Quality of life in oncology drug approvals: evidence from Swiss decisions in relation to global regulatory trends

Over the past two decades, patient-centred care has increasingly shaped both clinical practice [1, 2] and clinical trial design [1, 3] in oncology. As part of it, health-related Quality of Life (HRQoL) has emerged as an important measure of treatment benefit, complementing traditional survival outcomes such as overall survival (OS) and progression-free survival (PFS). The World Health Organization defines QoL as ”an individual’s perception of their position in life in the context of their culture, value systems, personal goals, expectations, standards, and concerns” [4], encompassing physical, psychological, social, and environmental dimensions of well-being. Methodologically, HRQoL is typically assessed through patient-reported outcomes (PROs), which capture patient self-evaluations of symptoms and functioning. They are often complemented by clinician-reported, observer-reported, as well as technology-reported outcomes that leverage wearables and digital health technologies for passive data collection [3, 5, 6]. Together, these approaches offer a multidimensional and patient-centred understanding of the impact of treatment.

While oncology trials have traditionally prioritized objective endpoints such as OS and PFS, regulatory agencies increasingly emphasize the inclusion of HRQoL and other patient centred outcomes in trial design to support benefit-risk evaluations [3, 7]. Incorporating these outcomes on HRQoL not only gives patients a voice, but also enables more informed, treatment-specific decisions, particularly when such data are reflected in product labels. This demand is reflected by the increasing number of guidance documents published by the FDA [8], EMA [9], and Swissmedic, the Swiss Agency for Therapeutic products [10]. However, regulatory submission trends indicated the opposite. Between 2012 and 2016, PROs were included in 70.3% of oncology submissions to the EMA and FDA, yet only 46.7% were reflected in the corresponding Summary of Product Characteristics of the EMA (SPCs, a regulatory document providing essential information for the safe and effective use of medicinal products [11]). During the same period, the FDA did not include any PRO-related statements in oncology labels [12]. More recent data show modest improvements: between 2017 and 2020, the proportion of oncology trials reporting HRQoL endpoints rose to 67.8% [13]. From 2018 to 2021, although PRO data were present in 66% of FDA oncology submissions, nearly half were excluded from labelling due to methodological concerns such as flawed trial design or inadequate measurement instruments [14, 15]. Similarly, EMA submissions from 2017 to 2021 included PRO data in 74.3% of cases, yet only 20% of SPCs referenced them [16]. Broader evaluations of European Public Assessment Reports (EPARs) confirm this disconnect: less than half of approved medicines mentioned PRO data at all [17]. Despite this growing emphasis, the inclusion of HRQoL and PRO measures in regulatory submissions is not currently mandated by any of these agencies.

Common methodological hurdles in incorporating HRQoL data into oncology trials include the reliance on significant missing data, inadequate multiplicity control, and inconsistent analytical standards [12, 15]. Heterogeneous methodologies, fragmented endpoint selection, and inconsistent reporting further limit the impact of HRQoL data on regulatory decision-making [18]. Gupta et al. noted that, despite half of cancer trials report HRQoL results, these data were frequently delayed or poorly integrated into primary publications, hindering timely interpretation [19]. Additionally, the lack of standardization in analysing and interpreting HRQoL data has long been a critical barrier. In response to these challenges, several methodological and reporting initiatives have emerged to improve the interpretability, robustness, and regulatory relevance of PROs. Notably, the SISAQOL-IMI consortium provides standardized recommendations for the analysis and interpretation of PRO endpoints in oncology trials [20]. Complementary guidance such as CONSORT-PRO [21] and SPIRIT-PRO [22] aim to enhance transparency and consistency in the reporting and protocol specification of PROs. Collectively, these initiatives reflect a shift in the field from problem identification toward structured solutions, although their uptake and impact on regulatory submissions remain variable.

Given this context, the present study aims to: (I) quantify the frequency and analyse the HRQoL domains included in Swissmedic oncology submissions from 2001 to 2020, and (II) identify gaps in HRQoL measurement practices and instruments utilized. Addressing these objectives directly contributes to ongoing discussions on patient-centred drug development and regulatory approval, advocating for better integration of patient perspectives in oncology care.

We conducted a retrospective study of all Swissmedic approvals for oncology drugs over two decades (2001–2020), including applications for new active substances (NAS) and indication extensions (IE). All applications approved for oncology indications during this period (both solid tumours and haematology) were included. Applications were excluded if they were withdrawn, rejected, indicated for non-oncology conditions, or lacked the required documentation.

