Physician-perceived utility of the EORTC QLQ-GINET21 questionnaire in the treatment of patients with gastrointestinal neuroendocrine tumours: a multicentre, cross-sectional survey (QUALINETS)

This was an international, observational, non-interventional, cross-sectional, multicentre study conducted in the context of routine clinical practice. In the course of a routine visit, participating patients completed two HRQoL instruments, the QLQ-C30 and QLQ-GINET21, and the participating physicians completed an ad hoc survey designed to gauge their opinion regarding the clinical utility of the QLQ-GINET21.

A total of 36 oncologists and endocrinologists from 34 outpatient clinics in Spain and Portugal participated in the study (NCT02853422). The participating clinics were selected to obtain a representative sample of clinical practice with a wide geographic spread. Reference centres for the treatment of NET in Spain and Portugal were pre-selected and invited to participate (100% agreed to do so). Given that these questionnaires are routinely used in clinical practice and in clinical studies, no specific training in the administration of these instruments was provided.

This study was approved by the Clinical Research Ethics Committee of the Hospital Clínic (Barcelona, Spain). All patients provided written informed consent.

Inclusion criteria were: age > 18 years; diagnosis of GI-NET during the past 12 months; ability to complete the study questionnaires and report self-perceived health status. Exclusion criteria were: concomitant participation in another clinical study and presence of any other serious illness.

Patient-related sociodemographic variables were: age, sex, employment status, marital status, and educational level. Clinical variables obtained from medical records were as follows: date of GI-NET diagnosis; diagnostic procedure; time since first therapeutic intervention; current medical treatment for GI-NET; treatment modifications (if any) during the visit and reason for change; performance status (European Cooperative Oncology Group, ECOG); symptoms; and concomitant diseases and treatments. The patient’s current health status was assessed by the treating physician and rated as good, neither good nor poor, or poor, according to his/her clinical judgement. In addition, the physician was asked to indicate the patient’s current health status as stable, improving, or worsening in comparison to prior visits. This assessment was based on a general clinical evaluation combined with data, when available, from biochemical, immunohistochemical, and/or imaging studies.

Physician-related variables were age, sex, and years of experience in treating patients with NETs. Centre-related variables were: hospital location (rural, metropolitan, urban) and hospital size (number of beds).

The primary outcome measures were the three dimension-specific clinical utility scores obtained on the ad hoc survey, categorized as high, moderate, or low clinical utility.

Univariate and multivariate multilevel statistical models were performed to explore the potential association between the clinical utility scores obtained on the ad hoc survey and patient-, physician-, and centre-related study variables.

Descriptive statistics (means, standard deviations [SD], medians, and range) were used for continuous variables. Interquartile ranges were used where appropriate. The number and percentage of patients per response category was used to describe the categorical variables with 95% confidence intervals (CI). Parametric or non-parametric tests were used, as appropriate, to compare means and proportions. Pearson’s correlation (for continuous variables) and ANOVA (for categorical variables) were used to evaluate clinical utility.

Cronbach’s α was calculated for each dimension of the ad hoc survey to determine internal consistency. Values > 0.70 were considered acceptable. Statistical significance was set at p < 0.05. The SAS Statistical software program, v. 9.01 (SAS Institute; Cary, NC, USA) was used to perform all statistical calculations.

A total of 36 physicians from 34 institutions participated in the study. Sociodemographic data were available for 35 of the 36 physicians (97.2%). The mean (SD) age was 42.7 (7.7) years and 51.4% were female. The mean (SD) years of experience treating NETs was 12.7 (6.7). Nearly all (34/35; 97.1%) of the participating physicians worked in metropolitan/urban areas, most (60.0%) in large hospitals (> 500 beds). Slightly less than half (48.6%) of the physicians treated > 20 NET patients annually.

Of the 209 patients initially recruited, 10 were excluded for failure to meet the inclusion criteria (serious illness [n = 3], failure to complete the QLQ-GINET21 [n = 4], and inclusion after study deadline [n = 3]). Thus, 199 patients (Spain, n = 194; Portugal, n = 5) were included in the final analysis. Of these patients, 92.3% completed both study questionnaires. Table 3 describes the patients’ sociodemographic and clinical characteristics (n = 199).

Table 4 shows the mean (SD) scores with 95% CI for the three dimensions assessed on the ad hoc survey.

Figure 1 shows the percentage of all responses for each dimension categorised as high, moderate, or low utility. The highest-rated dimension was questionnaire characteristics while the lowest was therapeutic and clinical decision-making.

Figure 2 shows the percentage of responses that rated that item as having high (3–4 points), moderate (2 points), or low clinical utility (0–1 point). As that figure makes clear, the highest rated aspects were items 1, 2, 7, 11, 12, and 13 (high utility on > 85% of responses) while the lowest rated were items 3, 4, 5 (high utility in < 70%). See Table 2 for a description of the individual items.

All patient-related variables that were close to statistical significance (P < 0.2) on the univariate regression analyses were entered, together with the significant physician- and centre-related characteristics, in the multilevel models. The results obtained in the final model (model 3) are shown in Table 5 and the key findings for each of the three dimensions are summarized below in Table 6.

The dimension that received the lowest clinical utility ratings (considered “high utility” in only 65.8% of cases) was therapeutic and clinical decision-making. However, if we group the “high” and “moderate” utility scores, this percentage rises to approximately 95%, indicating an overall moderate-to-high support for the clinical utility of the QLQ-GINET21 for clinical decision-making. The relatively lower perceived utility of this dimension (Fig. 1) versus the other dimensions is not surprising given that this dimension comprises the most complex aspects of patient care, and patient-reported data are only one piece of the complex clinical picture that the clinician must consider. Moreover, the capacity of this instrument to help improve clinical decision-making is limited by the fact that the minimal clinical important differences (MCID) for this instrument have yet to be defined[23].

