Health-related quality of life in prostate cancer

Abstract

Background. With new treatment options, the prognosis of prostate cancer (PCa) has improved in recent decades, and health-related quality of life (HRQoL) has become an important outcome of treatment. HRQoL scores are also essential for health economic analyses concerning treatment options for the disease. This study assesses HRQoL scores in different health states of PCa, compares the results obtained by different HRQoL instruments, compares the HRQoL of PCa patients with that of the general population, and explores factors associated with the resultant HRQoL scores. Material and methods. An observational cross-sectional study among PCa patients in the Helsinki and Uusimaa Hospital District between September 2009 and December 2010. A total of 630 PCa patients (aged 43–92) assessed their HRQoL with the generic 15D and EQ-5D, as well as the cancer-specific EORTC QLQ-C30 questionnaires. Patients were divided into five mutually exclusive groups based on disease state: Loc1 (local disease, first six months after diagnosis; n = 47), Loc2 (local disease, 0.5–1.5 years after diagnosis or recurrence; n = 158), Loc3 (local disease, more than 1.5 years after diagnosis; n = 317), Metastatic (after detection of metastases; n = 89) and Palliative care (n = 19). Multivariate analysis served to evaluate the factors associated with the HRQoL scores. Results. The utility scores were highest at baseline. Markedly impaired HRQoL was seen first at the more advanced states of the disease. All HRQoL instruments studied were consistent in all states of the disease, yet the HRQoL scores obtained varied widely. Symptoms of fatigue and pain, and background variables of financial difficulties and age were the most important factors associated with poor HRQoL. Conclusions. All instruments provided valuable insight into PCa patients’ overall HRQoL. Management of cancer-related symptoms is important in maintaining patients’ HRQoL, but more attention should also focus on financial difficulties.

View correction statement:

Corrigendum

Prostate cancer (PCa) is the most common type of cancer among men, and its burden in both economical and clinical terms is considerable [1,2]. PCa is a disease found mainly in men over 60 years of age. Although PCa patients are relatively old on average, the latest five-year relative survival ratio in Finland, with 93% survival rates, is promising [3].

With better treatment options for an ageing population, interest has been growing in issues concerning psychological distress and health-related quality of life (HRQoL) beyond mere survival. Many treatment alternatives are currently available from which patients, under the guidance of clinicians, may choose the best options for them. Randomized studies and systematic reviews usually compare the effectiveness of different treatment procedures to each other, normally in terms of survival, but thus far none of the existing therapies, including watchful waiting, have proved transcendent [4–6]. Consequently, even with a variety of treatment options, controversy about the selection of primary treatments remains. Furthermore, studies reporting on HRQoL issues, which help patients to choose the best treatment option regarding quality of life, have thus far been few.

One suggested research approach has therefore sought not only to focus on patients’ clinical response to a given treatment, but also to study which treatments succeed in maintaining patients’ HRQoL [7]. Furthermore, to improve patients’ well-being, it is essential to identify the predictors of poor quality of life among PCa patients.

Multiple HRQoL instruments have been developed during the past three decades, but thus far none of them has emerged as a preferred option or as a gold standard. To evaluate differences between instruments and their domains within PCa patients, we studied two generic instruments and one cancer-specific instrument.

Our study was designed to answer three questions: 1) What is PCa patients’ HRQoL compared to that of the general population; 2) do all HRQoL instruments used here provide similar results; and 3) how do different background factors affect patients’ HRQoL.

Methods

The data were collected in the Helsinki and Uusimaa Hospital District that provides specialist medical care for the approximately 1.4 million inhabitants of Southern Finland. The study was a cross-sectional observational survey approved by the local Ethics Committee (registration number 207/13/03/02/2008). Patients were enrolled between September 2009 and December 2010.

Patient sample

Patients over 18 years of age diagnosed with PCa were eligible for the study. A research nurse identified patients from hospital records by date of diagnosis; questionnaires were then mailed to them. Recently diagnosed patients and those receiving only palliative care were enrolled when visiting the hospital. All patients meeting the inclusion criteria were invited to participate. Non-respondents received one reminder letter.

