Quality-of-life measures in pharmacogenomic studies: a systematic review

Pharmacogenomics is a rapidly evolving field that combines pharmacology and genomics to understand how an individual’s genetic makeup influences their response to medications [1]. By tailoring drug selection and dosage based on a patient's genetic profile, pharmacogenomics holds the potential to optimise treatment efficacy and minimise adverse drug reactions [2]. If pharmacogenomics can optimise an individual's health, it will positively affect their quality of life (QoL). Therefore, integrating quality-of-life measures in pharmacogenomic studies is crucial, as it ensures personalised treatment that enhances overall well-being. The World Health Organisation defines QoL as an individual’s perceived position within the context of their culture value systems, personal ambitions, expectations, standards, and concerns [3]. A variety of QoL instruments have been used across health-economic research [4] yet their application in pharmacogenomics remains poorly characterised. For example, Arnold et al. [4] compared indirect preference-based instruments (e.g., European quality of life five dimensions (EQ-5D), short form six dimensions (SF-6D), and health utilities index (HUI)) with direct elicitation techniques (standard gamble, time trade-off) and found that the former systematically produce lower utility scores. In the pharmacogenomic context, O’Shea et al. [5] evaluated genotype-guided therapy in patients with multimorbidity and polypharmacy, demonstrating reductions in hospitalisations, costs, and adverse drug events. Although they list QoL among their outcomes, they neither report specific QoL scores nor name the instruments used. Consequently, it remains unclear which QoL tools are most commonly applied in pharmacogenomic studies and how their results are presented. We conducted a systematic review by analysing pharmacogenomic studies that assessed QoL using data from selected databases. Our study aimed to identify the QoL tools currently in use, and to determine whether these studies employ generic tools, disease-specific tools, or a combination of both. Additionally, we investigated whether the studies provided justification for their choice of QoL tools and how they reported their scoring system (i.e. domain-specific vs. overall scores) which is central for ensuring the appropriateness, validity, reliability and comparability of the findings.

The Preferred Reporting Items for Systematic reviews and Meta-Analyses search extension (PRISMA-S) guidelines was considered for this systematic review [6]. The PRISMA-S checklist is provided in Appendix 1.

We conducted a systematic review and used a search strategy through Cochrane, Embase, Ovid and PubMed. The literature search was conducted from inception until April 30, 2024. We assigned two members to search through Cochrane, and one member for each of the other databases. We entered the same search terms, “Pharmacogenetics” AND (“quality of life” OR “PROMS” OR “PREMS” OR “health related quality of life” OR “'EuroQol” OR “utility” OR “generic” OR “time trade off” OR “standard gamble” OR “SF-6D” OR “EQ-5D”) into the databases. However, on PubMed, using the term “utility” produced too many results, so we entered “health utility” instead to narrow down the results. The keywords were selected by considering the titles and abstracts of full publications relevant to our study, areas our study covers, and names of QoL tools. The inclusion of terms like 'time trade off' and ‘standard gamble’ was intentional, as these are direct utility-elicitation methods used to measure the health-state utility, which are essential for calculating Quality-Adjusted Life Years (QALYs) and are often reported alongside QoL data. This strategy ensured a comprehensive capture of studies while focusing on those that report QoL measures, which are important for understanding the impact of pharmacogenomic interventions on patient QoL.

Our inclusion criteria consisted of randomised clinical trials, cohort studies and cross-sectional studies. Both generic and disease-specific QoL measurements related to pharmacogenomics were included. Our exclusion criteria consisted of non-English and non-human studies, abstracts in the publications that mentioned QoL with no relation to pharmacogenomics, mention of pharmacogenomics but no specification to the QoL tool used, and studies without full text. Studies that mentioned the EQ-5D questionnaires without the specific version of the tool, or studies that utilise EuroQol VAS scale alone were excluded as well. Detailed reasons for the exclusion of studies are provided in Appendix 2. After obtaining our initial search results, each team member performed title and abstract screening on their assigned database. During this process, we evaluated every record against our predefined inclusion and exclusion criteria, documenting reasons for including or excluding each article. We also identified and removed duplicates across databases. For each database, we recorded the total number of hits retrieved, the number of duplicates removed, and the counts of articles included and excluded.

Upon completing title and abstract screening, we compiled the studies deemed eligible into a master Excel spreadsheet to facilitate full‐text review. ach article underwent two independent screenings: first by the reviewer who performed its initial title/abstract assessment, and subsequently by a second team member to verify data accuracy. We extracted the following information for each study: the type and name of the QoL instrument, the authors’ rationale for tool selection, whether the measure was generic or disease-specific, and whether domain-specific or overall utility scores were reported. We also recorded which team member screened each article (and when), along with study authorship, publication year, country, language, clinical condition, participant demographics (age and recruitment source), study design, a brief description of the intervention or exposure, and funding source. We applied our predefined inclusion and exclusion criteria at this stage, recording reasons for exclusion where applicable Once each article had been screened twice, we obtained our final set of included papers and confirmed with the third reviewers (ZA and DA). The completed PRISMA flow diagram was used to summarise our search strategy and screening across all the databases (Fig. 1). Although we initially planned to include risk of bias assessment in the PROSPERO protocol, we did not pursue it, given the exploratory aim of this review, to identify and characterise the QoL instruments employed in pharmacogenomic studies rather than to evaluate intervention effectiveness. This study has been registered in PROSPERO with the following number CRD42024542177.

