Introduction

Worldwide, in 2012, there were 14.1 million people living with cancer, 7.4 million in men and 6.7 million in women [1]. By 2035, an estimated, 24 million individuals will be diagnosed with cancer [1]. Technological advances have played a key role in improving survival through early detection and better treatment [2, 3]. As a result, approximately 75 % of people with cancer will now live 5 years or more [2, 3], but still many will experience a wide range of physical and psychological sequelae from the disease process itself and its treatment [4, 5]. These sequelae matter to patients and impact negatively on their quality of life (QOL).

It remains a challenge in a busy clinical setting to systematically obtain information on the different symptoms, functions, and health aspects of QOL. As a result, these concerns are often not systematically queried nor documented despite the fact that this information is important to guide patient-centered care.

One measurement technology designed specifically to obtain personalized information on areas of concerns to patients is personalized QOL measures [6–10]. The literature indicates that multi-item questionnaires [11–13] correlate only moderately with personalized measures indicating that standardized measures may not capture what is important to individuals. One such measure constructed to capture important concerns of the patient is the Patient Generated Index (PGI).

PGI is a personalized measure designed to both query and document QOL concerns and, thus, could potentially be a valuable clinical and research tool to evaluate changes in health outcomes in patients undergoing cancer care. However, its validity with respect to more standard QOL measures has not been fully established in the context of advanced cancer when QOL concerns predominate. The global objective of this study is to contribute evidence to validate the PGI as quality of life measure in people with advanced cancer. The specific objective is to identify similarities and differences in ratings of global quality of life between personalized and standard measures.

Methods

Subjects

The data for this study comes from a study of anorexia/cachexia in people with advanced cancer, and the data collection included a comprehensive assessment of factors contributing to overall quality of life [14].The target population of this study was people with advanced cancer of any origin. The sample was recruited from two tertiary care university hospitals. This was a cross-sectional study using data from the first assessment post-diagnosis, before start of oncology treatment.

The sample included people with unresectable stage 3A, 3B, or 4 non-small cell lung cancer (NSCLC); stage 3 or 4 upper gastrointestinal (GI) cancer; stage 4 colorectal, hepatobiliary, or head and neck (ENT) cancers; breast and prostate cancers with visceral metastases; all stages of pancreatic cancers; an estimated life expectancy of 3 months or more; and an Eastern Cooperative Oncology Group performance status score of 0 to 3 [15, 16]. People with symptomatic brain metastases were excluded from the study as well as those who were not able to follow the instructions.

Measures

The measurement framework for this study was based on the Wilson-Cleary model [17] of health-related quality of life. In addition to the personalized measure, both generic and cancer-specific QOL measures were used. Measures were chosen based on brevity, comprehensiveness, and inclusion of items on quality of life in addition to symptoms and function.

Personalized measure

The PGI, a personalized measure to identify the impact of specific conditions on QOL [7], has been used and validated in both noncancer populations [7, 18] and a variety of cancer populations [9, 10, 19–21].

PGI is completed in three steps: (1) patients identify the most important five areas of their life affected by cancer; (2) patients rate how much each area has been affected using a scale from 0 to 10, where 0 is the worst imaginable and 10 exactly as they would like it to be and how much; (3) patients now imagine that they have 12 tokens to spend to improve on the selected areas and they allocate these tokens to the areas according to their own priority. A global index is calculated by multiplying the ratings for each area in step 2 by the proportion of tokens given to that area in step 3, which are then summed to produce an index from 0 to 100 with higher scores indicating higher QOL. According to Ruta [7], the final score indicates the extent to which the reality falls short of patient’s hopes and expectations for those areas of life for which they would most value an improvement [7, 22].

Generic health-related quality of life measures

Two measures were used: the EQ-5Dindex from the EuroQoL group [23, 24] and the SF-6D derived from the comprehensive SF-36 Health Survey. The EQ-5D™ is a brief generic measure that provides a descriptive system consisting of five dimensions (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression). Each dimension has three levels of response: no problems, some problems, and extreme problems. The EQ-5D™ describes 243 health states each with a unique index value calculated using country-specific weights [25–27] with values for EQ-5Dindex ranging from −0.6 to 1.0 [25, 28]. The EQ-5Dindex has been used extensively in population health and cancer research [29–32].

