotherapy, or to interpret such differences when they emerge.

Notwithstanding these differences, longitudinal studies demonstrate lower cognitive performance following chemotherapy compared with healthy or with cancer no-chemotherapy cohorts (e.g., Collins et al., Reference Collins, Mackenzie and Kyeremanteng2013; Deprez et al., Reference Deprez, Amant, Smeets, Peeters, Leemans, Van Hecke and Sunaert2012). Although the time course and domains impacted are being actively studied, there is evidence that immediate and delayed recall, attention, working memory, executive function, and processing speed can be affected (e.g., Ahles et al., Reference Ahles, Saykin, McDonald, Li, Furstenberg, Hanscom and Kaufman2010; Collins et al., Reference Collins, Mackenzie, Stewart, Bielajew and Verma2009; Deprez et al., Reference Deprez, Amant, Smeets, Peeters, Leemans, Van Hecke and Sunaert2012; Wefel et al., Reference Wefel, Saleeba, Buzdar and Meyers2010), with small to moderate effect sizes (Jansen et al., Reference Jansen, Miaskowski, Dodd and Dowling2007; Jim et al., Reference Jim, Phillips, Chait, Faul, Popa, Lee and Small2012; Stewart et al., Reference Stewart, Bielajew, Collins, Parkinson and Tomiak2006). These effects are dose dependent: more chemotherapy is associated with worse cognitive performance (Collins et al., Reference Collins, Mackenzie, Tasca, Scherling and Smith2012; Collins et al., this issue; van Dam, Schagen, & Muller, Reference van Dam, Schagen, Muller, Boogerd, v.d. Wall, Droogleever Fortuyn and Rodenhuis1998). Cognitive symptoms tend to improve following chemotherapy, but the speed of improvement is variable across individuals, and objective dysfunction persists in a subset of survivors for months or even many years after treatment has ended (Ahles & Saykin, Reference Ahles and Saykin2001; Ahles et al., Reference Ahles, Saykin, McDonald, Li, Furstenberg, Hanscom and Kaufman2010; Coates et al., Reference Coates, Abraham, Kaye, Sowerbutts, Frewin, Fox and Tattersall1983; de Ruiter et al., Reference de Ruiter, Reneman, Boogerd, Veltman, van Dam, Nederveen and Schagen2011; Deprez et al., Reference Deprez, Amant, Smeets, Peeters, Leemans, Van Hecke and Sunaert2012; Jenkins et al., Reference Jenkins, Shilling, Deutsch, Bloomfield, Morris, Allan and Winstanley2006; Koppelmans et al., Reference Koppelmans, Breteler, Boogerd, Seynaeve, Gundy and Schagen2012; Schagen et al., Reference Schagen, Muller, Boogerd, Mellenbergh and van Dam2006; Silverman et al., Reference Silverman, Dy, Castellon, Lai, Pio, Abraham and Ganz2007; Tannock et al., Reference Tannock, Ahles, Ganz and Van Dam2004; Wefel et al., Reference Wefel, Saleeba, Buzdar and Meyers2010).

Why do some patients decline while others remain stable or improve? Perhaps some patients learn to cope with cognitive changes and develop compensation strategies. In others, the problem may be variability in performance, which may not be evident in one standard neuropsychological assessment or when using traditional neuropsychological measures (Bernstein, Catton, & Tannock, this issue).

Possible physiological underpinnings of cognitive decline after chemotherapy have been outlined in several reviews, and the causes are certainly multi-factorial. Symptoms appear to be associated with damage in white matter microstructure and cerebral vasculature (Ahles & Saykin, Reference Ahles and Saykin2007). Proposed mechanisms include altered integrity of the blood brain barrier, increased oxidative stress, and release of inflammatory cytokines (Ahles & Saykin, Reference Ahles and Saykin2007; Seruga, Zhang, Bernstein, & Tannock, Reference Seruga, Zhang, Bernstein and Tannock2008). Reductions in structural integrity of cortical white matter tracts, smaller gray matter volumes, axonal injury, and changes in neural activation in response to cognitive demands are consistent with this idea (de Ruiter et al., Reference de Ruiter, Reneman, Boogerd, Veltman, van Dam, Nederveen and Schagen2011; Deprez et al., Reference Deprez, Amant, Smeets, Peeters, Leemans, Van Hecke and Sunaert2012; Ferguson, McDonald, Saykin, & Ahles, Reference Ferguson, McDonald, Saykin and Ahles2007; Inagaki et al., Reference Inagaki, Yoshikawa, Matsuoka, Sugawara, Nakano, Akechi and Uchitomi2007; Kesler et al., Reference Kesler, Wefel, Hosseini, Cheung, Watson and Hoeft2013; Koppelmans et al., Reference Koppelmans, Breteler, Boogerd, Seynaeve, Gundy and Schagen2012; McDonald, Conroy, Ahles, West, & Saykin, Reference McDonald, Conroy, Ahles, West and Saykin2012). This body of work suggests that cancer and/or its treatment affect the brain, even when treatments are not targeting the brain.