Clinical background information regarding patients’ disease stage and treatments administered within the last three months were collected from hospital records. Patients were divided into five mutually exclusive groups based on disease state: less than six months after diagnosis (Loc1), the following 12 months (Loc2), subsequent years of remission (Loc3), metastatic disease (Metastatic), and palliative care (Palliative).

HRQoL questionnaires

We used two generic self-administered HRQoL instruments: the 15D and the EQ-5D, as well as the cancer specific EORTC QLQ-C30 questionnaire. All instruments are widely used, validated, and standardized.

The 15D is a generic instrument with 15 dimensions (mobility, vision, hearing, breathing, sleeping, eating, speech, excretion, usual activities, mental function, discomfort and symptoms, depression, distress, vitality, and sexual activity) with five possible responses for each. The 15D index score ranges from 0 (= being dead) to 1 (= full health), and for this estimation we used the Finnish valuation algorithm for utility scores and profiles. The minimal clinically important difference (MID) in the 15D has been estimated at 0.03 [8,9].

The EQ-5D is a generic five-dimensional utility instrument which also includes a visual analog scale (VAS). For each of the dimensions (mobility, self-care, usual activities, pain or discomfort, and anxiety or depression), the responder can choose from three different options. We used the most commonly used UK time-trade-off (TTO) tariff, which generates a utility score between -0.594 ( = worst health, 0 = being dead) to 1.0 ( = full health). The VAS produces patients’ self-perceived estimate of their HRQoL on a vertical 100-point scale (0 = worst imaginable health state; 100 = best imaginable health state). The MID for the EQ-5D TTO is about 0.08 [10].

The cancer-specific EORTC QLQ-C30 yields a global health status and five functioning scales (physical, role, social, emotional, and cognitive functions). Fatigue, nausea/vomiting and pain are presented on their own scales, as do six single symptom items (dyspnea, insomnia, appetite loss, constipation, diarrhea and financial difficulties). EORTC produces symptom and functioning profiles, but not utility values. The scoring was performed according to the EORTC Scoring Manual [11,12].

Patients’ background information was collected simultaneously with questionnaires, which covered details regarding marital status, education, occupational status, treatments received outside hospital, and informal care. We compared utility scores to scores from the general population obtained in the representative Finnish Health 2000 Health Examination Survey [13].

Analysis

Key demographic characteristics are reported as proportions and HRQoL utility scores in different states of PCa as unadjusted means, standard errors and confidence intervals. The mean EQ-5D and 15D scores and the 15D profile for each disease state were compared with those of the general population using the Student's independent samples t-test. For comparison, the samples of the general population were weighted to reflect the age, gender, and education distribution of the patient samples.

The determinants of HRQoL (scores from EQ-5D, 15D and VAS) were analyzed using the ordinary least square regression model. Separate models for localized and metastatic PCa were built in a stepwise regression, in which the choice of predictive variables is carried out by an automatic procedure. Clinical and demographic factors and EORTC symptoms served in the models as explanatory variables. We performed the statistical analyses using SPSS 20 [14], and p-values ≤ 0.05 were considered statistically significant.

Results

Study population

We approached a total of 1025 patients, of whom 630 (61.5%) responded. Of these, 522 patients had local disease, of whom 47 were in the Loc1 group, 158 in the Loc2 group, and 317 in the Loc3 group. Of the 108 patients with metastases, 89 were receiving oncologic treatments (Metastatic group), and 19 palliative treatments only (Palliative group) (Table I).

Table I. Patient characteristics.

