Pilot Study of the Prospective Identification of Changes in Cognitive Function During Chemotherapy Treatment for Advanced Ovarian Cancer

Many cancer patients receiving chemotherapy develop persistent changes in cognitive function, characterized by declines in memory, attention, and executive function.^{[1] and [2]} Neurocognitive symptoms may be associated with discrete or multiple etiologies, including direct or indirect effects of cancer on the central nervous system, comorbid neurologic or psychiatric diagnoses, and diffuse and specific effects of cancer treatment, including radiation therapy, immunotherapy, and chemotherapy,^{[2] and [3]} as well as fatigue, pain, and anemia, all of which have been associated with poorer performance on neurocognitive testing among cancer patients.⁴ Even mild cognitive changes can be expected to influence quality of life and the ability to function effectively in performing day-to-day activities.

Although the perception of cognitive decline is a common complaint among individuals treated with chemotherapy, it is poorly understood and limited efforts have been made to identify the extent of this problem among women with ovarian cancer. To date, the few studies documenting the neuropsychological consequences of ovarian cancer and its treatment have shown that patients report cognitive problems but that these problems were not quantifiable using objective measures due to the lack of sensitivity of standard instruments to the subtle changes that occur during cancer treatment.^{[5], [6] and [7]}

Although studies of cognitive function among oncology patients have used instruments that have been validated in their own disciplines and with a variety of diseases, the evidence is emerging that they are not comprehensive or appropriate tools for the detection and evaluation of chemotherapy-related change in cognitive function.⁸ Furthermore, the likelihood of having these tests conducted in a similar manner across multiple institutions, sites, and interviewers with any degree of consistency is very low. This study was designed as a pilot study of the identification of chemotherapy-related changes in cognitive function among women with advanced ovarian cancer using a Web-based assessment tool (Headminder, Inc., New York, NY).⁷ The primary goal of the current study was to determine if there is evidence of changes in the cognitive function domains of attention, processing speed, and reaction time as well as self-reported changes in the memory, sensory-perception, and cognitive-intellectual domains of cognitive function during chemotherapy among women with newly diagnosed advanced ovarian cancer.

Materials and Methods

All study methods and procedures were reviewed and approved by the University of Arizona Institutional Review Board. Eligible patients included women with a histologically or pathologically confirmed diagnosis of stage III–IV epithelial ovarian or primary peritoneal cancer who were prescribed at least six courses of platinum-based therapy. Patients were excluded if they had a prior history of any cancer (other than nonmelanoma skin cancer), chemotherapy, radiation therapy, erythropoietin treatment (within the last 6 months), or severe head injury. Initially, patients were excluded if they received intraperitoneal therapy, but the protocol was later amended to permit the use of any platinum-based therapy, regardless of route of administration.

Assessment Tools

After providing informed consent, patients completed a neurocognitive battery of tests and the Functional Assessment of Cancer Therapy—Neurotoxicity (FACT-Ntx, to assess patient-reported neuropathy).^{[9] and [10]} The neurocognitive evaluation included both a computerized, Web-based and a patient-reported assessment. The Web-based assessment was provided by HeadMinders, Inc.^{[7] and [11]} and was a modified version of the Cognitive Stability Index. The modified battery was comprised of two warm-up tasks and three empirically-derived cognitive factors: Processing Speed (Animal Decoding and Symbol Scanning subtests), Attention (Number Recall and Number Sequencing subtests), and Reaction Time (Response Direction 1 and Response Direction 2 subtests). The subtests have been validated against traditional neuropsychological tests in healthy and clinical populations, including cancer patients.¹² Cognitive domain correlations in the battery's healthy normative sample range from 0.52 to 0.74, and correlations are similar or higher in clinical populations. Test–retest reliability of the factor scores between first and second administrations ranges from 0.74 to 0.82.¹² This Web-based neurocognitive assessment tool is 21 CFR Part 11– and Health on the Net (HON)–compliant to ensure patient confidentiality. Prior to undergoing the Web-based cognitive tests, all study participants completed a keyboard proficiency test as a “warm-up task” to the computerized assessment.

The patient-reported cognitive function tool used was the Patient Assessment of Own Functioning Scale (PAF).^{[13], [14] and [15]} The PAF includes eight scales that are grouped into the nature of the ability being considered. The Memory, Sensory-Perceptual, and Cognitive-Intellectual subscales of the PAF are included in this self-assessment questionnaire. Respondents are asked to rate on a six-point scale, from almost always to almost never, how often they experience a particular kind of difficulty in their everyday lives. For this study, the Memory and Cognitive-Intellectual subscales of the PAF were used, similar to other clinical research protocols investigating cognitive changes during chemotherapy treatment.¹⁵ The PAF has been shown to be directly related to the Minnesota Multiphasic Personality Inventory (MMPI)¹³ and to be highly correlated with other cognitive impairment indices, such as the American College of Rheumatology neuropsychology research battery of tests.¹⁶ Of note, self-reported cognitive change has not been shown to correlate formal assessments of cognitive function among individuals who have experienced cancer.^{[17], [18], [19], [20] and [21]}

