Characteristics of Clinical Studies Conducted Over the Total Product Life Cycle of High-Risk Therapeutic Medical Devices Receiving FDA Premarket Approval in 2010 and 2011

Importance  The US Food and Drug Administration (FDA) approves high-risk medical devices, those that support or sustain human life or present potential unreasonable risk to patients, via the Premarket Approval (PMA) pathway. The generation of clinical evidence to understand device safety and effectiveness is shifting from predominantly premarket to continual study throughout the total product life cycle.

Objective  To characterize the clinical evidence generated for high-risk therapeutic devices over the total product life cycle.

Design and Setting  All clinical studies of high-risk therapeutic devices receiving initial market approval via the PMA pathway in 2010 and 2011 identified through ClinicalTrials.gov and publicly available FDA documents as of October 2014.

Main Outcomes and Measures  Studies were characterized by type (pivotal, studies that served as the basis of FDA approval; FDA-required postapproval studies [PAS]; or manufacturer/investigator-initiated); premarket or postmarket; status (completed, ongoing, or terminated/unknown); and design features, including enrollment, comparator, and longest duration of primary effectiveness end point follow-up.

Results  In 2010 and 2011, 28 high-risk therapeutic devices received initial marketing approval via the PMA pathway. We identified 286 clinical studies of these devices: 82 (28.7%) premarket and 204 (71.3%) postmarket, among which there were 52 (18.2%) nonpivotal premarket studies, 30 (10.5%) pivotal premarket studies, 33 (11.5%) FDA-required PAS, and 171 (59.8%) manufacturer/investigator-initiated postmarket studies. Six of 33 (18.2%) PAS and 20 of 171 (11.7%) manufacturer/investigator-initiated postmarket studies were reported as completed. No postmarket studies were identified for 5 (17.9%) devices; 3 or fewer were identified for 13 (46.4%) devices overall. Median enrollment was 65 patients (interquartile range [IQR], 25-111), 241 patients (IQR, 147-415), 222 patients (IQR, 119-640), and 250 patients (IQR, 60-800) for nonpivotal premarket, pivotal, FDA-required PAS, and manufacturer/investigator-initiated postmarket studies, respectively. Approximately half of all studies used no comparator (pivotal: 13/30 [43.3%]; completed postmarket: 16/26 [61.5%]; ongoing postmarket: 70/153 [45.8%]). Median duration of primary effectiveness end point follow-up was 3.0 months (IQR, 3.0-12.0), 9.0 months (IQR, 0.3-12.0), and 12.0 months (IQR, 7.0-24.0) for pivotal, completed postmarket, and ongoing postmarket studies, respectively.

Conclusions and Relevance  Among high-risk therapeutic devices approved via the FDA PMA pathway, total product life cycle evidence generation varied in both the number and quality of premarket and postmarket studies, with approximately 13% of initiated postmarket studies completed between 3 and 5 years after FDA approval.

In the United States, the Food and Drug Administration (FDA) predominantly grants marketing approval through the Premarket Approval (PMA) pathway for novel high-risk medical devices, which are defined as those that support or sustain human life, prevent illness, or present potential, unreasonable risk to patients.¹ The PMA pathway requires premarket clinical evidence providing reasonable assurance of device safety and effectiveness and permits supplemental applications whenever postapproval changes are made to the device.^1,2 Recently, concerns have been raised that the clinical studies supporting FDA approval of high-risk devices lack adequate rigor and are prone to bias.^3-5

To accompany these premarket assessments, the FDA has become increasingly committed to evaluating device safety and effectiveness throughout the “total product life cycle,”⁶ an approach that involves ongoing study and reevaluation for as long as devices remain in use.⁷ As one part of this approach, the FDA can require additional clinical studies as a condition of PMA approval through its postapproval studies (PAS) program.⁸ However, PAS may often be small,⁹ delayed,^9,10 or not generalizable^5,11; in addition, only one-quarter of PAS required by the FDA between 2005 and 2011 were completed.⁹ Beyond FDA-required postmarket studies, complementary sources of evidence may be generated through studies initiated by manufacturers or independent investigators, although manufacturers may conduct postmarket clinical studies primarily to comply with regulatory requirements.¹²

As the FDA adopts more flexible premarket evidence requirements for devices in an effort to expedite patient access to new technologies,^13,14 the information generated from postmarket studies will become increasingly important in guiding regulatory and clinical decisions. Our objective was to characterize the clinical studies of high-risk therapeutic devices initially approved via the FDA PMA pathway between 2010 and 2011 to better understand the amount and quality of evidence generated over the total product life cycle.

