A 15 Year Review (2006-2020) of Patient-Reported Outcomes (PRO) in United States Oncology Product Labeling and Trends in Sponsor Size and Oncology Experience

Background : Despite wide use of patient reported outcomes (PRO) tools in clinical development, resulting data is rarely incorporated into the US label. This study reviewed oncology product labels approved by the Food and Drug Administration (FDA) between 2006 and 2020 to determine if the rate of PRO inclusion in labeling has meaningfully changed. Sponsors were assessed to identify demographic trends in achieving PRO label success. Methods : FDA-approved drugs were searched utilizing the Drugs@FDA database by month from January 2006 to December 2020 for novel drug and biologic approvals. Labels and product summary basis of approval (SBA) were reviewed for inclusion of PRO data. Sponsor size and experience were determined for each product in the year of initial approval. Results : 155 oncology products received initial approval between 2006-2020, of which only 7 contained PRO data in the label. More than half (53.5%) of products had PRO data described in the SBA. Over time, PRO information increasingly been included in the product marketing application. Sponsors utilizing PRO data tend to be experienced in oncology development and larger in size. Conclusions : There has been no meaningful change in inclusion of PRO data in oncology product labeling over the past 15 years. Recent FDA guidance and initiatives may provide additional clarification to support appropriate PRO tools to support label inclusion as well as another forum for distributing PRO data publicly.


4
Questionnaire-Core 30 (EORTC QLQC30), EuroQol 5-dimensional questionnaire (EQ-5D), and Functional Assessment of Cancer Therapy: General (FACT-G) are the most commonly used general cancer measurement tools observed in oncology clinical trials. 6,7,[9][10][11][12] While the FDA does not specify which assessment tools to use, their 2009 Guidance for Industry for PRO outcome measures to support labeling claims does provide the characteristics they should contain. 3 Acceptability of PRO tools by health authorities varies. In fact, Gnanasakthy, et al. (2019) found that between 2012-2016, nearly half of oncology product labels received PRO label language when data obtained by these tools was included in the market approval submission package to the European Medicines Agency (EMA), while no label language was granted by the FDA for the same products. Specifically, the EQ-5D, EORTC QLC-C30 with disease specific module, and FACT-G with disease specific module, had the highest rates of inclusion in the EU label. 6 This difference in standards highlights the dynamic value-framework built to support each region's health authority, payer, and prescriber requirements. In regions with single or national payer systems, it is not enough to simply get market approval, rather an evaluation of the overall cost effectiveness of a product is conducted to set price ceilings. In contrast, in regions where there is a private payer market such as the US, the health authority focuses on the product's efficacy and safety while each independent payer separately evaluates overall value, and the level of evidence needed to influence payer decisions will vary. 13

Data interpretation
The FDA has stated that PRO instruments should be reliable, valid, and be able to detect change (i.e., sensitivity). 3 Supplementary, Rock, et al. (2007) note that interpretability of PRO data is reliant on three factors: 1) that the tool being used is capturing all relevant data; 2) that the results are consistent; and 3) that the results are clinically relevant to the population being studied even 5 if statistically significant. Achieving these requirements is challenging as oncology patient populations are complex even within the same indication. As a result of this complexity, general assessment tools may not be sufficient to support label claims. To help mitigate this, disease specific modules have been developed, for example, EORTC QLQ-C30 has roughly 60 disease specific questionnaires; however, not all such modules are validated. Additionally, most tools contain a multitude of measures or domains, and results may be positive in some domains, while negative or unchanged in others, challenging the robustness of findings. 14 Ultimately, the expectation is that the findings support the more conventional study endpoints (i.e., overall survival, overall response rate, duration of response, etc.). On the other hand, even if findings are statistically positive, if the changes are small, they may not be meaningful. One must also consider the potential for multiplicity when analyzing the data and ensure that the hypothesis being tested is well defined. Finally, studies which are not randomized or blinded have a high likelihood of confounding results due to patient bias. 3,7 To combat several of these issues, FDA recommends developing an end point model which can be discussed with the agency during developmental meetings. Such models tie specified primary and secondary endpoints, their outcome measure(s) and assessment tool(s), to a clearly defined label claim. 3 The challenge here, is knowing what label content is desired early in the development program, which is often difficult. 8