The SF-6D is a generic measure (39, 40) with six multi-level dimensions covered by the SF-36 Health Survey: physical functioning, role limitation ((combined role-physical and role-emotional)), social functioning, bodily pain, and mental health. Each dimension contains four to six levels of function or limitation. Therefore, the SF-6D index classifies 18,000 unique health states. The SF-6D index is calculated using preference weights obtained from a sample of the general population using the standard gamble and ranges from 0.3 (worst health state) to 1.0 (best health state) [33]. The SF-6D has been widely used and validated in cancer population [34–39].

Cancer-specific quality of life measures

Two single-item and one multi-item measures were used. The single items were: (1) Edmonton Symptoms Assessment Scale (ESAS-Version 1) [40] QOL, with a rating scale from 0 to 10 with 10 being the worst possible QOL, and (2) McGill Quality of Life Questionnaire (MQOL) [15, 41–44], also with a 0 to 10 rating scale with 10 being the best QOL. The multi-item questionnaire was the existential subscale from the MQOL with six items, each rated on 0 to 10, scale with 0 meaning the worst and 10 being the best. The ESAS is a brief cancer-specific measure of symptom severity and it is a valid and reliable tool to help in the assessment of nine symptoms commonly experienced by cancer patients (pain, fatigue, nausea, depression, anxiety, drowsiness, appetite, itching, and shortness of breath) and one item for overall quality of life [45, 46].

McGill QOL questionnaire was developed to measure QOL at all stages of the disease trajectory for people with a life-threatening illness such as cancer [15, 41–44]. There are five MQOL sub-domains: physical symptoms (items 1–3), physical well-being (item 4), psychological symptoms (items 5–8), existential (items 9–14), and support (items 15 and 16). The existential sub-domain captured aspects related to meaning of life considering the last 2 days. Many studies have shown that the total score and the sub-domain scores of the MQOL have a good validity (construct validity) and reliability (internal consistency reliability) for people with cancer and other palliative populations [15, 41–43, 47].

All of the measures used in this study, with the exception of the ESAS and the MQOL, have legitimate total scores [48] based on weighting the dimensions before summing. Therefore, from the ESAS and MQOL, we chose the two single items for QOL and the one sub-domain (existential) of the MQOL.

Procedure

This study was approved by the Institutional Review Board of the Faculty of Medicine of McGill University. People who met the study criteria were approached by a member of their primary oncology and medical team in order to obtain verbal consent. Once consent was granted, study personnel contacted the patients to explain the goals and procedures of the study. All participants gave written informed consent. At the first assessment, patients were interviewed on all study measures by trained research personnel.

Analysis

The first step was to identify the most frequently nominated areas using the PGI measure and to calculate the PGI score. Secondly, scores were calculated for all the measures and comparisons between personalized measure (PGI), generic measures (SF-6D, EQ-5Dindex), and the cancer-specific measures (ESAS-qol, MQOL single items, and MQOL existential) made using generalized estimating equations (GEE) to account for the clustering of the questionnaire scores within person. All measures were transformed to range from 0 to 100 with 100 being the best QOL.

Spearman’s correlation coefficient was used to measure the strength of the association between the PGI and the other QOL measures. The regression coefficients from the GEE model were used to calculate the difference between measures and the corresponding 95 % CIs, accounting for the correlation among measures. Also calculated from the regression coefficients were effect sizes (β/SE).

The Bland-Altman plot was used to depict differences between the measures under study here: the PGI and the EQ-5Dindex, and the PGI and the SF-6D. While developed to depict agreement, in this context, the graphic representation permits a visualization of the variability in the magnitude of the latent variable QOL arising from using different measures of the latent construct. The latent construct is represented, on the x-axis, by the average of the scores of the two measures. The unique contribution to the latent construct from the different measures is shown on the y-axis as the difference between the two measures. If both measures equally represent the latent construct throughout its range, there will be no discernible pattern to the distribution of the difference along the latent construct.

