The early predictive value of frailty for health-related quality of life among elderly patients with cancer receiving curative chemotherapy

Changes in health-related quality of life (HRQOL) among elderly patients with cancer before and after receiving curative treatment, such as chemotherapy, have always been an important consideration in physician–patient treatment decision-making. Although frailty assessment can help predict the effects of chemotherapy, there is a lack of relevant literature on its effectiveness in predicting post-chemotherapy HRQOL. Therefore, this study investigated the early predictive value of pre-chemotherapy frailty assessment for post-chemotherapy HRQOL among elderly patients with cancer receiving curative chemotherapy. From September 2016 to November 2018, this study enrolled elderly patients with cancer aged ≥ 65 years (N = 178), who were expected to receive chemotherapy at three hospitals in Taiwan. The mean age of patients was 71.70 years (SD = 5.46 years) and half of them were female (n = 96, 53.9%). A comprehensive geriatric assessment was performed to measure frailty in 178 participants one week before receiving chemotherapy (T0). Further, the HRQOL of the elderly patients with cancer was assessed again, four weeks after chemotherapy (T1). After controlling for demographic variables, this study evaluated the predictive value of frailty for HRQOL using a hierarchical regression analysis. A total of 103 (57.9%) participants met the frailty criteria. The results showed that 31.1%–56.7% of the variance in the seven domains of HRQOL could be explained by demographic variables and the presence or absence of frailty. This suggests that the presence or absence of frailty is an important predictor of the illness burden domain (β = 9.5; p < .05) of HRQOL. Frailty affects the illness burden domain of HRQOL in elderly patients with cancer. Finally, the administration of frailty assessments before treatment is recommended as a reference for patient treatment decision-making.


Introduction
In March 2018, the proportion of the elderly adults in Taiwan exceeded 14%, and Taiwan officially became an aging society. Taiwan is expected to become a super-aged society in 2030. Compared with Japan, which has the largest elderly population in the world, Taiwan is estimated to have transformed into a super-aged society three years earlier [1]. The increase in the elderly population has become an important issue in several countries worldwide [2]. Notably, two-thirds of the patients with cancer worldwide are older than 65 years [3], and it is estimated that the number of elderly patients with cancer will increase by 14 million by 2035 [4]. Since 1982, cancer has been the leading cause of death in Taiwan, and the median age at death has increased annually [5]. With an increase in the aging population and incidence of cancer, the importance of assessing elderly patients with cancer receiving anticancer treatment (e.g., radical resection, radiotherapy, and chemotherapy) has increased, as some patients cannot endure intensive treatment and thus experience severe life-threatening complications [6,7].
Although the effectiveness of chemotherapy in cancer control has been demonstrated in many clinical trials [8], treatment-related side effects can reduce the health-related quality of life (HRQOL) of patients with cancer, especially the elderly [9]. The HRQOL of elderly patients with cancer is unique [3]. Therefore, the European Organization for Research and Treatment of Cancer (EORTC) developed a specific measurement tool for HRQOL in elderly patients with cancer: the EORTC QLQ-ELD14 [3,10]. However, the relevant literature has mainly focused on cross-sectional studies [10][11][12], overlooking longitudinal studies.
Frailty refers to a clinical symptom in which the function and reserve capacity of multiple physiological systems are diminished, affecting the ability to maintain body stability while responding to stressful events or leading to several adverse postsurgical outcomes [13]. Scholars have emphasized that frailty is correlated with aging [14] and that the level of frailty in elderly patients with cancer is negatively correlated with HRQOL [15]. Scholars have conducted a systematic review of frailty-related studies; one study found that the mean prevalence of frailty in elderly patients with cancer was 43% [16], and another reported a rate of 68% [17]. Frailty assessment helps predict the effects of chemotherapy in patients with cancer [18]. Therefore, experts have encouraged elderly patients with cancer to undergo a frailty assessment before treatment [6,19] . A systematic review indicated that the CGA should include the assessment of at least five domains, and failure to meet the criteria of two domains should be considered the defining criteria of frailty [16].
Changes in the HRQOL of patients with cancer before and after treatment have always been an important issue for medical teams, patients, and their family members [7], and an important consideration in treatment decision-making for elderly patients with cancer [24]. Therefore, there is a need to investigate the predictive value of frailty for post-treatment HRQOL in elderly patients with cancer before chemotherapy. The results can be used as a reference for treatment-related decision-making by clinicians for elderly patients with cancer (providing anticancer treatment, reducing anticancer treatment intensity, or not providing anticancer treatment) [6], which was the main motivation for this study.