Missing data
The FDA's 2018 Guidance for Industry on oncology clinical trial endpoints clearly notes that "missing data and infrequent assessments can complicate the evaluation of symptom data particularly for time to deterioration analysis." In addition, to demonstrate improvement of symptoms, patients must be symptomatic at baseline. 15 Despite this, missing data, including 6 baseline data, is one of the primary reasons FDA rejects PRO data to support label claims. 6,7,9,10 Oncology patients tend to have significant health complications and complex treatment plans which involve a high number of office visits and the length of time to complete a HRQL is an increased burden on the patient. Collection of baseline data also means that study sponsors must also have the PRO instrument in place at the time of study initiation, which is not always the case, especially for products which get approval based on earlier phase studies. Missing data must then be censored which may result in false results or inability to interpret outcomes.
Methods to overcome missing data, such as missing completely at random (MCAR) and missing at random (MAR) analyses, as well as completing the HRQL by proxy present their own challenges and are not considered appropriate by FDA. 7 Notwithstanding the challenges collecting and analyzing PRO data can have, there are sponsors who do it properly and have achieved labeling claims in the US. The sponsors of the three products approved in 2011 with PRO label language, Zytiga, Jakafi, and Xalkori, include Janssen, Incyte Corporation, and Pfizer respectively. When it comes to new and costly ventures in oncology product development, many sponsors take a watch-and-wait approach, allowing other sponsors to work through the challenges and define a clear path to success. No previously published literature was found examining the sponsors who have successfully achieved PRO labeling, nor those who have failed.
Previously published evaluations of oncology PRO labeling included labels approved up to 2016. [4][5][6] The purpose of this study is to build upon the available published data by reviewing oncology product labels approved by the FDA between 2006 and 2020 to determine if the rate of PRO inclusion in oncology labeling has meaningfully changed over this 15-year period and 7 assess if the reasons for denial have remained consistent. In addition, product sponsors were appraised to identify the types of sponsors achieving PRO label success.

METHODS
Following methodology outlined in Gnansakthy et al., 2012 and 2016, FDA-approved drugs were searched utilizing the Drugs@FDA database by month from January 2006 to December 2020 for novel drug and biologic approvals, were searched for oncology products, and then oncology indications were verified by a second investigator. 4,5 Only products whose label "Initial U.S. Approval" date matched the NDA/BLA approval date within the database were included, eliminating older products which were noted as original submissions. Supplemental submissions and generic products were excluded.
Following identification, labels were searched for PRO inclusion, with particular focus on the Clinical Studies section of the package insert. Drug approval packages were reviewed for inclusion of PRO data application submitted to FDA for approval. Within the drug approval packages the Medical Review was examined when available, and in the absence of a Medical Review, the Multidisciplinary Review was utilized. In circumstances where PRO was included in the drug approval package but was not included in the product label, the FDA reviewer's reason(s) for rejecting the data were documented. The PRO assessment tool was recorded for all PRO data found in the drug approval packages and label. The type of FDA application review was noted for all products as follows: standard, priority, and/or accelerated. Finally, orphan drug status was also flagged.
Sponsor demographic data, including company size (number of employees) and the number of marketed oncology products the sponsor had at the time of product approval, was retrieved by searching the company product portfolios. The number of sponsor oncology products approved 8 before or within the year of each product within the dataset was recorded. When needed, such as when a sponsor no longer exists, Wayback Machine (wayback.archive.org) was used to access archived company websites. Sponsors were then categorized as follows: (1)

