Participants
A convenience sample of 281 patients with non-CNS cancers completed the FACT-Cog-v3 online. The inclusion criteria were: (1) age between 18 and 65 years old; (2) diagnosis of non-CNS cancer; (3) undergoing or having received treatments for cancer; (4) ability to read and write Portuguese; and (5) Portuguese nationality. Patients with (1) psychiatric or communication disorders, and/or other serious medical condition; (2) CNS metastasis; and (3) diagnosis of dementia, epilepsy, brain injury (stroke, head injury), and drug or alcohol abuse, were excluded since these conditions might impact on cognitive functioning. Of the total sample, 266 participants additionally filled out the QLQ-C30 and, of those, 258 participants also filled out the Hospital Anxiety and Depression Scale (HADS) (see Measures section for a description of these instruments).
Procedure
Volunteer cancer patients were recruited through online advertisement disseminated across Portugal. An online survey (LimeSurvey®) located on a server from the University of Aveiro was used to collect data from participants. Participants were a self-selected sample who replied to advertisements posted on social media (Facebook), specifically in support groups, blogs/forums, cancer-related information groups, and pages of national cancer associations that accepted to collaborate in the dissemination of the study, targeting Portuguese adult cancer patients; national cancer associations were also invited to collaborate in disseminating information about the study by e-mail to their associates. Advertisements invited potential participants to access a link to the survey. Those who clicked on the link were then given detailed information about the study’s goals, inclusion criteria, and ethical statements. Participants were informed that their participation was voluntary and confidentiality of the data was ensured. Cancer patients who agreed to the study conditions provided their informed consent by clicking on the “Yes” option to the question “Do you accept to participate in this study?”. The survey was open for four months, between January and April 2021. The protocol took approximately 30 min to complete. Participants’ ethical treatment was safeguarded, in accordance with the Declaration of Helsinki [21] and the guidelines of the American Psychological Association [22]. The Ethics and Deontology Committee of the University of Aveiro (22 January 2020/ No. 30/2019) approved all the procedures of this study.
Measures
Participants completed a global self-report questionnaire assessing sociodemographic (e.g., age, education, occupation) and clinical variables (e.g., cancer diagnosis, treatments, brain injuries).
The version 3 of the FACT-Cog [7] used in this study was translated into universal Portuguese by the FACIT team, using an iterative methodology [23, 24]. For the present study, authorization was requested from FACIT to test its psychometric properties. Figure 1 presents an overview of the translation process performed by FACIT as well as a schematic representation of the validation process described in the present article.
The FACT-Cog-v3 is a 37-item self-response measure to assess cognitive concerns of cancer patients, consisting of four subscales. For CogPCI (20 items; 0–80) and CogOth (4 items; 0–16) items, the patient indicates how often the situation occurred during the last 7 days, on a 5-point Likert scale (“0 = Never” to “4 = Several times a day”); and for CogPCA (9 items; 0–36) and CogQoL (4 items; 0–16), a 5-point Likert scale (“0 = Not at all” to “4 = Very much”) is used to indicate the severity of each situation taking into account the last week. Although two items of CogPCI and two items of CogPCA are not currently scored under the FACT-Cog-v3 scoring algorithm, according to FACIT, they may be included if some additional analyses (i.e., internal consistency and individual item-total correlation coefficients) are conducted to confirm that the items have a good fit with the scale. Therefore, the 37 items were used in this study to test its psychometric properties [19]. Except for the CogPCA subscale, negatively worded items are reverse scored prior to summing all the items. Higher scores indicate better PCF and better QoL. The reliability and validity of these scores have been established [14, 16], including the preliminary evaluation of the Portuguese version that revealed good psychometric properties regarding reliability and concurrent and convergent validity [25].