The FACT-Ntx is a validated instrument^{[9] and [10]} that was used to evaluate neurotoxicity. This scale includes 11 items: nine to assess neurotoxicity, one to assess bodily weakness, and one to assess anemia. Neurotoxicity may affect a patient's ability to use the keyboard in the computerized neurocognitive evaluation. This complete assessment battery of tests was completed at baseline (within 5 days of initiation of chemotherapy) and again during follow-up assessments at cycle three and cycle six of chemotherapy. The medical record was reviewed and data were abstracted related to chemotherapy medications, all concomitant medications, and blood test results (eg, hemoglobin, CA-125).

Statistical Plan

This prospective study was exploratory in nature and designed to collect pilot data to determine if there is evidence of neurocognitive change in attention, processing speed, response time, or self-reported cognitive function during the course of chemotherapy among women being treated for advanced ovarian cancer. The purpose of this study was to obtain preliminary estimates of the incidence and degree of cognitive decline to aid in the planning of future studies. While prior estimates of cognitive function were not available for this population, power analyses demonstrate that with a target recruitment goal of 30 patients, a McNemar's test has 78% power at the 0.05 level of significance to detect a significant decline in impairment in a cognitive domain if 12 patients are found to have impairment prior to course six of treatment (but not at course three) and if as few as two patients demonstrate impairment prior to course three but not at course six. This study was therefore powered to detect declines in one or more of the domains that may have occurred at less than both of the study time points following the baseline assessment.

To be considered fully evaluable, patients had to have completed at least one follow-up neurocognitive evaluation and may not have received antipsychotic neuropsychological medications during the study (eg, chlorpromazine, haloperidol, clozapine). Antidepressants and antianxiety medications (eg, serotonin/norepinephrine reuptake inhibitors or benzodiazepines) were permitted and use was recorded throughout study participation. A summary score for each cognitive domain (processing speed, reaction time, and attention) was recorded at each assessment time point using the HeadMinder Web-based assessment. This summary score was assessed by time (processing speed and reaction time), measured to the hundredth of a second, and by number of errors (attention). If a cognitive domain summary score at a follow-up assessment time declined at least one standard error of measurement (SEM) from baseline, the patient was considered to have experienced a decline at that time point. For the purposes of this article, such declines are referred to as “impairments” within the cognitive domain under investigation. A cognitive index score (CIS) was calculated as the number of cognitive domains impaired for the time point. The range of a CIS is 0–3, with zero equal to no impairment on any cognitive domain and three equal to impairment on all cognitive domains. Patients with only one cognitive domain decline (CIS = 1) at any one of the follow-up assessment time points were considered as having possible cognitive function decline. Patients with more than one cognitive domain impairment (CIS >1) at any follow-up assessment time points were considered as having evidence of cognitive function decline. The incidence of cognitive function impairment was determined by the percentage of patients who experienced any cognitive domain impairment (including possible and evidence of decline) at any follow-up assessment.

A repeated-measures analyses of variance (ANOVA) was used to further explore the neurocognitive values at the various time points during the study. Many of the neurocognitive values were not normally distributed but skewed either positively or negatively, so the square roots of the values were used in the analyses. Since this is an exploratory analysis, no corrections for multiple comparisons were performed.

The patient-reported cognitive function instrument (PAF) contains items scored on a Likert-type scale from almost never to almost always (range 0–5). Patient-reported outcomes as measured with the PAF are measured as mean scale values, ranging from 0, indicating no impairment, to 5.0, indicating complete impairment. PAF score ranges indicate low (≤1.25), medium (1.26–1.92), and high (≥1.93) levels of cognitive impairment.¹³ A total FACT-Ntx score was obtained; lower scores represent greater neurotoxicity, ranging from 0 (extreme neurotoxicity) to 44 (no neurotoxicity). The total score was reported, with adjustments made for missing values as described elsewhere.²²

Results

Thirty patients were enrolled in this study; however, two were later deemed ineligible, and one was unable to complete the baseline neurocognitive assessment prior to chemotherapy and was withdrawn from the study, resulting in 27 patients available for assessment. Five of these patients did not complete all neurocognitive assessments. The primary reason for nonadherence to the study schedule was clinical scheduling (eg, chemotherapy was administered prior to the neurocognitive assessment). The characteristics of eligible patients are provided in Table 1. The majority of patients were receiving intravenous chemotherapy (intraperitoneal therapy was at first not permitted but later was allowable following an amendment to the protocol) and taking concomitant sleep, antianxiety, and/or antidepressant medications outside of every 3- to 4-week chemotherapy regimen (primarily zolpidem, lorazepam, sertraline, and/or trazodone).