	Loc1	Loc2	Loc3	Metastatic	Palliative	All patients
Respondents (n)	47	158	317	89	19	630
Demographic factors
Age, mean (SD)	68.5 (8.2)	68.3 (8.2)	69.1 (7.8)	71.4 (7.9)	75.5 (8.3)	69.4 (8.1)
Higher education (%)	28 (59.6)	52 (32.9)	173 (54.6)	48 (53.9)	7 (36.8)	279 (55.4)
Married/cohabiting (%)	38 (80.9)	80 (50.6)	256 (80.8)	69 (77.5)	14 (73.7)	457 (82.0)
Disease-related factors
Time after diagnosis, years (SD)	0.14 (0.5)	1.2 (0.1)	3.4 (1.0)	5.4 (4.4)	8.1 (4.5)	3.0 (2.6)
Time after metastases, years (SD)	–	–	–	2.2 (2.2)	2.6 (2.0)	2.3 (2.2)
Bone metastasis (%)	–	–	–	81 (91.0)	18 (94.7)	99 (15.7)
Treatments, within last 3 months
Radiotherapy (%)	0 (0.0)	3 (1.9)	2 (0.6)	10 (11.2)	8 (42.1)	23 (3.7)
Antiandrogen therapy (%)	0 (0.0)	7 (4.4)	16 (5.0)	19 (21.3)	3 (15.5)	45 (7.1)
LHRH-analog (%)	1 (2.1)	25 (15.8)	25 (7.9)	64 (71.9)	5 (26.3)	120 (19.0)
Chemotherapy (%)	0 (0.0)	0 (0.0)	3 (0.9)	36 (40.4)	3 (15.8)	42 (6.7)
Zoledronic acid (%)	0 (0.0)	0 (0.0)	0 (0.0)	17 (19.1)	2 (10.5)	19 (3.0)

Patients’ ages ranged from 44 to 93 years (mean 69). Most of them were married or cohabiting (82%) and had higher education (55%). The mean time after diagnosis was 3.0 years (Loc1 0.14; Loc2 1.2; Loc3 3.4; Metastatic 5.4; Palliative 8.1). Bone metastases were found in 92% of patients with metastatic disease (Table I).

Within the last three months, 19% of the patients had received luteinizing hormone releasing hormone (LHRH) analog treatment, which was a widely used option in both the Metastatic and Palliative groups (72% and 26%). Patients with metastatic disease underwent chemotherapy (Metastatic group 41%) and radiotherapy (Palliative group 42%) as well as zoledronic acid treatment (Table I).

HRQoL

The 15D scores of the study population ranged from 0.34 to 1.00 (mean 0.87). Altogether 30 patients (5%) were in full health (i.e. their 15D score was 1). The EQ-5D scores of the study population ranged from -0.166 to 1.00 (mean 0.84). With the EQ-5D, a ceiling effect was evident: 266 patients (42%) were in full health. The VAS score ranged from 1 to 100, with a mean of 76.4 (Table II).

Table II. Mean health scores in different disease states.

	EQ-5D				15D				VAS
	Mean	n	SD	95% CI	Mean	n	SD	95% CI	Mean	n	SD	95% CI
Loc1	0.90	46	0.19	0.84–0.96	0.91	47	0.09	0.89–0.94	81.11	46	14.87	76.69–85.52
Loc2	0.89	91	0.14	0.86–0.92	0.89	157	0.09	0.87–0.90	76.73	92	16.69	73.27–80.18
Loc3	0.87	309	0.19	0.85–0.89	0.88	310	0.11	0.87–0.89	79.80	311	16.24	77.99–81.62
Metastatic	0.74	85	0.27	0.69–0.80	0.80	88	0.12	0.78–0.83	66.79	86	21.80	62.12–71.46
Palliative	0.59	17	0.22	0.48–0.70	0.67	19	0.10	0.62–0.72	48.67	18	19.10	39.17–58.17
All patients	0.85	621	0.03	0.83–0.86	0.87	621	0.11	0.89–0.94	76.36	553	18.56	74.81–77.91

The mean scores of all instruments were consistently lower in the more advanced disease states (Table II). However, as long as the disease was local, HRQoL scores remained close to the baseline level; significantly impaired HRQoL scores were seen only after the disease had progressed to the metastatic stage. Both generic instruments gave higher scores in the Loc1 and Loc2 groups than those of the general population (EQ-5D also in Loc3 state) (Table III). In the Metastatic state the mean EQ-5D and 15D scores were both clinically and statistically significantly lower than in the Loc3 state, and in the Palliative state, lower than in the Metastatic state. Lowest mean values were thus seen in the end of life care. The 15D profile highlighted the difference between local and metastatic disease in terms of decrements on different dimensions of HRQoL (Figure 1).