Afbeelding vergroten

This systematic review aimed to investigate the types of QoL measures used in pharmacogenomic studies. We found that a variety of questionnaires, particularly disease-specific measures, were utilised, which have not been extensively explored in previous research. However, this diversity also reflects a lack of consistency in QoL assessment and specifically reporting practices (i.e. domain-specific vs. overall scores) across studies.

A significant finding from this systematic review was that most of the studies reported domain-specific outcomes rather than overall QoL utility scores [9‐11, 18]. This is particularly concerning, as domain scores cannot be directly used in health economic evaluations. These analyses rely on utility values (i.e. overall scores) for the prioritisation of healthcare resources. The absence of overall utility scores limits the ability to compare interventions or prioritise treatments based on their impact on patients’ QoL.

Moreover, most studies did not fully report how their QoL instruments performed in terms of reliability and validity. Only Isaza et al. [8] provided both reliability and validity evidence. Three others (Hajj et al. [13]; Hajj et al. [17]; Agulló et al. [9]) reported validity alone, and only Martin et al. [11] noted the minimally clinically important difference. This lack of psychometric detail limits our ability to interpret and generalise QoL findings. Consistent reporting of reliability, validity, and responsiveness is therefore essential for robust QoL assessment in pharmacogenomic research.

The most common generic QoL measure used in the studies was the SF-36, which was employed by Bohlen et al. [12]., Martin et al. [11]., and Schricker et al. [15]. This is a widely used 36-item instrument that covers eight domains to assess overall health status, including physical, mental, and social functioning [19]. The SF-36 is applicable across a wide range of diseases and conditions, making it suitable for use in various patient groups. Another generic QoL measure that was utilised was SF-12 in the study by Agulló et al. [9], a condensed version of the SF-36, designed to provide a concise measure of health-related QoL [20]. The World Health Organisation developed WHOQOL-BREF [3], another generic QoL tool, which was used in the study conducted by Isaza et al. [8]. It is designed to provide a comprehensive assessment of an individual's QoL across 4 domains: physical health, psychological health, social relationships, and environment. The Q-LES-Q-SF, a generic QoL measure used by Brennan et al. [16]., is a short form of the original Q-LES-Q, consists of 16 items that cover various aspects of life focusing on enjoyment and satisfaction, derived from the original 93-item form [21].

For pharmacogenomic studies in cancer patients, the EORTC-QLQ-C30 questionnaire was frequently used. This is a generic, 30-item instrument that measures QoL in all cancer patients across key health domains such as physical functioning, emotional functioning, fatigue and pain [21]. Both studies, Riccetti et al. [18] and Hajj et al. [17] used this tool in their respective pharmacogenomic studies.

Only one of the studies employed a disease-specific measurement. In this study, Lebbé et al. [10] used the FACT-Melanoma tool to assess the QoL in patients with melanoma. FACT-Melanoma is one of the two validated QoL tools for use in melanoma patients, and it incorporates four main domains: physical well-being, social/family well-being, emotional well-being, and functional well-being [22]. The melanoma-specific subscale focuses on issues and symptoms particularly relevant to melanoma patients, such as skin symptoms, side effects of treatment, and other melanoma-specific QoL concerns.

Some pharmacogenomic studies used both generic and disease-specific QoL measures. For instance, studies conducted by Oberkampf et al. [14] and Ciruelos et al. [7] used EORTC QLQ-C30 as a generic QoL tool for cancer. For the disease-specific measure, Oberkampf et al. [14] and Ciruelos et al. [7] used QLQ-BN20 and QLQ‐CIPN20, respectively. The QLQ-BN20 was designed to measure the QoL in patients with brain tumours, including the specific symptoms and concerns associated with this condition [23]. The QLQ-CIPN20 was developed to assess the impact of chemotherapy-induced peripheral neuropathy on the QoL of cancer patients. The questionnaire consists of 20 items divided into three subscales that encompass specific sensory, motor, and autonomic symptoms associated with chemotherapy-induced peripheral neuropathy [24].

While most studies included in this review did not provide explicit justifications for their choice of QoL instruments, Isaza et al. [8] highlighted the reliability of the WHOQOL-BREF in assessing QoL among heroin-dependent patients receiving methadone. However, even in this case, the justification was relatively broad and lacked specificity regarding context and population relevance. Similarly, Hajj et al. [13] acknowledged the EORTC-QLQ-C30 as a widely recognised and validated assessment tool, their justification for its use in assessing QoL in palliative cancer patients receiving morphine was insufficient. They did not provide a clear rationale for why this generic QoL measure was the most suitable choice, especially considering the availability of disease-specific modules that might have been more appropriate for this patient population. Furthermore, the study did not discuss the sensitivity of EORTC-QLQ-C30 to changes in QoL within this context or compare it to other potentially more suitable measures. In their discussion, the authors themselves recognised that disease-specific modules could be more fitting, highlighting a gap in their rationale for selecting the EORTC-QLQ-C30 for their study. The absence of well-articulated rationales across studies limits the ability to evaluate whether the selected instruments were appropriate for capturing outcomes meaningful to patients. Clear justification for QoL instrument selection is essential for advancing methodological rigour and ensuring that tools are suitable for the intended population and research objectives.

Pharmacogenomic studies employed diverse QoL instruments, hindering consistent and reliable reporting. Future studies should justify QoL tool selection and report both overall and domain-specific outcomes consistently to enable valid comparisons and inform decision-making.