Is EQ-5D-5L sensitive enough to detect treatment-related changes in health status of prostate cancer patients? A nationwide Norwegian longitudinal study from the prostate cancer registry

All incident cancer cases and pre-cancerous lesions are registered with the Cancer Registry of Norway (CRN) which includes information on basic diagnosis (location and morphology) and treatment-related data [14, 15]. Pathology reports are submitted directly from the pathology laboratories to the CRN, and the database is continuously updated and matched with information from The National Population Register on vital status and migration. The registration is compulsory, and it has been estimated that the CRN is more than 99% complete on a minimum of diagnostic information on solid tumors [16]. The Norwegian Prostate Cancer Registry (NoPCR) includes all incident PCa cases (ICD 10, localization C61) with individual information on pathological and clinical data sent by an electronic cancer report from the clinicians.

In this nationwide prospective study (2017–2019), CRN invited all PCa patients diagnosed with invasive PCa in Norway from January 2017 through December 2019 to participate in a three-year survey on Men’s Health. The survey collected patient-reported outcomes (PROs) on a large scale with the intention to test and justify whether PROMs should be implemented for all new incident PCa cases in the Norwegian Prostate Cancer Registry (NoPCR).

PCa patients were identified and invited electronically directly through CRN. Patients who used an official digital mailbox (Digipost/eBoks), or the official Norwegian health portal “Helsenorge.no” (since 2020), were invited digitally, while the remaining patients were invited by postal mail. The postal invitations included the questionnaire and a pre-paid return envelope. Patients included from the NoPCR were eligible to participate if they were aged 18 years or older and diagnosed with invasive PCa (n = 8002), regardless of the extent of the disease and treatment choice. PCa patients were excluded if they had other prior cancer(s) or PROM were not assessed within 2 months of diagnosis prior to start of treatment (baseline(T1)) and/or approximately 6 months after treatment (T2) (> 277 days and < 427 from T1 to T2). Simultaneously a reference group from the general population, matched 8:10 by ten-year age groups and geographic region of residence with no previous prostate cancer history, was also invited to participate. The reference group was included to evaluate the longitudinal stability of the EQ-5D-5L in individuals without PCa and to provide a contextual comparator for interpreting changes in the PCa cohort. Although no formal power calculation was performed due to the population-based design, the resulting sample sizes are assumed to provide adequate precision for the planned analyses.

Socio-demographic characteristics included age (years), bodyweight (kg), completed education (≤ 12 years) and being married/cohabitant (yes/no) at baseline. For the PCa patients, basic diagnosis- and treatment-related data included time of diagnosis, SEER grade of PCa stages (localized, regional, distant metastases, unknown), and treatment (radical prostatectomy, radiotherapy).

The PROMs included the disease-specific instruments EORTC-QLQ-C30 and EPIC-26 and the generic instrument EQ-5D-5L, providing a unique opportunity in comparing the generic EQ-5D-5L and disease-specific instruments with respect to validity and responsiveness to change. PROs were assessed at baseline preferably within 2 months of diagnosis (T1) prior to start of treatment, and approximately 6 months after treatment (T2). The PROMs were scored according to published algorithms [17‐21].

The European Quality of Life 5-Dimensions questionnaire (EQ-5D-5L) is a generic health utility questionnaire consisting of five single items (domains): mobility, self-care, usual activities, pain/discomfort and anxiety/depression (dimensions of health) with ratings on five levels of perceived problems from no problems (1) to being unable to do/having extreme problems [22]. A higher score on EQ-5D-5L reflect more problems. The level scores are presented as global health indices with a weighted total value for health status [21, 23, 24]. The second part is a Visual Analogue Scale (VAS) ranging from 0 (worst health state) to 100 (best health state) and used as an overall measure of perceived health status, Higher scores on EQ-VAS reflect better perceived health. Evidence supports the validity of the EQ-5D-5L for use in the Norwegian general population [18, 19, 21]. In the present study, we used the population norms for UK since the statistical package R did not yet include the updated Norwegian population norms [19].