Figure 1. 15D Profiles in different states of PCa.

Table III. PCa patients’ HRQoL compared to that of the general population, 15D, EQ-5D.

	Loc1 ∆	Loc2 ∆	Loc3 ∆	Metastatic ∆	Palliative ∆
EQ-5D	0.103***	0.089***	0.043***	−0.054*	−0.157**
15D index	0.025	0.012	−0.001	−0.068***	−0.177***
15D dimensions
Mobility	0.019	0.033***	0.012	−0.066**	−0.260***
Vision	0.008	0.036***	0.005	−0.003	−0.102*
Hearing	−0.013	0.018	0.005	−0.026	−0.183**
Breathing	0.017	0.037	0.012	−0.054*	−0.104
Sleeping	0.012	−0.009**	−0.026*	−0.058*	−0.121*
Eating	−0.005	0.009	0.003	0.003	−0.032
Speech	0.022***	0.017	0.012	0.019*	−0.056
Elimination	−0.010	−0.044**	−0.043***	−0.134***	−0.293***
Usual activities	0.029	0.047***	0.015	−0.106***	−0.272***
Mental function	0.081***	0.048***	0.037***	0.021	−0.045
Discomfort and symptoms	0.082***	0.065***	0.046***	−0.080***	−0.180***
Depression	0.032	0.016	0.009	−0.026	−0.137***
Distress	0.003	0.001	−0.010	−0.029	−0.165***
Vitality	0.036	0.010	−0.003	−0.110***	−0.255***
Sexual activity	−0.042	−0.229***	−0.205***	−0.446***	−0.544***

∆, difference compared to the general population (positive for better scores compared to general population).

* = p-value < 0.05; ** = p-value < 0.01; *** = p-value < 0.005.

Of the 15D dimensions, only sleeping, excretion, and sexual activity were statistically significantly worse in the Loc2 group than in the general population. In the Loc1 group, patients fared better than did the general population on the dimensions of speech, mental function, and discomfort and symptoms, and in the Loc2 group on the dimensions of mobility, vision, usual activities, mental function, discomfort and symptoms (Table III).

The EORTC QLQ-C30 showed that the patients with metastatic disease exhibited more symptoms than did patients with local disease; the Palliative group was the most symptomatic (Figure 2). Reported symptoms followed the progression of the disease stages. Only pain, diarrhea and financial difficulties were reported more often in the Loc1 group than in the Loc2 group, and constipation occurred in the Loc2 group more often than the Loc3 group.

Figure 2. EORTC QLQ-C30 functionality scales in different states of PCa.

The mean EORTC QLQ-C30 Global Health score was 75 (range: Loc1 81 to Palliative 49). The EORTC yielded different functioning scores that were as consistent as other HRQoL instruments studied here (Figure 3).

Figure 3. EORTC QLQ-C30 symptom scales in different states of PCa.

Determinants affecting HRQoL

Our first models in localized and advanced disease showed that EORTC symptoms and background variables affected HRQoL, but results varied across instruments (Table IV). The adjusted R square was relatively high, ranging from 0.471 to 0.738 (Table IV). Our analysis based on a localized PCa model showed that fatigue, pain and financial difficulties were significant factors influencing HRQoL assessed with all of the instruments, and dyspnea, insomnia, and age assessed with two of the three instruments. In an advanced disease model, fatigue and pain lowered the HRQoL scores as assessed with all instruments, and age lowered HRQoL as assessed with the 15D and VAS.

Table IV. Factors associated with HRQoL in PCa.