Participants and data collection
This study used part of the data from a large-scale prospective, longitudinal, and observational study (frailty assessment for high-risk patients with cancer undergoing curative intent antitumor treatment) for analysis [25]. This study administered a structured questionnaire to understand the frailty of patients with cancer before treatment and the changes in HRQOL before and after treatment. Purposive sampling was used to enroll patients from three hospitals in Taiwan who met the inclusion criteria: (1) patients with solid tumor or lymphoma aged � 65 years who were expected to receive chemotherapy, (2) patients who were willing to sign the informed consent form, and (3) patients who were conscious and able to complete the questionnaire interviews. Conversely, the following exclusion criteria were used: (1) patients with cancer with cognitive impairment who could not complete the questionnaire interviews, and (2) patients whose objective was not to cure cancer.
After the protocol was reviewed and approved by an institutional review board (no. 201600916B0), the research assistants explained the research purpose and methods to the patients. The patients completed the questionnaire after signing an informed consent form. The research assistants performed the first assessment within one week before the patients received chemotherapy (T 0 ). The assessment included a basic attribute information sheet, frailty assessment, and HRQOL scale. Four weeks after the patients received chemotherapy, another assessment (T 1 ) was conducted to collect data using the HRQOL scale.

CGA
The CGA was used to assess seven domains: functional status (activities of daily living/instrumental activities of daily living), nutrition, comorbidity, falls, polypharmacy, social support, and mood [19,21]. Since relevant societies do not recommend the use of consistent domains, following the suggestion of scholars that the CGA should assess at least five domains, this study chose the above-mentioned five domains with the highest proportion of CGA impairment. Further, this study deemed patients with impairments in more than two domains frail [16,17]. The assessment tools and cut-off standards for the five domains considered in this study are listed in Table 1.

HRQOL scale
This study assessed the HRQOL of elderly patients with cancer using the EORTC QLQ-ELD14 developed by the European Organization for Research and Treatment. The scale comprises 14 items asking patients about their HRQOL in the past week, each with a scoring range of 1-4 points, and is divided into five multi-item scales (three questions for mobility, two questions for worries about others, three questions for future worries, two questions for maintaining purpose, and two questions for illness burden) and two single items (joint stiffness and family support). The scale score was converted to 0-100 points based on the EORTC score conversion criteria. Higher scores indicated worse HRQOL (domains of mobility, worries about others, future worries, illness burden, and joint stiffness). Conversely, for the domains of maintaining purpose and family support, the higher the score, the better the HRQOL [10].

Statistical analysis
The data from the questionnaires were arranged, coded, and archived on a computer. Statistical analysis was performed using SPSS 22.0 (IBM, Armonk, NY, USA). Regarding descriptive statistics, age, as a basic attribute, was a continuous variable that was presented as a mean and standard deviation. Categorical variables are expressed as frequencies of use and percentages. To understand the unique effect of frailty on HRQOL (T 1 ) and its predictive value, an approach was taken to control for confounding variables, which included basic attributes exhibiting a significant relationship with the outcome variable in the univariate analysis as well as HRQOL (T 0 ). Subsequently, the effect of frailty on the different domains of HRQOL was analyzed using a hierarchical regression model. The two sets of variables were entered sequentially: (1) the basic attribute variables that had a significant effect on the specific domain of HRQOL in univariate analysis and HRQOL (T 0 ) and (2) the frailty variable.