Results

A total of 192 persons completed the PGI at baseline, and their characteristics are shown in Table 1. Table 2 lists the 11 most frequent areas of QOL affected by advanced cancer, at time of diagnosis, as documented using the PGI. Also indicated is whether or not the areas of QOL concern are present in the other study measures: EQ-5Dindex, the SF-6D, the MQOL, and the ESAS. Overall, participants nominated 114 areas of QOL concern, using the PGI.

Table 1 Demographic and clinical characteristics of the study population
Full size table
Table 2 The most frequent areas identified by cancer participants using PGI compared with items in generic measures and specific disease measure
Full size table

Participants nominated a wide range of QOL concern affected by cancer with fatigue being the most common, nominated by 39 % of participants, to mobility, nominated by 3.5 %. Fatigue was present as an item in the SF-6D, the MQOL, and the ESAS but not in the EQ-5Dindex. Sleep function, the second most frequent area, was found only in the MQOL. Pain was the only area that was included in all of the measures. In the five top commonly nominated areas, the MQOL included all the areas (five areas) important to the people with advanced cancer, followed by the ESAS (four areas), the SF-6D (three areas), and then the EQ-5Dindex (two areas).

The average rating from the PGI was substantially lower (37 ± 25) than the average QOL ratings from generic measures (EQ-5Dindex (mean 66 ± 17) and SF-6D (mean 65 ± 14)). The correlations between the PGI and the generic measures were low (0.16 and 0.22 for EQ-5Dindex and SF-6D, respectively); the correlation between the two generic measures was moderate (0.53). The magnitude of the difference between the PGI and generic measure ranged from −28.5 to −27.5; effect size estimates ranged from −14.3 to −13.4 (see Fig. 1a). The results comparing the PGI with the cancer-specific QOL measures are shown in Fig. 1b. Correlations between the PGI and these measures were also low (range 0.12 to 0.18), and the magnitude of differences were similar for the MQOL single item (−26.7) and the ESAS single item (−26.1) and larger for the MQOL existential single item (−43.3).

Fig. 1
figure 1

The mean and standard deviation values for a PGI and generic measures with differences, Spearman’s correlation, and 95 % CI calculated using generalized estimating equations and b PGI and cancer-specific measures

Full size image

Figure 2 represents the relationship between the PGI and the generic measures (EQ-5Dindex and SF-6D) using the Bland-Altman plot. The y-axis represents the difference between the PGI and each measure, EQ-5Dindex (Fig. 2a) and SF-6D (Fig. 2b). The x-axis represents the latent construct of QOL as the average between the two measures. The gray line, which is parallel to the x-axis, represents the line of equality (0 differences between measures). The middle red line represents the mean difference between the two measures. The upper and lower red lines represent the corresponding 95 % CI. The black dots represent the score of each participant on both measures. At the left end of the x-axis when the participant’s QOL was poor, the score from the PGI was lower (except for one observation) than the score obtained from the EQ-5Dindex (Fig. 2a). This was also observed for the middle range (shown in rectangle), where the majority of the observations were below the line of equality. However, at the right end of the x-axis, when the participant’s QOL was good, there was no discernible pattern to the difference between the two measures. Similar results were found for the comparison between the PGI and SF-6D (Fig. 2b).

Fig. 2
figure 2

Relationship between the PGI and the generic measures: a Bland-Altman plot of the relationship of the PGI and the EQ-5Dindex and b Bland-Altman plot of the relationship of the PGI and the SF-6D

Full size image

Discussion

The PGI identified a total of 114 areas of QOL concern, with the top three, fatigue, sleep disruptions, and pain. Other studies that have used the PGI in cancer samples also found this wide range of areas [9, 19]. Our findings are concordant with the literature on the importance of fatigue, sleep, and pain to QOL, as the most recent systematic reviews endorsed their importance [48–53].