	EQ-5D		15D		VAS
	Std coefficent	P-value	Std coefficent	P-value	Std coefficent	P-value
Localized disease	R^2 = 0.564		R^2 = 0.738		R^2 = 0.471
Fatigue	−0.003	***	−0.002	***	−0.318	***
Dyspnea			−0.001	***	−0.098	***
Pain	−0.003	***	−0.001	***	−0.102	***
Insomnia			−0.001	***	−0.078	***
Age	−0.002	***	−0.002	***
Financial difficulties	−0.001	***	−0.001	***	−0.117	***
Diarrhea			−0.001	***
Watchful waiting****			−0.023	***	−3.903	*
Cohabiting	0.037	*
Higher education	0.028	*			2.530	*
Appetite loss	−0.002	***
Antiantrogen treatment					−8.439	***
Time after diagnosis/recurrence					0.940	*
Constipation			−0.001	***
Nausea vomiting			−0.001	*
Metastatic disease	R^2 = 0.714		R^2 = 0.712		R^2 = 0.492
Age			−0.002	*	−0.002	*
Fatigue	−0.005	***	−0.003	***	−0.320	***
Financial difficulties	−0.002	**
Watchful waiting****	0.145	***
Dyspnea			−0.001	***
Pain	−0.003	***	−0.001	***	−0.300	***
Antiantrogen treatment	−0.082	*

* = p-value < 0.05; ** = p-value < 0.01; *** = p-value < 0.005.

****watchful waiting is defined here as patients who do not receive any cancer treatments/operations.

In a further analysis, we excluded EORTC symptoms and focused on background variables. This revealed that higher education, age and financial difficulties were the most important factors associated with HRQoL scores. In advanced disease, financial difficulties and age proved to be the most significant variables. Nonetheless, adjusted R square was lower than in the first models, in which symptoms were included.

Discussion

HRQoL is currently an important element in the selection of treatment for PCa. Most HRQoL studies in PCa have focused on presenting results from one or more treatment alternatives at the same state of disease. We attempted to capture a more holistic view by studying patients from all disease states and compared the results of several widely used HRQoL instruments. Accurate assessment of HRQoL and an understanding of the factors related to different HRQoL instruments are critical to patient counseling as well as to the decision making process, both at the clinical level and in health economic assessment.

Unsurprisingly, our study showed that the HRQoL of PCa patients differs in different states of the disease. However, as long as the disease remained localized, patients’ HRQoL remained at a relatively high level. Both generic instruments produced higher scores in the Loc1 and Loc2 groups – and the EQ-5D also in the Loc3 group – than those found among the general population standardized for gender and age. One explanation for this finding could be that a significant proportion of patients enter PCa treatment because of elevated prostate-specific antigen (PSA) levels found in opportunistic testing. As PSA testing has not been recommended at the national level, such opportunistic testing in Finland is currently limited mainly to occupational health services.

Based on the Finnish national level health survey, localized PCa patients also valued their HRQoL higher than did patients with other types of cancer: in our study, the mean 15D score was 0.89, and the mean EQ-5D score was 0.87, compared to the mean values for other cancer types (0.86 and 0.74, respectively) [15]. PCa patients with local disease, by their own judgment, are evidently doing better than other cancer patients or their age peers. National level HRQoL comparison is collected a decade ago, and our comparison is the mean value for other cancer types, it is therefore fair to say that this comparison is only indicative.

In patients with Metastatic disease, the HRQoL scores were clinically and statistically significantly lower than in those with localized disease. This deterioration was most pronounced in patients in whom the disease had progressed to the Palliative care state, in which the mean utility scores were the lowest found in our study: 0.67 for the 15D, and 0.59 for the EQ-5D, respectively. In light of previous studies, this was an expected finding. However, comparison between states must be performed with caution, as this was a cross-sectional study with different patients in groups representing different states.

The cross-sectional study design is probably the most important limitation of our study. It may have been better to follow the same patients throughout their disease progression, but this would have taken much more time, and we probably would have needed to enroll far more patients to have a sufficient number of patients in later phases of the disease. Also the response rate of 61.5% may be seen as a limitation although it is comparable with that seen in other studies. It is possible that the non-respondents may have had more severe disease, but due to the study design we did not have access to their medical records to verify whether that is true. Based on our previous experiences with similar surveys, we, however, have no reason to expect that non-respondents would, on average, be significantly different from respondents regarding disease severity.