The European Organization for Research and Treatment of Cancer Quality of Life Questionnaire (EORTC-QLQ-C30) focuses on several HRQoL aspects of cancer patients and is not PCa specific [25]. The instrument contains five functional subscales including physical functioning (five items), role functioning (two items), emotional functioning (four items), cognitive functioning (two items) and social functioning (two items), a global health status question, and nine symptom subscales (fatigue, pain, nausea and vomiting, dyspnea, insomnia, appetite loss, constipation, diarrhea and financial impact). A higher score indicates better quality of life for functional scales and global health status, and lower quality of life for symptom scales [25]. Evidence supports the validity of the EORTC QLQ-C30 for use in Norwegian patients with head and neck cancer [26] and the Norwegian general population [27].

The Expanded Prostate Cancer Index Composite short form (EPIC-26) is a disease-specific instrument comprises 26 items across the following five domains: urinary incontinence; urinary irritation/obstruction; bowel function; sexual and vitality functions; hormonal disturbance. In addition to measuring functioning within each domain, the instrument also assesses the extent to which patients perceive these areas as problematic, with a recall time the last four weeks [28]. Domain scores were transformed to a 0–100 scale, with higher scores representing better quality of life and functioning [28, 29]. Evidence supports the validity of the EPIC-26 for use in Norwegian non-metastatic PCa patients undergoing radical prostatectomy or prostatic radiotherapy [29].

Descriptive statistics were conducted using IBM SPSS Statistics Version 26.0 software (IBM Corp, Armonk, NY, USA). Paired t-test were performed with latent change scores in a structural equation model (SEM) framework [30], using the R package Lavaan version 6.12 [31]. The SEM models were estimated by using Maximum likelihood.

Change between T1 (pre-treatment) and T2 (post-treatment) was estimated using a latent change score model (LCS), in which a latent variable represents the change for each outcome. This approach was chosen because it allows estimation of correlated change across multiple PROMs within a single multivariate framework and handles missing data efficiently using Full Information Maximum Likelihood (FIML). FIML yields unbiased estimates under missing at random (MAR) assumptions [32], and including outcome variables in the same model provides auxiliary information that supports the plausibility of MAR [33]. Age, SEER stage, and treatment type were not included as auxiliary variables to support the MAR assumption for FIML. Because an LCS model with time points is fully saturated, with zero degrees of freedom and a perfect fit to the data, model fit indices are not informative (df = 0) and no fit indices are reported.

The magnitude of change is expressed by Cohen’s d which was calculated by dividing the change score of the outcome variable by the standard deviation of the baseline score on this variable. A Cohen’s d of 0.20 was considered small effect size, 0.50 as medium effect size and 0.80 as large effect size. Correlations of 0.1, 0.3 and 0.5 were considered as small, medium and large effect sizes, respectively [34]. Figure 1 illustrates the latent change score model structure for two outcome variables (A and B). A visual summary, presented as a bar plot of effect sizes across instruments, was generated using R.

Afbeelding vergroten

For the PCa group, changes in EQ-5D-5L domain scores from T1 to T2 were statistically significant but small. Three domains (Self-Care, Usual Activities, Pain/Discomfort) increased slightly, suggesting modest deterioration (Cohen’s d = 0.12–0.19), while Mobility showed a slight improvement (Table 2). Changes were below the threshold for small effect sizes. The reference group also showed small deteriorations from T1 to T2 (effect sizes 0.04–0.17), consistent with minor age-related changes.

The mean EQ-5D-5L utility index for PCa patients decreased from 0.93 to 0.91 (mean change − 0.02; d =  − 0.20), indicating a small deterioration in overall health status. The reference group showed similarly small declines (0.92 to 0.91; d =  − 0.13) (Table 2).

For QLQ-C30, most domains showed minor changes (mean change < 5 points). The exception was social functioning, which declined from 88.4 to 78.6 (mean change − 9.85; d =  − 0.52; p < 0.001), indicating a clinically important deterioration (Table 3).