The QLQ-C30 [12, 26] is a 30-item self-response questionnaire that was used to assess health-related QoL. This scale includes a global health status/QoL subscale, functional and symptom subscales, and single items. Each of the items is scored on a 4-point Likert scale (“1 = Not at all” to “4 = Very much”), except the items of the global health/QoL subscale (modified 7-point linear analogue scale). The scores for each subscale range from 0 to 100, with higher scores for functional scales and global health/QoL representing better functioning and QoL, while higher scores in the symptom subscales and single items are indicative of worse symptoms. Of interest in this study was the cognitive functioning, global health/QoL, fatigue, and sleep disturbance subscales. Good psychometric properties were found on the Portuguese validation study [26]. In this study, the subscales used have shown acceptable Cronbach’s alpha: Cognitive Functioning = 0.79, Fatigue = 0.88, Pain = 0.88, Nausea/Vomiting = 0.70, and Global Health Status/QoL = 0.91.
This study included use of the HADS [27, 28], a 14-item self-response questionnaire, useful in recognizing emotional components of physical illness. The HADS consists of two subscales, each with seven items, one measuring anxiety (HADS-A) and one measuring depression (HADS-D); these items are answered on a 4-point Likert scale. Each subscale has a score ranging 0–21 points; higher scores indicate a higher level of anxious and depressive symptoms. Good psychometric properties were found on the Portuguese validation study [28]. In this study, Cronbach’s alpha was acceptable (0.86) for both subscales.
Statistical analysis
Statistical analyses were performed with the Statistical Package for the Social Sciences (IBM SPSS, version 28.0; IBM SPSS, Inc., Chicago, IL) and with the lavaan package for R [29, 30].
Descriptive statistics were first calculated for sample’s demographic and clinical characteristics. Measurement characteristics, i.e., mean scores, standard deviations (SD), and range, are presented for each subscale.
Reliability, through internal consistency, was measured using the following techniques and cut-off recommendations: mean of the inter-item correlation (adequate if > 0.30), corrected item-total correlation (adequate if > 0.50) [31], and Cronbach’s alpha (acceptable if > 0.70 and high if > 0.90) [32,33,34].
To test criterion validity of the scale, concurrent validity was established via correlation coefficients between the scores of the FACT-Cog-v3 and the QLQ-C30 cognitive functioning subscale.
Construct validity was determined by factorial, convergent, and discriminant validity. Confirmatory factor analysis (CFA) was used to test the hypothesis that the construct of PCF, as assessed by the FACT-Cog-v3, is composed of four separate factors of CogPCI, CogOth, CogPCA, and CogQoL [7]. Mardia’s Test was performed to assess the multivariate normality of the sample. Regarding sample size requirements for CFA, rules-of-thumb vary from five to 10 subjects per variable, including a minimum of 100 subjects [34] or a range of 200–300 individuals [35, 36]. A CFA using weighted least squares with mean and variance adjustment (WLSMV) estimator was conducted. We considered the following goodness-of-fit indices and respective cut-off recommendations for good adjustment [31, 37,38,39,40]: Chi-Square (χ2); Comparative Fit Index (CFI; 0.90 ≤ CFI ≤ 0.95); Tucker-Lewis Index (TLI; 0.90 ≤ TLI ≤ 0.95); Root Mean Square Error of Approximation (RMSEA; 0.05 ≤ RMSEA ≤ 0.08); and Standardized Root Mean Square Residual (SRMR ≤ 0.08). Local model fit was assessed through the items’ standardized factor loadings (λ ≥ 0.50) and individual reliability (R2 ≥ 0.25) [31, 40].
Convergent and discriminant validity were assessed using the Fornell and Larcker [41] criterion and by correlations with external criteria. Convergent validity of the measurement model can be assessed by the average variance extracted (AVE; AVE ≥ 0.50) and construct reliability (CR) for each factor (CR ≥ 0.70) [41], and discriminant validity is supported when the AVE for a construct is greater than the squared interconstruct correlations [31]. Convergent validity was also assessed by examining the correlations between FACT-Cog-v3 subscales and HADS and QLQ-C30 fatigue, sleep disturbance, and global health status subscales. Discriminant validity was further examined through the correlation between FACT-Cog-v3 subscales and QLQ-C30 pain and nausea/vomiting subscales.
Following the guidelines presented by Ratner [42], the correlations were classified as weak (0–0.3), moderate (0.3–0.7), and strong (> 0.7–1.0).
All significance tests were conducted using a significance level of p < 0.05.