Prostate cancer-specific anxiety in Dutch patients on active surveillance: validation of the memorial anxiety scale for prostate cancer

The original US MAX-PC was developed for identifying and quantifying anxiety in men with prostate cancer and was designed for self-administration [3]. It consists of 18 items, divided into 3 subscales: ‘prostate cancer anxiety’ (11 items; example: ‘Any reference to prostate cancer brought up strong feelings in me.’—Not at all, rarely, sometimes, often), ‘PSA anxiety’ (3 items; example: ‘I have been so anxious about my PSA test that I have thought about delaying it.’—Not at all, rarely, sometimes, often), and ‘fear of recurrence’ (4 items; example: ‘Because cancer is unpredictable, I feel I cannot plan for the future.’—Strongly agree, agree, disagree, strongly disagree). Total score ranges from 0 to 54, with 54 indicating maximum anxiety. Scores on the 3 scales range from 0 to 33, 0 to 9, and 0 to 12, respectively. The total MAX-PC was subdivided into the three subscales as it was designed to tap three specific aspects of prostate cancer-related anxiety, although the reliability of the ‘PSA anxiety’ subscale was found to be weak in the original publication [3]. To our knowledge, the MAX-PC has never been translated or used in an active surveillance setting [3‐6].

The translation to Dutch followed standardized forward–backward procedures [7]. First, three forward translations by native Dutch with a medical background with fluency in English were performed and pooled into a common version after a consensus meeting. Second, a native English speaker fluent in Dutch and also with a medical background translated this provisional Dutch version back into English, while being blinded to the original version. This back translation showed some discrepancies with the source document, but these were mainly related to the wording and not to the specific meaning of items. Consensus was reached by discussion. The Dutch version of the MAX-PC was first tested face-to-face at the respondent’s home in 5 participants, who completed the questionnaire while thinking aloud in the presence of a researcher. Afterward, guided by a checklist, potential problems in acceptance or comprehension and time necessary for questionnaire completion were explored and discussed with the participant. Only minor problems, different for all 5 participants, were found during this process. These did not indicate a need to adapt the Dutch version of the MAX-PC. The Dutch version of the MAX-PC can be found at https://​www.​prias-project.​org/​modules/​news/​article.​php?​storyid=​12.

Our study group consisted of Dutch patients who had been recently diagnosed with prostate cancer, elected active surveillance as the initial treatment option, and who consented to participate in the prospective protocol-based PRIAS (Prostate Cancer Research International: Active Surveillance) study on active surveillance [8]. If diagnosed between May 2007 and May 2008, these men received a questionnaire within 6 months after diagnosis. Men are medically eligible for the PRIAS study if they have small, localized, and well-differentiated prostate cancer. The medical ethics committee at the Erasmus University Medical Centre in Rotterdam in The Netherlands (coordinating centre) approved this study (number 2004-339).

Besides the MAX-PC, the questionnaire included the Decisional Conflict Scale (DCS, 16 items with 5 response options: total score 0–100, with 100 indicating maximum decisional conflict) to assess decisional conflict on the choice for active surveillance [9]; the Centre for Epidemiologic Studies Depression scale (CES-D, 20 items with 4 response options, total score 0–60, with 60 indicating maximum depression) [10]; the abridged State Trait Anxiety Inventory (STAI-6, 6 items with 4 response options, total score 20-80, with 80 indicating maximum generic anxiety) [11]; and the Short Form health survey 12 (SF-12, 12 items with 2–6 response options, mean score 50, with a standard deviation of 10 in the general US population) of which we used the Mental Component Summary (MCS) score [12].

Floor or ceiling effects causing an abnormal score distribution were considered to be present when >15% of respondents had the lowest or the highest possible score [13].

We estimated the internal consistency by calculating the Cronbach’s alpha coefficients for the total MAX-PC and the three subscales. An alpha of 0.7 is generally regarded as sufficient for group comparisons [14].

An initial confirmatory factor analysis (CFA) model was fitted in which each item was assigned to one of the three underlying factors, similar to the original publication, to verify that the original factor structure was present in our data [3]. Correlated factors were allowed in this model. The fit of the model was improved by freeing fixed parameters according to the sequence implied by the modification indices. When we found equal or nearly equal modification indices, priority was given to freeing fixed parameters in the covariance matrix of the errors over for instance fixed factor loadings [15]. Model fit was assessed with the chi-square (χ²) test, the root mean square error of approximation (RMSEA), and comparative fit index (CFI).

Construct validity was assessed by comparing MAX-PC total score and the three subscale scores with DCS, CES-D, STAI-6, and SF-12 scores, respectively, using Pearson correlation coefficients. Correlations with r > 0.3 were considered relevant [16, 17]. We hypothesized that higher scores on the total MAX-PC and subscales have to be related to higher scores on DCS, CES-D, and STAI-6, and to lower SF-12 MCS scores; that correlations were highest with STAI-6, as this measure also is anxiety specific, and lower with DCS, CES-D, and SF-12 MCS; and that correlations with the ‘PSA anxiety’ subscale were lower, because this is a very specific subscale, previously found to show lower construct validity. We tested differences between correlations with MAX-PC total for significance with a bootstrap procedure to obtain standard errors [18]. At least 75% of the results should be in accordance with a priori hypotheses [13].

For statistical analysis, the commercially available software Statistical Package for the Social Sciences, (version 15.0; SPSS, Inc, Chicago, IL, USA), S-Plus (version 8.0; TIBCO software, Palo Alto, CA, USA), and LISREL (version 8.72; Scientific Software International, Lincolnwood, IL, USA) [19] were used. A P-value of ≤0.05 was considered statistically significant.