RESULTS
The final dataset consisted of 155 unique oncology products approved between 2006-2020.
Three products, Bavencio ® (avelumab), approved in 2017, and Tazverik ® (tazemetostat) and Gavreto ® (pralsetinib), both approved in 2020, received approval under two separate applications; however, these products were only counted as a single unique product in the database. Over the 15-year review span, 7 products (4.5%) had labels containing PRO data. One product, Zirabev ® (bevcizumab-bvzr) notes PRO as a secondary study outcome; however, no data or summary of results is provided. Therefore, Zirabev was not counted as having PRO in the label. Additionally, Xalkori (crizotinib) was noted by Gnanasakthy, et. al (2016) as having PRO in the label; however, the exact contribution of PRO data supporting label language, specifically visual safety events, could not be verified by this research group in the initial product label. The Visual Symptom Assessment Questionnaire is mentioned as being added to the pivotal Xalkori study during an amendment, but no further discussion on the tool is provided within the drug approval package, and it is therefore difficult to determine what, if any, data is directly linked to the initial label content. The addition of PRO language was found in the Xalkori product label under the header "Description of selected adverse drug reactions: Vision disorders" in 2013, two years following initial approval; however, as this was a modification to the label post-initial approval, this product was not counted as having PRO at the time of initial approval. Two labels containing PRO were approved in 2011, one in 2017, three in 2019, and one label in 2020. Figure 1 shows approval of PRO containing labels over time from 2006-2020. In addition, PRO was reviewed as part of the submission pack and discussed in the product approval summary for 87 products (53.5%). The majority of products with PRO in the approval summary were approved between 2016-2020 (n=53/87; 60.9%). Inclusion of PRO data in the approval summary by year as a portion of all yearly approvals is presented in Figure 2.
As described in Table 1, three of the seven products that were granted PRO labeling, Rituxan Hycela ® , Herceptin Hylecta™, and Phesgo™, utilized tools that measured preference for route of administration. The results are described under the "Patient Experience" subheading in the clinical studies section of each of the three labels. For these products, patients were asked to complete a preference questionnaire post-treatment to document their preference for subcutaneous or intravenous administration and their reasoning. Two of the seven products, Zytiga and Nubeqa™, used the Brief Pain Inventory-Short Form, which is designed to assess the severity of pain in chronically ill or cancer patients and the impact of pain on their daily functioning. PRO measurement tools specific to the disease indication were implemented for Jakafi and Inrebic®, both indicated for myelofibrosis, which used the Myelofibrosis Symptom Assessment Form version 2.0 to measure changes in the patient's Total Symptom Score. This tool allows patients to report on the six core symptoms of myelofibrosis (night sweats, itching, abdominal discomfort, early satiety, pain under ribs on left side, and bone or muscle pain). These data were included as additional clinical endpoints for the clinical studies for both products.
Experienced sponsors made up 25% (17/68) and moderately experienced sponsors comprised 19.1% (13/68) of the remaining products with no PRO information in the summary approval as shown in Figure 4. As a percentage of each experience type, those that had PRO information in the approval summary were primarily experienced sponsors (37/54; 68.5%), followed by moderately experienced (18/31; 58.1%), and little experience 32/70; 45.7%). primary objective or a primary endpoint. This is notable, as the most common reason for FDA denial of PRO inclusion in product labeling was due to the exploratory nature of these endpoints.

This analysis reveals the trends in PRO inclusion
The FDA analysis of PRO label inclusion for Phesgo stated that, "The results of the patient 13 preference questionnaire benefit from a reasonable study design, straightforward instrument, and high completion rate…". 16 Of the four remaining product labels containing PRO, two utilized the Brief Pain Inventory-Short Form and two utilized the Total Symptom Score for myelofibrosis (MFSAF). As the majority of PRO tools described in the approval packages included generic assessments such as the EORTC QLQ-C30, Functional Assessment of Cancer Therapy for multiple indications, and EQ-5D, it remains clear that the FDA still requires a definitive link between the PRO tool measurements and the indication studied. Sponsors appear to continue to struggle to find and appropriately implement the right tool for their specific tumor type within their development programs. These findings remain consistent with the Gnanasakthy, et. al.
(2016) study, which recognized similar limitations to PRO inclusion, such as open-label study design and lack of disease-specific tools. 5 Most products received priority approval (114/155; 73.5%), and about a third (50/155; 32.3%) received accelerated approval. It may therefore be reasonable to consider that complete PRO data may not be available at the time of submission of the marketing application, thereby rendering its inclusion into the product label unlikely. Upon review, this explanation is unfounded, as 4 of the 7 products with PRO in the label received priority review. Additionally, the percent of products with standard approval which did not contain PRO in the approval package was nearly twice that of products with standard approval which did contain PRO in the approval package. Accelerated approval did not appear to have any bearing on inclusion of PRO or not.
Sponsor size and experience in the oncology field were considered as possible influences on acceptance of PRO language in the product label, as one may expect that sponsors with greater resources and experience would have a larger number of products incorporating PRO adequately in each development program. Indeed, Rituxan Hycela, Phesgo, Herceptin Hycela were all sponsored by Genentech, a large biotechnology company with considerable experience with oncology products. Additionally, Nubequ and Zytiga, sponsored by Bayer and Janssen, reflect PRO product labeling by large companies. This is further supported by the findings that PRO inclusion in the summary package is led by more experienced sponsors, while those that did not contain PRO were headed by sponsors with little experience. While sponsor size did not vary meaningfully between PRO inclusion or exclusion in the summary package, sponsor disposition cannot be excluded as a factor in the appropriate use and acceptance of PRO data supporting the product label.
This study was limited to review and inclusion of PRO data for initial approval of novel oncology products and did not evaluate PRO inclusion in label supplements. This limitation is highlighted by the exclusion of Xalkori as an oncology label including PRO data, as the specific mention of the Visual Symptom Assessment Questionnaire (VSAQ-ALK) was not incorporated into Section 6 of the label until after initial approval. Additionally, due to the often expedited nature of oncology product development, collection of PRO following initial approval is not uncommon; therefore, future analysis should include both initial approvals as well as all clinical label supplements.