We developed a systematic method to extract and classify HRQoL data from Swissmedic assessment reports (Fig. 1). First, all relevant oncology assessment reports were manually retrieved from Swissmedics internal document storage and filtered by application type (NAS/EI), oncology indication, submission and approval dates, and approval status (step 1 in Fig. 1). Due to the manual nature of the document retrieval process, only oncology assessment reports up to the year 2020 were included in the analysis, as these had been extracted and made available before the start of this study. While an application refers to a submission by a sponsor, an assessment report is an internal Swissmedic document containing the results of the application review. Relevant metadata, such as the creation date, were automatically extracted from the files. For each application identifier, the drug name, submission and approval year, and active substance were then extracted from the assessment reports (step 2). For further analysis, the indication was then classified as either solid tumour or haematology for further analysis (step 3). As a single application can cover multiple oncology indications the number of indications (N = 362) is higher compared to the number of applications (N = 342). Subsequently, the latest SPCs were collected as published on Swissmedicinfo, an online platform providing detailed information on medicines authorized in Switzerland, and matched with the application drug name (Step 4). Similarly, as in step 3, the same SPC is often linked to multiple assessment reports (i.e., one drug can have several oncology indications). Consequently, the total number of SPCs is lower than the total number of assessment reports, since a single SPC may cover multiple indications (one drug = one SPC = several indications).

Afbeelding vergroten

The next steps involved identifying HRQoL concepts of interest. This included general concepts such as Quality of Life and patient perspective that were used to assess whether HRQoL terminology was used. Any application containing at least one concept was marked as including HRQoL terminology. Furthermore, HRQoL domains such as psychological health, with associated keywords (e.g., mental health) were defined following the QoL-framework proposed by the WHO [23]. Furthermore, specific HRQoL measurement instruments, such as the EORTC QLQ-C30, used to represent the patient-perspective were investigated. Lastly, digital health technologies were defined that could have been used to collect HRQoL data beyond questionaries including for example electronic PROs and wearables (step 5). As most of the assessment reports and all SPCs are available only in German, we have included the concept in both languages, German and English. To avoid introducing bias by including two languages, we ran our search twice: once on the German list and once on the English list. We then reconciled the results at the concept level. Specifically, we created a small translation lexicon that mapped each German HRQoL keyword to its English equivalent. We then merged any duplicate hits so that for instance ”Lebensqualität” and ”quality-of-life” were counted as one conceptual instance. Only then did we compute our frequency statistics. To determine whether a given application contained reference towards QoL, we applied a binary classification approach based on the keywords defined in the General HRQoL group. If there was at least one keyword present, the report was classified as “includes HRQoL concepts”. The list of terms and their corresponding bilingual mapping can be found in Appendix B.

To ensure comprehensive identification, approximate string-matching techniques were employed (step 6). These methods accounted for variations in terminology, typographical errors, and alternative spellings, increasing the accuracy of term detection within the reports [24]. To quantify this improvement, we conducted a case study on the term “patient-reported outcomes” in Kluetz et al. [3]. Manual review of the article identified 25 instances of the term. An exact-match search recovered 23 of these occurrences (92% recall), whereas the fuzzy-matching approach implemented successfully retrieved all 25 instances (100% recall) by accounting for both hyphenated (”patient-reported outcomes”) and non-hyphenated (”patient reported outcomes”) spellings. Given the presence of images, tables, and print screens in some of the assessment reports, optical character recognition was employed, allowing text to be captured from images and Table [25]. The implementation of an automated term matching process within the reference files is intended to guarantee reproducibility and facilitate efficient repetition of new concepts that were introduced at a later stage during the project. However, we recognize potential limitations, such as the risk of missing nuanced or non-standard HRQoL concepts, and challenges in capturing the subjective, multidimensional nature of HRQoL constructs. To address these issues, we integrate multiple expert review cycles and iteratively refine our scripts and keyword lists used, ensuring that the methodology remains robust and comprehensive (further described in Appendix A and the list of keywords in Appendix B).

Finally, in-depth reviews of the results and review of the assessment reports were performed to provide information on the inclusion and impact of HRQoL in the regulatory decision-making process (step 8). All analyses were conducted using Python 3.12, and the code is available upon request.

Our findings reveal a significant increase in both new active substance and expanded indication applications over the two-decade period (2000–2020). The higher overall numbers in IE applications reflect a strategic shift towards expanding the use of existing therapies, particularly for solid tumours, by seeking additional indications alongside the development of new agents. Despite this increase and the emphasis on patient-centred care by regulatory agencies worldwide [8‐10], the integration of HRQoL data in regulatory submissions remains fragmented: The average HRQoL inclusion rate across the observation period was 63.16% (69.35% in NAS applications and 59.63% in IE applications). This aligns well with previous studies, for example, Sherry et al. [32] found that HRQoL appears both under evaluated and underreported in phase 3 oncology RCTs; similarly, Gupta et al. [19] found that HRQoL outcomes were underreported in FDA approvals for oncology drugs, with significant delays and inconsistencies in reporting. Furthermore, Gupta et al. reported that only 50% of the analysed trials associated with FDA approvals between 2015 and 2020 included HRQoL data while only 42% of it was available by the time of FDA approval. Additionally, Paratore et al. show that not only is the frequency of reporting low, but only in 38.5% (FDA) and 39.4% (EMA), HRQoL data demonstrate clinically significant improvements beyond minimal clinically important differences [33], while in our analysis only 28.70% of the submissions indicate a HRQoL benefit (see Fig. 4B).