Basic demographic attributes of elderly patients with cancer
This study included 178 elderly patients with cancer from the original database who met the inclusion criteria, with a mean age of 71.70 years (SD = 5.46 years, range = 65-96 years). Most patients were aged � 70 years (n = 99, 55.6%), female (n = 96, 53.9%), and married (n = 142, 79.8%). Regarding education, most patients were elementary school graduates (n = 117, 65.7%). Most participants were unemployed (n = 155, 87.1%). Regarding primary caregivers, half the patients (89, 50%) were cared for by their spouses, whereas the other half were cared for by non-spouses. Lymphoma was the most common cancer type among patients, accounting for 38.8% (69 cases) of the cases. Concerning cancer stage, 13 patients (7.3%) had stage I cancer, 64 patients (36.0%) had stage II cancer, 66 patients (37.1%) had stage III cancer, and 33 patients (18.5%) had stage VI cancer. All patients underwent chemotherapy, and 109 patients had solid tumors and received chemotherapy after surgery. Regarding the ECOG performance status, 102 patients (57.3%) had a score of 0. Most patients did not smoke (68.0%), drink (74.2%), or chew betel nuts (89.1%). Most patients (75.3%) had other chronic diseases ( Table 2). All patients received traditional chemotherapy via injection, and with the exception of one patient who received it via nasogastric tube, they could eat orally. Three patients underwent a colostomy.

Prevalence of frailty in elderly patients with cancer
This study used the CGA to assess frailty. The assessment included domains such as nutrition (MNA-SF), mood (GDS-4), functional status (ADL and IADL), comorbidity (CCI), polypharmacy, social support, and falls. The order of impairment domains was nutrition, mood, functional status, comorbidity, polypharmacy, social support, and falls (Table 3). A systematic literature review [16] found that most scholars used seven omains [26][27][28][29][30] or five domains [31-33] to perform CGA and failed to meet the criteria for two domains to define frailty. When seven domains were used for the CGA, 110 participants (61.8%) met the criteria for frailty. Contrastingly, when the five domains with the highest impairment rates were considered for assessment, 103 participants (57.9%) met the criteria for frailty ( Table 4).

The early predictive value of frailty assessment for patients' HRQOL
After controlling for the basic variables with a significant effect on HRQOL at baseline in step 1, the researcher included the presence or absence of frailty based on five domains of the CGA. In step 2, a regression analysis was performed on the various domains of HRQOL. The presence or absence of frailty was a significant predictor only for the illness burden domain of HRQOL.
The results indicated that 31.1-56.7% of the variance in the seven HRQOL domains could be explained by demographic variables and the presence or absence of frailty. Further, the presence or absence of frailty accounted for 1.4% of the variance in the illness burden domain (β = 9.5%, p < .05) for post-chemotherapy HRQOL in elderly patients with cancer (Table 5) and was a significant predictor for this particular domain.