CONCLUSIONS
The collection, analysis, and interpretation of PRO in oncology is a key topic for regulators, sponsors, prescribers, as well as patients. In June 2021, the FDA released a draft guidance entitled "Core Patient-Reported Outcomes in Cancer Clinical Trials", which provides considerations for identifying the appropriate PRO tool to utilize, as well as when PRO assessments should be conducted and other protocol design considerations. The guidance removes some ambiguity around FDA's thinking on acceptable PRO tools specific to oncology 15 and notes that additional, more detailed Patient Focused Drug Development guidances are being drafted. 17 The ongoing desire to make PRO data publicly accessible is also apparent in the Oncology Center of Excellence's pilot project, "Project Patient Voice", initiated in 2020 as an online platform intended to list details of clinical trials and PRO data in a way that is easy for prescribers and patients to interpret. It is acknowledged by FDA that such data is currently limited in product labels, and cautions that such information should not be used alone in making treatment decisions. 18 As both the guidance and pilot project are both relatively new, it is unclear what the full impact they will have on inclusion of PRO in product labeling will be. Based on the increase, albeit slow, in inclusion of PRO assessments in oncology clinical programs observed over the 15-year period assessed in this study, it is likely that use of PRO tools may become even more commonplace given the alternative forum for providing the information to prescribers and patients. However, this alternative data sharing platform may hinder sponsor desire for inclusion of PRO data in the product label itself, resulting in a continued low number of oncology product labels containing PRO data.

Consent for publication
Not applicable.

Availability of data and materials
All data generated or analyzed during this study are available at Drugs@FDA.com, Google.com, and Wayback.archive.org.   Figure 1 Oncology labels containing patient report outcome data. The total number of oncology labels containing patient report outcome (PRO) data receiving initial market approval are depicted by year from 2006-2020.

Figure 2
Approval summaries containing patient report outcome data. The total number of summary basis of approvals for oncology products containing patient report outcome (PRO) data receiving initial market approval are depicted by year from 2006-2020.

Figure 3
Inclusion of patient report outcome in summary approval package by sponsor size. Product summary basis of approvals for approved oncology products between 2006-2020 are categorized sponsor size and whether they contain patient report outcome (PRO) data or not. Sponsor size was determined by number of employees as follows: small (1-499 employees), medium (500-9,999 employees), and large (10,000 or greater employees).

Figure 4
Inclusion of patient report outcome (PRO) in approval summary by sponsor experience. Product summary basis of approvals for approved oncology products between 2006-2020 are categorized sponsor experience and whether they contain patient report outcome (PRO) data or not. Sponsor experience was determined by number of oncology products approved each year as follows: little experience (0-2 products at time of approval), moderate experience (3-5 products at time of approval), and experienced (6 or more products at time of approval).