Afbeelding vergroten

The analysis of HRQoL concepts reveals an ongoing imbalance in the representation of patient experiences in regulatory considerations. Although aspects related to physical health, such as fatigue or pain, are frequently reported, others, such as psychosocial health and social support, are rarely addressed. This emphasis on physical symptoms and functional outcomes reflects a narrow focus on the physical domain, often at the expense of a broader, multidimensional understanding of patient well-being. This is further accentuated when comparing the assessment reports with the public SPCs: Of all the HRQoL dimensions, physical health was the only domain mentioned more frequently in the SPCs than in the assessment reports. This may be because physical health is more closely aligned with clinical endpoints and adverse event profiles, making it easier to measure and more relevant to regulatory communication and thus is more often the regulatory focus of the PRO data submitted, as suggested by Kluetz et al. [3]: “The FDA [.] has suggested focusing its analysis of submitted PRO data on the core concepts of disease symptoms, symptomatic adverse events, and physical function”. Interestingly, two SPCs (nivolumab and carfilzomib) include patient-reported physical health-related outcomes in both the EMA and Swissmedic versions, but not in the FDA version (see Appendix E). Taken together, these findings suggest that HRQoL information is progressively filtered as it moves from internal assessment reports to public product information. Although assessment reports capture a broader range of patient-reported outcomes, only a limited subset is retained in SPCs. This reflects regulatory relevance, evidentiary standards and communication constraints, rather than a disregard for HRQoL data.

A contributing factor to the inconsistent inclusion of HRQoL data in regulatory submissions is the lack of standardized methods and validated instruments across diverse patient populations and disease contexts [7]. To address these challenges, several international initiatives have emerged: SISAQOL for harmonizing HRQoL analysis standards [20], PROTEUS for improving the quality and usability of PRO data [34], PROMIS for standardized patient-centred outcome measures [30], and COMET for developing core outcome sets [35]. Although not yet universally adopted, the promotion of such frameworks could bridge the disconnect between the growing emphasis on HRQoL and its inconsistent representation in regulatory submissions, leading to more robust patient-centred outcomes.

As outlined in the results, digital methodologies were notably absent. Although the early cut-off date (end of 2020) may partially explain this, it also points toward challenges in including novel measurements due to insufficient guidance on their implementation and validation. Recent regulatory-supported [5, 36] and nonprofit-led projects [37] have initiated feasibility assessments of these digital methodologies, yet their validation, and thus regulatory acceptance, remains pending.

This study has several limitations. First, our analysis was limited to Swissmedic’s internal documents until the end of 2020. As a result, our study does not reflect any developments or changes in regulatory practice that occurred thereafter. Furthermore, our methodology focused on the presence or absence of HRQoL concepts in the regulatory documents, rather than assessing the quality or depth of the HRQoL data provided. This binary approach does not capture the richness of the information or how it may have influenced regulatory decisions. Variations in terminology, the potential use of synonyms as well as using our own, non-validated translation lexicon could also have led to underreporting of HRQoL mentions, introducing bias into our analysis. This approach also assumes that sponsors have included all HRQoL data collected during clinical trials in the submitted evidence, which may not be the case for exploratory endpoints, which are often known to be HRQoL/PRO [17]. Despite these limitations, the study provides valuable information on the integration of HRQoL data in regulatory submissions and highlights the need for more consistent and comprehensive inclusion of patient-centred outcomes to improve drug evaluation processes.

Our study examined the integration of health-related Quality of Life data in Swiss oncology submissions from 2001 to 2020, revealing persistent gaps in the representation of patients lived experiences. Although physical symptoms such as fatigue and pain were frequently discussed, the psychological, social and environmental dimensions were given far less consideration. To move beyond supplementary use, pharmaceutical companies must incorporate fit-for-purpose, validated HRQoL measurement instruments into the design of studies early, aligning endpoints with patient values and ensuring statistical readiness for hypothesis testing. To support this evolution, regulatory agencies should even further strengthen their guidance on the systematic inclusion of robust HRQoL data, formulating clearer expectations and harmonized standards. Aligning international efforts will be key to ensure that future therapies better reflect the full range of patient needs and measure what matters to patients.