As Fig. 2 shows, the items with the greatest perceived clinical utility in this dimension were the items that help to provide a better understanding of the patient’s perspective regarding the impact of the disease (item 1), the aspects that most impact QoL (item 2), and current disease status (item 7). The relatively high clinical utility of these items suggest that physicians are interested in obtaining a greater understanding of the patient’s perspective. By contrast, the items with the lowest perceived utility (all of which received a score of 3 or 4 in < 70% of cases) were items 3 (treatment selection), 4 (changing dose or treatment regimen), and 5 (establishing frequency/number of follow-up visits).

The three items on this dimension all received similar utility ratings (around 80%), indicating that, in general, the physicians considered the questionnaire to be of value in improving communication. Item 8 on this dimension assesses the value of the questionnaire to improve doctor-patient communication in general. Item 9 refers to the value of the QLQ-GINET21 to explore overlooked aspects of the disease. Item 10 refers to the value of the questionnaire to help determine whether the patient has received enough information about the disease. Overall, the findings on this dimension suggest that the QLQ-GINET21 may help to improve doctor-patient communication. a finding that is in line with previous reports regarding the QLQ-C30 and similar instruments [18‐20].

This dimension, which broadly assesses the design and ease of use of the QLQ-GINET21, was the most highly rated, indicating that the physicians in this survey believe that the questionnaire is sufficiently brief and easy to understand, clearly-worded, and easy to complete. Moreover, these results are supported by the high completion rate (92.3%) of the QLQ-GINET21 in our cohort, a finding that is consistent with the 90% completion rates reported by Yadegarfar et al. in the validation study for this instrument [15]. Notably, the lowest-rated item (80.4%) in this dimension was item 14 “the characteristics of the questionnaire enable its routine use in clinical practice”, indicating that some physicians believe there are impediments to using the questionnaire in routine practice. In this regard, the QLQ-GINET21 is generally administered together with the QLQ-C30, thus increasing the time needed for completion. Indeed, a pilot study conducted by Van Leeuwaarde et al.[24] to assess patient (rather than physician) perceptions of the QLQ-GINET21 found that 13% of the respondents considered the questionnaire either too long or too tiring. This finding, considered together with our results, suggests that the instrument could potentially benefit from further refinement, such as removing the least valued aspects (as rated by physicians and/or patients) to reduce the time needed to complete the QLQ-GINET21. For example, in our survey, 7.5% of respondents indicated that item 10 (which asks whether patients have received enough information about the disease) had little clinical utility. Despite the proven utility of HRQoL questionnaires, their use in clinical practice is sometimes limited due to time limitations [25], which may be particularly relevant in patients with GI-NETs in whom fatigue is a common symptom.

A secondary, exploratory aim of this study was to identify patient-, physician-, or centre-related variables that might be associated with perceived clinical utility. On the multilevel model, high NET patient volume (> 30 patients/year) was associated with higher perceived utility on the therapeutic and clinical decision-making dimension, perhaps because physicians who treat more patients have greater experience in using and interpreting the results of the QLQ-GINET21. Conceivably, as physicians become more familiar with the instrument, this could increase their ability to use it more effectively and interpret it more quickly. Two other physician-related variables (older age and less experience) were associated with low clinical utility on two dimensions (doctor-patient communication and questionnaire characteristics). The association between older physician age and lower clinical utility is consistent with previous reports [26, 27]. However, in our study, this finding appears to contradict our other finding that less experience in treating NETs was also associated with low clinical utility. These contradictory findings are difficult to explain but may be related to the relatively small number of participating physicians (n = 36). The routine use of PROMs in clinical practice has increased substantially in the past two decades. Numerous studies have demonstrated the value of these instruments to inform care [28] and to improve communication [12, 13], symptom control, and HRQoL [12, 14]. Such instruments are particularly valuable in chronic diseases with a high symptom burden such as GI-NETs. However, the development and validation of these instruments is a complex, time-consuming process that requiring an ongoing evaluation process through field-use and studies such as ours. The QLQ-GINET21 is a relatively new instrument and it is imperative that we continue to refine it for the benefit of both patients and clinicians alike. Studies such as ours and that by Van Leeuwaarde et al. [29] can provide valuable data that can be used to improve these questionnaires. For instance, those authors found that while most (83%) of the patients who completed the survey considered that the instrument provided a moderate (30%) or good (53%) indication of their HRQoL, nearly one in five (18%) patients considered it a “poor” representation of their HRQoL. These findings, considered together with the results of our study, underscore the importance of assessing PROMs such as the QLQ-GINET21 from the perspective of both clinicians and patients.

This study has several limitations. First, the number of participating centres and clinicians was relatively small, making it difficult to draw statistically significant conclusions, especially regarding to the physician- and centre-related variables. In addition, this was a cross-sectional study with only a single HRQoL assessment per patient.

Another limitation is that the QLQ-GINET21 was designed for use at an aggregate level, rather than for individual patients. Since the MCID for this instrument have not been defined and validated[23], this could have impeded the clinician’s ability to accurately interpret the items, potentially leading the participants to overrate the true utility of certain items. By contrast, the main strength of this study is that it is the first to assess the perspective of physicians regarding the clinical utility of the QLQ-GINET21 in routine clinical practice.