New approaches lay the groundwork for better understanding this issue. The International Cancer and Cognition Task Force, a multidisciplinary group of experts in neuropsychology, clinical health psychology, and medical oncology issued recommendations to address study design discrepancies that should facilitate comparisons across clinical trials (Wefel, Vardy, Ahles, & Schagen, Reference Wefel, Vardy, Ahles and Schagen2011). Recommendations included: longitudinal studies that incorporate pre-chemotherapy assessments; appropriate comparison groups; a brief standardized test battery shown to be sensitive to cognitive changes after chemotherapy (i.e., Controlled Oral Word Association Test, Trailmaking Test, Hopkins Verbal Learning Test-Revised); control variables that account for physical and psychosocial factors; and analyses that address reliable change. Collins et al.'s elegant study design (this issue) incorporates these recommendations along with state of the art statistical techniques. They show that the chemotherapy dose-dependent cognitive decline largely remits 1 year after treatment due to improved working memory, although deficits persist in a subset of survivors.

New statistical approaches to test development may provide stronger links between subjective complaints and cognitive performance following cancer treatment. For example, revision of the Neurocognitive Questionnaire using item response theory, a technique that provides discrimination and difficulty parameters for test items, resulted in a self-report measure that is significantly correlated with memory and executive function performance in childhood cancer survivors with moderate effect sizes (Kenzik et al., Reference Kenzik, Huang, Brinkman, Shenkman, Robison, Hudson and Krull2012). In the future, application of these approaches to questionnaires examining cognitive dysfunction in adult cancer patients may result in screening tools that can identify those at risk for cognitive impairment.

Neuroimaging techniques reveal relationships between chemotherapy and cognition that are undetected by behavioral measures. In one of the most frequently cited functional magnetic resonance imaging (fMRI) studies, Ferguson et al. (Reference Ferguson, McDonald, Saykin and Ahles2007) evaluated performance on an n-back task in identical twins, one of whom had previously received chemotherapy for breast cancer. The twins performed equally well and showed comparable effects of working memory load on task performance. However, the twin who had cancer showed greater bilateral prefrontal and posterior parietal activation during the task, and had substantially more white matter hyperintensities bilaterally on structural MRI, than her unaffected sister. Those results suggest that chemotherapy for breast cancer can affect brain structure and function.

More recent longitudinal imaging studies using fMRI or DTI add knowledge about brain changes during and after chemotherapy (for reviews, see Saykin, de Ruiter, McDonald, Deprez, & Silverman, Reference Saykin, de Ruiter, McDonald, Deprez and Silverman2013; Simó, Rifà-Ros, Rodriguez-Fornells, & Bruna, Reference Simó, Rifà-Ros, Rodriguez-Fornells and Bruna2013). These approaches lead toward developing diagnostic and predictive biomarkers of cognitive decline after chemotherapy. For example, Kesler et al. (Reference Kesler, Wefel, Hosseini, Cheung, Watson and Hoeft2013) applied multivariate pattern analysis (MVPA) to fMRI, and accurately differentiated default-mode network connectivity in breast cancer survivors who received chemotherapy from those who did not receive chemotherapy and healthy controls. Hosseini & Kesler (this issue) applied this technique to examine prefrontal cortex connectivity during an executive function task (go-nogo) and found a similar pattern of results with discrimination between women who were treated with chemotherapy and those who were not, despite equivalent task performance. This approach provides evidence of changes in neural circuitry “at rest” and when executive functions are challenged. The authors note that MVPA is concerned with reliability of a difference between groups rather than the existence of a difference between groups. This is consistent with the suggestion that variability underlies cognitive difficulties experienced after chemotherapy (Bernstein et al., this issue).

Finally, in addition to implementation of rigorous approaches to study design and novel methodologies, theoretically driven studies are critical for advances in this field. Recent suggestions that cancer treatments place survivors at risk for premature aging (Ahles, Root, & Ryan, Reference Ahles, Root and Ryan2012; MacCormick Reference MacCormick2006), provides another framework for future studies.