We constructed a sample of high-risk therapeutic devices initially receiving US marketing approval via the FDA PMA pathway between January 1, 2010, and December 31, 2011, using the publicly accessible PMA database (eFigure 1 in the Supplement).¹⁵ We selected this sample period to ensure that the majority of relevant trials were registered on ClinicalTrials.gov—an online public clinical trials registry maintained by the National Library of Medicine—in compliance with the 2007 FDA Amendments Act.¹⁶ We used information on device type listed within the FDA database to exclude all nontherapeutic (ie, diagnostic) devices,¹⁷ including detection kits, molecular assays, and imaging machines. Based on information within the publicly available FDA Summary of Safety and Effectiveness Data (hereafter referred to as the FDA Summary) linked to each original PMA application,¹⁸ we further excluded therapeutic devices that were previously marketed in the United States for another indication.

Using information within the PMA database, we classified each device in our sample by the following characteristics: approval year, medical specialty area,¹⁷ review type (normal/expedited), implantable designation (yes/no), and lifesaving designation (yes/no). We also characterized their recall history by searching the FDA’s online Medical Device Recalls Database using PMA application numbers and recording the highest recall class for each affected device (Class I-III).¹⁹

We primarily identified clinical studies using ClinicalTrials.gov; with the exception of small feasibility studies, the 2007 FDA Amendments Act required that all device studies ongoing as of December 2007 be registered on ClinicalTrials.gov.¹⁶ We used an inclusive, multistep search algorithm (Figure) that first screened for device trade names and then included a broader set of search terms with relevant manufacturer names, device descriptors, or both (eAppendix 1 in the Supplement). We used all relevant device trade names, component trade names, manufacturer names, and device descriptors in constructing ClinicalTrials.gov search terms, incorporating information contained in the PMA database, FDA Summaries, FDA webpages,^20,21 and manufacturer websites. We excluded studies with an enrollment status of “not yet recruiting,” “suspended,” or “withdrawn.” Included studies were required to either identify the medical device by trade name or examine a technologically equivalent, unnamed device that was both attributable to the correct manufacturer by study description or sponsorship and studied in a setting consistent with the marketing history outlined in the FDA Summary. All searches were performed by one author (V.K.R.) in October 2014 and confirmed for inclusion by a different author (J.S.R.). If an identified study compared 2 or more devices in our sample, the study was counted once for each device.

After our search of ClinicalTrials.gov, we then reviewed all feasibility and pivotal studies described in FDA Summaries and PAS listed within the FDA PAS database²²; pivotal studies are those that serve as the primary basis for the FDA’s premarket evaluation of device safety and effectiveness.²³ Studies described solely within FDA documents were included, even if not registered on ClinicalTrials.gov (eAppendix 1 in the Supplement). For devices with substudies conducted in support of the PMA application, we considered each named substudy with FDA-required follow-up of the premarket cohort as a separate PAS.

For all identified studies, we abstracted the following information from ClinicalTrials.gov, FDA documents, or both (eAppendix 2 in the Supplement): enrollment number, study status (completed, ongoing, terminated/unknown), primary completion date (ie, final data collection for primary end points), and study type (pivotal study, FDA-required PAS, or manufacturer/investigator-initiated study). Studies listed on ClinicalTrials.gov as “completed” were categorized as completed, studies listed as “recruiting,” “enrolling by invitation,” or “active, not enrolling” were categorized as ongoing, and studies listed as “terminated” or “unknown” were categorized as terminated/unknown. Studies with a primary completion date prior to initial FDA marketing approval were categorized as premarket; all other studies were categorized as postmarket.

We then abstracted additional information on study features and end point characteristics for all pivotal, completed postmarket, and ongoing postmarket studies; nonpivotal premarket and terminated/unknown studies were excluded from further analysis because the information available was often insufficient for characterization. We collected the following additional study features (eAppendix 3 in the Supplement): funder, centers, location, registry design, blinding, study groups, comparator, randomization, and indication. For all postmarket studies, we determined whether the described indication for use differed from the original FDA-approved indication; if there was insufficient information to make a determination, we categorized the indication as not differing (eAppendix 2 in the Supplement). We also categorized each primary end point as safety or effectiveness (eAppendix 2), and further examined all effectiveness end points, recording the duration of longest follow-up (using the prespecified duration for ongoing studies) and classifying end points as clinical outcomes (clinical outcomes or objectively measured function or symptom scales) or surrogate markers of disease based on an established framework and a recent Institute of Medicine report.^24,25 All data were abstracted by one author (V.K.R.); all characterizations of indications and primary end points were confirmed by a different author (J.S.R.), with conflicts resolved by consensus.