EPIC-26 showed the largest and clinically most meaningful changes from baseline (T1) to 6 months after treatment (T2). Sexual function declined substantially (65.2 to 27.1; mean change − 38.0; d =  − 1.42; p < 0.001), and urinary incontinence scores declined from 92.7 to 68.4 (mean change − 24.3; d =  − 1.86; p < 0.001), indicating markedly worse urinary control. Urinary irritation/obstruction increased modestly (mean change 4.3; d = 0.30), suggesting more urinary difficulties (Table 4). Changes in the other EPIC-26 domains were minor.

A latent change score model examining associations between changes in EQ-5D-5L and the disease-specific measures is presented in Table 5. Change in social functioning (QLQ-C30) showed small-to-moderate correlations with several EQ-5D-5L domain scores as well as the utility index and VAS (r values up to 0.40). In contrast, correlations between changes in EQ-5D-5L scores and changes in sexual function and urinary incontinence (EPIC-26) were consistently small (r < 0.25), indicating limited overlap between these constructs.

Correlations among the disease-specific instruments (EORTC QLQ-C30 and EPIC-26) were also small across the functional and symptom domains (r ≤ 0.30), as shown in Supplemental Tables 1–2.

Analyses stratified by SEER stage indicated that EQ-5D-5L was more sensitive to differences by disease severity across all domains and the summary score. However, limited power prevented further subgroup analyses.

Strengths of the present study include the population-based design and the availability of diagnostic factors available at the time of diagnosis, though we had limited in-depth information on type of PCa treatment and severity of disease (SEER). The present study is part of a three-year survey on men’s prostate-related health. We have presented and discussed the preliminary and final results with the steering group, including individuals with lived experience in the survey on men’s prostate-related health. As this study has a methodological focus, we have not had continuous involvement of with lived experience in the research process.

Although the response rate is quite reasonable for PROMS in a nationwide registry study, some degree of selection bias cannot be ruled out. Non-responders in large epidemiological studies tend to have more risk factors associated with vulnerable health [48]. On the other hand, it might be that the non-respondents had fewer physical and psychosocial health issues and thus were less motivated to participate in the study than those who participated. Unfortunately, we do not have information on reasons for non-participation, and the nature of any potential bias, if present, remains uncertain.

Our population-based cohort is likely to be more representative of the patient population encountered in clinical practice than those drawn from clinical trials. If selection bias influenced the prevalence of symptoms and functional limitations reported by participants with PCa, it may also have affected estimates of responsiveness to change of the EQ-5D-5L. Assessing these measurement properties was the main focus of our study. In any case, further studies are needed to confirm the performance of EQ-5D-5L when used in other PCa populations.

Finally, we acknowledge that use of the UK EQ-5D-5L value set may introduce some bias in utility estimation due to potential valuation differences between the UK and Norwegian contexts; however, evidence suggests that any impact on the direction or magnitude of the results is likely to be very modest.

Taken together, our findings indicate that although the EQ-5D-5L continues to provide value as a generic instrument, particularly for cross-disease comparisons and health-economic evaluations, its limited coverage of prostate-cancer–specific domains reduce its capacity to reflect the treatment-related changes experienced in this population. In this real-world cohort, clinically meaningful declines were captured only by disease-specific PROMs, underscoring the importance of selecting instruments aligned with the symptom profile of PCa. These considerations are highly relevant when interpreting HRQoL data collected in clinical practice and research settings.

EQ-5D-5L indicated good health for participants with PCa both before and after treatment, failing to capture clinically relevant problem areas, particularly in terms of sexual function, urine incontinence, which are pertinent for health status. Consequently, the EQ-5D-5L should not be the preferred questionnaire for measuring health status in patients with PCa. If used, it should always be supplemented with a PROM specific for prostate cancer or cancer in general to provide a more complete picture of patient-reported health.