This heterogeneity in QOL contributors is not reflected in standard QOL measures, either generic or cancer specific, as the combined measures used in our study queried only nine areas. All of the measures used in this study except the EQ-5Dindex included fatigue; sleep was only included in the MQOL. Pain, the third most frequently endorsed area, was the only domain that was included in all of the measures. In contrast, two of the ten top areas identified using the PGI (eating, family relationships; see Table 2) were not represented in any of the standard measures used here. ESAS, a symptom measure, captured all the important symptoms identified by the participants, except sleep disruption.

One of the most frequently used cancer-specific measures is the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire-C30 (EORTC QLQ-C30) [54]. While we did not include this 30-item measure, its content has been shown to capture 25 domains related to symptoms and function and also includes three questions, one each on health perception, overall QOL rating, and financial impact of cancer [55]. The functional domains on the EORTC were all nominated using the PGI and are present as items in the measures chosen for this study (see Table 2). Another widely used measure the Functional Assessment on Cancer Therapy—General Scale (FACIT-G) [56] comprises 27 items that capture 6 emotional impairments and 4 physical impairments (pain, fatigue, nausea, and insomnia); 2 items for work; 1 for role; and the remaining 14 items reflect aspects of QOL, general health, and environmental factors. These areas were all identified in the PGI.

This study found that the PGI consistently produced values for QOL that were lower than standard generic or cancer-specific measures of QOL. One reason for this is the scoring system which not only rates the severity of the impact of nominated areas but also weighs them in terms of their priority for intervention. For example, a person would get a low QOL score if an area has a large negative impact and receives a lot of “tokens,” therefore identified as a very important priority. Another reason is that the PGI allows patients to mention important QOL contributors that may not be captured by other measures. The other measures that use weights to derive total scores (EQ-5Dindex and SF-6D) have obtained these from the general population, not from people with cancer [57]. The consequence is that QOL is overestimated if measurement relies only on standard “one-size-fits-all” measures and/or weights. The PGI could be useful in the clinical setting as the combination of weight and priority for each contributor would probably be more effective in orientating interventions to improve overall QOL.

There was a low correlation between the PGI and the other study measures. This finding is supported by earlier studies that found low to moderate correlation between the PGI and EORTC and the PGI and the SF-36 [9, 19]. Also, our finding is in line with a study done in multiple sclerosis using the same methodology [58]. This further supports the importance of exploring the priority attached by the patient to the QOL contributors.

This was a study of a unique sample compromised entirely of people with advanced cancer before any oncological treatment for whom QOL is of primordial concern for making treatment and life decisions. Because a restricted number of QOL measures were included in the study based on the fewest items per questionnaire, in order not over burden the participants, direct comparisons with the well-known 30-item EORTC and the 27-item FACIT could not be performed.

The PGI may be an interesting instrument to evaluate the impact of clinical trials that target QOL. The standard measures tend to yield higher QOL ratings, and hence, comparisons with baseline may be disappointing as there is limited room for improvement with intervention. It would not be unusual for a cancer clinical trial to use the SF-6D or EQ-5D as an outcome [59]. If so, the effect of the intervention in a sample of patients with similar QOL as our study, considering a ½ SD effect size [60], would mean that the SF-6D would need to improve from a mean of 66 to a mean of 75 (and the EQ-5Dindex, from 65 to 72), whereas the same effect size on PGI would require a change from 37 to 49, a more achievable change. For the cancer-specific measures, the SDs were as high as the PGI but the means were also higher indicating an important change would be difficult to achieve.

Conclusion

PGI is a personalized measure and was developed to identify the impact of a health condition on QOL. This study showed that, in comparison to commonly used generic and cancer-specific QOL measures, the PGI covers the widest spectrum of health concepts as it allows respondents to nominate, weigh, and prioritize any areas that are important to their QOL, not just those presented to them. The PGI has additional advantages in that it is short and easy to administer and yields a value farther away from the ceiling and with a lower standard error in comparison to standard measures. If only one QOL measure is to be included, either in a clinical setting or for research, the PGI would satisfy many of the criteria for “best choice” [61]. When used in a cancer population, the PGI could be considered a cancer-specific QOL measure.

Financial disclosure

None.

Funding

This study received funding from the Terry Fox Research Institute.

Conflict of interest

The authors declare that they have no competing interests.