Patients
For this study, baseline data were used from an RCT on the efficacy of MCGP for cancer survivors [26]. Ethical approval for this study was provided by the Medical Ethical Committee of Leiden University Medical Center (NL34814.058.10). Information about the study protocol, participants, and primary outcomes has been published previously [26, 27].
Participants were recruited between August 2012 and September 2014. Inclusion criteria were as follows: cancer diagnosis in the last 5 years, treated with curative intent, main treatment completed (i.e., surgery, radiotherapy, chemotherapy), presence of an expressed need for psychological support, and at least one psychosocial complaint. Exclusion criteria were as follows: severe cognitive impairment, current psychological or psychiatric treatment elsewhere, and an insufficient mastery of the Dutch language. All criteria were ascertained during a telephonic screening interview.
Informed consent was obtained from all individual participants included in the study. Demographic characteristics were obtained by self-report: age, gender, marital status, education level, employment, religious background, other negative life events, and past psychological treatment. Illness-related characteristics included type of cancer, tumor stage, type of treatment, and time since treatment and were retrieved from medical records or by self-report, if medical records were unavailable.
Outcome measures
Psychological well-being was measured using the Dutch version of the SPWB [28]. This is a 39-item measure consisting of six subscales: self-acceptance (α = 0.81), positive relations with others (α = 0.83), autonomy (α = 0.84), environmental mastery (α = 0.76), purpose in life (α = 0.79), and personal growth (α = 0.071). Items were answered on a 6-point Likert scale, ranging from 1 (strongly disagree) to 6 (strongly agree). Subscale scores were calculated as the mean item score. Higher scores indicated greater well-being. The Dutch version has the same six subscales as the original version, although several items had to be removed to reach adequate fit. The Dutch version showed sufficient internal consistency and good construct validity [28].
The Dutch version of the PMP was used to measure personal meaning [17, 29]. This 39-item measure has five subscales: dedication to life (α = 0.89), fairness of life (α = 0.77), goal-orientedness (α = 0.89), relations with other people (α = 0.85), and relation with God (α = 0.86). Items were scored on a 7-point Likert scale from 1 (not at all) to 7 (a great deal). A higher score reflects a more important source of meaning. This measure was validated in Dutch cancer patients and showed good internal consistency and construct validity. Its number of items and factor structure differed from the original Canadian version. Of the originally 57 items, 18 had to be removed in the Dutch version, because of low or double loadings and the original factors “relations” and “intimacy” formed one factor in the Dutch version, as well as “fair treatment” and “self-acceptance” [29].
Posttraumatic growth was measured using the Dutch translation of the PTGI [25, 30]. This 21-item measure has five subscales: relating to others (α = 0.85), new possibilities (α = 0.80), personal strength (α = 0.79), spiritual change (α = 0.70), and appreciation of life (α = 0.75). Items were rated from 0 (not at all) to 5 (very great degree). Subscale scores were calculated as mean item scores and a higher score suggests stronger growth. A psychometric study of the PTGI in Dutch cancer patients showed good internal consistency, construct validity, and factorial validity. The Dutch version contains the same factors as the original version [30].
Statistical methods
Exploratory maximum likelihood factor analysis (EFA) with varimax rotation on all subscales of the SPWB, PMP, and PTGI was conducted to explore possible areas of overlap between psychological well-being, personal meaning, and posttraumatic growth. The number of factors to retain was based on the eigenvalues (> 1.0), the slope of the scree plot and parallel analysis. To assess the goodness-of-fit of the resulting model, this model was entered into a confirmatory maximum likelihood factor analysis (CFA) using the same sample. The following goodness-of-fit indices and thresholds were used: the χ2-test (p < 0.05), the root mean square error of approximation (RMSEA, < 0.06), the comparative fit index (CFI, ≥ 0.90), the Tucker-Lewis index (TLI, ≥ 0.90), and the standardized root mean square (SRMS, < 0.08). Missing data were presumed to be missing completely at random (MCAR).
When the model resulting from the EFA would not show adequate fit, two additional models would be considered. (1) In order to compare the result of the EFA with the null model (i.e., a model in which the subscales load on a factor that represents their own measurement instrument, revealing no areas of overlap), the goodness-of-fit indices would be calculated for this null model, as well, using CFA. (2) In order to explore the overlap between the SPWB, PMP, and PTGI further, structural equation modeling (SEM) would be used. Beginning with the null model, in which each measurement instrument formed a latent variable, represented by its subscales as manifest variables, the path with the highest modification index would be added to the model and the goodness-of-fit indices would be re-calculated. This procedure would be repeated until the model had an adequate fit. Correlations in the models were considered as low (< 0.5), moderate (≥ 0.5 and < 0.7), or high (≥ 0.7). All analyses were performed in IBM SPSS Statistics 24 or R 3.4.0, package Lavaan.