Discussion
This study assessed frailty using five domains of the CGA with the highest impairment rate (nutrition, mood, functional status, comorbidity, and polypharmacy) and used impairment in more than two domains as the defining criteria for frailty. The results showed that 57.9% of the participants met the criteria for frailty in two domains, consistent with the results of Mohile et al. (with a prevalence rate of 60%) [32]. However, Wedding et al. also performed a CGA using five domains and used impairment in more than two domains to estimate frailty. Their study found a frailty prevalence rate of 50% [33], which is lower than that reported in this study. The researcher suggested that although the number of CGA domains and judgment criteria of the two studies were the same, this study followed the suggestion of the ASCO to [30] used seven domains of the CGA, but employed the impairment of more than four domains as a judgment criterion for frailty. The results showed that the prevalence of frailty was only 10% [30], which was significantly different from that in this study. Inconsistencies in the defining criteria for frailty affect its prevalence. Unfortunately, the SIOG and NCCN have not offered specific suggestions on the defining criteria of the CGA. This study only followed the suggestions of most scholars to use the impairment of more than two CGA domains to judge frailty [16,26,27]. The frailty assessment results using five and seven domains of the CGA were similar (57.9% vs. 61.8%) in this study, and the medical environment in Taiwan is characterized by a high level of busyness. Further, the CGA requires a considerable amount of time and resources to be administered by adequately trained assessors [22]. Therefore, this study suggests the adoption of the five domains of the CGA as a more practical and advantageous approach for frailty assessment by clinical medical and nursing personnel. If a large-scale, multicenter, or multinational study is to be conducted in the future, it will be necessary to reach a consensus on the number of CGA domains and define the criteria before a comparison can be performed.
Experts believe that the CGA should include multiple domains [19-21] but suggest that there is no need to assess all domains [23]. A systematic review [16] found that most scholars used seven [26-30] and five domains [31-33] of the CGA to assess frailty. This review summarized relevant studies on CGA using seven domains and found that the following five domains had the highest impairment rates: functional status [6,26,27,29], nutrition [6,26,29], mood [6,26,27], comorbidity [6,26,27], and polypharmacy [6]. These domains are consistent with the five domains with the highest impairment rates in this study (nutrition, mood, functional status, comorbidity, and polypharmacy), indicating that the important dimensions of frailty could be similar between Western and Eastern populations.
According to the results of the regression analysis, the presence or absence of frailty only explained 0.1-1.4% of the variance in various domains of post-chemotherapy HRQOL in elderly patients with cancer. There were no significant differences in most domains (p > .05), except for the illness burden domain (p < .05). The results showed that the presence or has been rarely used. Further, this study is the first to predict the HRQOL after chemotherapy (using the EORTC QLQ-ELD14) by evaluating frailty before chemotherapy in elderly patients with cancer. This makes it difficult to compare and discuss these findings with those of previous studies. It is possible that some variables not measured in this study, such as financial Step 1 Step 2 Step 1 Step 2 Step 1 Step 2 Step 1 Step 2 Step 1 Step 2 Step 1 Step 2 Step 1 Step 2 toxicity, treatment side effects, and spiritual well-being, could be related to HRQOL, especially in elderly patients with cancer in advanced stages. Moreover, the scoring method of the EORTC QLQ-ELD14 differs from that of most HRQOL scales, which use the total score to assess the HRQOL of patients with cancer, such as the FACT-G [37][38][39][40][41], Functional Living Index-Cancer [42], QOL-Cancer Survivors [43], QOL-AntiCancer Drugs [44], and Multidimensional QOL Scale-Cancer [45]. The EORTC QLQ-ELD14 scale used in this study does not use the total HRQOL score for assessment, but utilizes the scores of various subscales to perform respective analyses [10]. This approach could have overlooked the wholeness of HRQOL [46]. The HRQOL of elderly patients with cancer is unique [3], and the EORTC QLQ-ELD14 scale has been developed accordingly [3,10]. However, this scale has rarely been used. The FACT-G, developed by Cella et al., could be a better HRQOL assessment tool [38]. It not only has good reliability and validity [37,39,40], it has also been widely used by scholars in various countries around the world [37,40,41]. The Chinese-translated version and reliability and validity data are also available [40]. Therefore, it could be a better choice for measuring HRQOL in elderly patients with cancer and future studies should consider using this scale.

Conclusion
This study is the first to investigate the early predictive value of frailty for HRQOL in elderly patients with cancer in Taiwan. The findings showed that the presence or absence of frailty is an important predictor of domains such as illness burden in elderly patients with cancer. Frailty was a common issue among elderly patients with cancer. Thus, it is recommended that elderly patients with cancer undergo frailty assessment before treatment. These findings could serve as a reference for treatment-related decision-making (providing anticancer treatment, reducing anticancer treatment intensity, or not providing anticancer treatment) or the development of treatment plans.