One of our goals was to understand better how different HRQoL instruments are able to capture different aspects of PCa patients’ HRQoL, and how the results of the instruments differ from each other. Although both generic instruments used in this study (the 15D and the EQ-5D, including VAS) yielded consistent results commensurate with disease progression, some interesting discrepancies also occurred. One of the most evident was the ceiling effect: 42% of the patients were in full health according to the EQ-5D, but only 5% according to the 15D. Previous studies have also identified ceiling effects, which is a result of the different features of the instruments, mainly the different descriptive systems and ranges of health states and the shape of the distribution of scores [16–21].

The 15D profile revealed that discomfort and symptoms, excretion, sexual activity, and usual activities were the dimensions in which most of the patients’ problems, compared to general population, occurred. In metastatic disease, the mean dimension scores of mobility and vitality, and in local disease, mental function were statistically significantly worse than in the general population. The 15D profile also highlighted the difference between local and metastatic disease in terms of decrements in different dimensions of HRQoL.

For our analysis of factors affecting HRQoL across different disease states and different instruments, we developed two linear multiple regression models: one for local disease and the other for metastatic disease. The analyses revealed that, beyond symptoms, financial difficulties and age were the most important explanatory factors impairing HRQoL in both regression models. These results are in line with those of earlier published studies [22,23]. Treatment options failed to explain the variance in patients’ HRQoL values. Further research is still needed to fully understand factors influencing HRQoL. Our ordinary least squares (OLS) regression approach may have its limitations as the distributions of the scores produced by the generic instruments analyzed here failed to satisfy the distributional assumptions required by traditional OLS. Although, based on the literature, the choice of using a more complex regression method has rather little significance for the 15D, this may not hold true for the EQ-5D [15]. Nevertheless, further studies are needed to fully understand the challenges of using different methodological approaches and to reveal which HRQoL instrument is best suited for PCa.

Our study provided estimates of PCa patients’ HRQoL from two generic and one cancer-specific instrument. Our study contributes to the current literature in that our results can be used both on the clinical level, as well as in the health economic evaluation of treatment interventions. Further work is still needed, however, a similar patient grouping approach can also serve when evaluating costs among PCa patients’ disease states.

Conclusion

As the incidence of PCa, and its treatment options, have increased in the past two decades, better understanding how the disease affects HRQoL in a real world setting is vital. Our study provides information on the HRQoL of PCa patients in various states of the disease. The HRQoL of PCA patients appears to be surprisingly good prior to metastatic progression of the disease. Comparisons of widely used HRQoL instruments have revealed that they produce somewhat different utility scores, but these instruments have otherwise performed in a fairly similar manner commensurate with the disease progression. The widely used EQ-5D, however, produced a marked ceiling effect, with 42% of PCa patients reporting full health (i.e. a utility score of one). Nevertheless, all instruments studied are applicable, despite some known limitations, for use in evaluating the HRQoL of PCa patients. However, policy makers must be aware of these features when interpreting the results of health economic analyses, as results may differ depending on the instrument used. One of the key findings in our study was that, beyond the rather obvious factors such as symptoms and age, financial difficulties seem to be the most important determinant related to the poor HRQoL of PCa patients. Consequently, any means to remove financial barriers to high-quality treatment could yield higher HRQoL results in future.

Acknowledgement

We would like to thank Stephen Harris for linguistic assistance.

Declaration of interest: This work was supported by the Cancer Society of Finland and GlaxoSmithKline Oy, Finland. No study sponsors were involved in the study design or in the collection, analysis, or interpretation of data, nor were they involved in the writing of the manuscript or in the decision to submit the manuscript for publication. However, ST and NF are employees of GlaxoSmithKline. All authors have contributed to the study conception and design, the acquisition, analysis and interpretation of the data, and the writing the manuscript.