Prostate cancer patients’ quality of life assessments across the primary treatment trajectory: ‘True’ change or response shift?

Figures & data

Figure 1. Patient flow. Note. All percentages refer to n = 522 eligible patients.

Table 1. Descriptive statistics of the sample.

	M	SD
Age (years)	64.7	7.1
	n	%
Marital status
Married	333	82.8
Single/divorced	51	12.7
Widowed	18	4.5
Education
Less than junior (<10 years; basic secondary school)	133	33.0
Junior (10 years; middle-level secondary school)	147	36.6
Senior (high school graduate)	121	30.1
Other	1	0.3
Occupational status
Retired	247	61.4
Employed	144	35.8
Unemployed	9	2.2
Missing	2	0.5
Tumor stage
T1	15	3.7
T2	213	53.0
T3	135	33.6
T4	1	0.2
TX, missing	38	9.5
Gleason score
Below 6	2	0.5
6	61	15.2
7	277	68.9
Above 7	32	8.0
Missing	30	7.5
Treatment
Surgery	328	81.6
Radiotherapy	53	13.2
Multiple	20	5.0
Missing	1	0.2

Table 2. Change of observed EORTC QLQ-C30 functioning subscale scores from baseline to 3-months follow-up.

		Baseline	At 3 months	p-Value	Mean difference	Observed change (d)	‘True’ change (d)	Response shift (d)
PF	M	95.0	89.7	<0.001	−5.3	−0.44	−0.24	−0.21
	SD	11.2	14.9
RF	M	91.5	79.9	<0.001	−11.5	−0.47	−0.26	−0.14
	SD	18.1	25.2
EF	M	73.7	77.2	<0.001	3.5	0.17	−0.53	0.69
	SD	21.6	23.3
CF	M	89.5	87.4	0.014	−2.1	−0.13	−0.37	0.25
	SD	17.1	17.5
SF	M	85.9	77.0	<0.001	−8.9	−0.35	n.a.	n.a.
	SD	20.6	24.2

CF, Cognitive Functioning; EF, Emotional Functioning; PF, Physical Functioning; RF, Role Functioning; SF, Social Functioning. d, Cohen’s d effect size; n.a., not applicable. Please note that rounding errors may have occurred in summing up effect sizes for ‘true’ change and response shift. For RF, the d-values of ‘true’ change and response shift do not sum up to the d-value of observed change, due to methodical reasons (see also Discussion).

Figure 2. Structural equation model, with the EORTC QLQ-C30 subscales indicating the latent variable quality of life. Note. Rectangles represent observed variables, circles latent variables, single-headed arrows the impact of one variable on another, and double-headed arrows the covariance between pairs of variables, respectively. The numbers are the parameter estimates of the resulting response shift model: (standardized) factor loadings (λ), (unstandardized) intercepts (τ), residual variances (ɛ) and residual correlations. Parameter estimates separated by a slash represent separate baseline and follow-up assessment estimates; otherwise, unstandardized parameters were constrained to be equal across assessments. Bold numbers indicate response shift. λ indicates factor loadings, τ intercepts, and ɛ residual variances, respectively. CF, Cognitive Functioning subscale; EF, Emotional Functioning subscale; EORTC, European Organization for Research and Treatment of Cancer; PF, Physical Functioning subscale; QoL, quality of life; RF, Role Functioning subscale; SF, Social Functioning subscale.

Table 3. Zero-order correlations of the model variables.

	Baseline				At 3 months
	PF	RF	SF	CF	PF	RF	SF	CF
PF
RF	0.57				0.66
SF	0.24	0.39			0.46	0.55
CF	0.21	0.36	0.39		0.36	0.34	0.38
EF	0.23	0.34	0.55	0.47	0.51	0.51	0.58	0.52

CF, Cognitive Functioning; EF, Emotional Functioning; PF, Physical Functioning; RF, Role Functioning; SF, Social Functioning. All correlations p < 0.05.