We used descriptive statistics to characterize our high-risk therapeutic device sample. We calculated median enrollment numbers for nonpivotal premarket, pivotal premarket, FDA-required PAS, and manufacturer/investigator-initiated postmarket studies and used the Kruskal-Wallis test to assess for a difference between these 4 study types. We then used descriptive statistics to characterize all other features of pivotal premarket, completed postmarket, and ongoing postmarket clinical studies; FDA-required PAS and manufacturer/investigator-initiated postmarket studies were categorized together to provide a holistic perspective of completed and ongoing postmarket evidence generation. Analyses of primary effectiveness end points were conducted at the end point level because some studies had multiple primary effectiveness end points and some studies had only safety end points. We then used χ², Fisher exact, and Kruskal-Wallis tests as appropriate to examine for differences in features and primary effectiveness end points between these 3 study types. Analyses were performed using Microsoft Excel 2011 and JMP version 10.0 (SAS Institute). All statistical tests were 2-tailed, and we used a type I error rate of .05 in testing enrollment number. To account for multiple comparisons, we used type I error rates of .006 and .0125 in testing all other study features (9 comparisons) and end point characteristics (4 comparisons), respectively.

Between 2010 and 2011, the FDA granted initial marketing approval for 28 high-risk therapeutic devices via the PMA pathway: 21 (75.0%) were implantable and 9 (32.1%) were life-sustaining (Table 1 and eTable in the Supplement). About half (n = 15; 53.6%) were for cardiovascular conditions. Ten devices (35.7%) were recalled at least once, with 1 (3.6%) undergoing a Class I recall (highest-risk: reasonable probability of serious health problems or death) and 1 (3.6%) voluntarily withdrawn from market.

We identified 286 clinical studies of these high-risk therapeutic medical devices (eFigure 2 in the Supplement): 82 (28.7%) premarket and 204 (71.3%) postmarket, among which there were 52 (18.2%) nonpivotal premarket studies, 30 (10.5%) pivotal premarket studies, 33 (11.5%) FDA-required PAS, and 171 (59.8%) manufacturer/investigator-initiated postmarket studies. A total of 44 (84.6%) nonpivotal premarket studies were reported as completed, as were all 30 (100.0%) pivotal premarket studies (Table 2). In contrast, 6 (18.2%) FDA-required PAS and 20 (11.7%) manufacturer/investigator-initiated postmarket studies were reported as completed, with 23 (69.7%) and 130 (76.0%) reported as ongoing, respectively; 2 (6.1%) FDA-required PAS were pending.

The median number of nonpivotal premarket studies per device was 1 (interquartile range [IQR], 0-2), while 26 (92.9%) devices received FDA approval on the basis of a single pivotal premarket study. At least 1 PAS was required by the FDA for 19 (67.9%) devices; nearly all (n = 29; 87.9%) were ordered as a condition of approval for the original PMA application, while the remainder (n = 4; 12.1%) were ordered after market introduction as a condition of approval for a supplemental PMA application. The median number of manufacturer/investigator-initiated postmarket studies was 3 (IQR, 1-6). We were unable to identify any postmarket studies (including completed, ongoing, or terminated/unknown studies) for 5 (17.9%) devices; 3 or fewer studies were identified for 13 (46.4%) devices overall.

Median enrollment was 65 patients (IQR, 25-111), 241 patients (IQR, 147-415), 222 patients (IQR, 119-640), and 250 patients (IQR, 60-800) for nonpivotal premarket, pivotal, FDA-required PAS, and manufacturer/investigator-initiated postmarket studies, respectively. Median enrollment was lower among completed FDA-required PAS and manufacturer/investigator-initiated postmarket studies (202 patients [IQR, 126-694] and 100 patients [IQR, 43-252], respectively) than among ongoing postmarket studies (300 patients [IQR, 120-1115] and 300 patients [IQR, 60-1011], respectively).

Although only 3 of 28 (10.7%) devices in our sample were coronary stents, 75 of 179 (41.8%) completed and ongoing postmarket studies (including FDA-required PAS) examined these devices. Among these coronary stent studies, median enrollment was 572 patients (IQR, 237-2000), whereas median enrollment was 135 patients (IQR, 50-326) for the 104 studies of all other devices. Focusing on the 10 devices in our sample that were recalled at least once, 67 of 104 (64.4%) “noncoronary stent” completed and ongoing postmarket studies examined these devices; median study enrollment was 130 patients (IQR, 50-318) for recalled devices and 165 patients (IQR, 40-346) for nonrecalled devices.

Study features were characterized for 209 studies: 30 (14.4%) pivotal premarket studies, 26 (12.4%) completed postmarket studies, and 153 (73.2%) ongoing postmarket studies (Figure). Whereas all pivotal studies were solely funded by industry (30/30 [100.0%)] and virtually all were multicenter (28/30 [93.3%]) and enrolled US patients (29/30 [96.7%]), fewer postmarket studies were supported by industry (completed: 17/26 [65.4%]; ongoing: 91/153 [59.5%]), were multicenter (completed: 18/26 [69.2%], ongoing: 92/153 [60.1%]), and enrolled US patients (completed: 15/26 [57.7%], ongoing: 63/153 [41.2%]) (all P values ≤.002) (Table 3). Pivotal and postmarket study design features were otherwise broadly similar, as approximately 10% were designated registries, roughly three-quarters were unblinded, and nearly half were single-group and thus had no comparator. Among multigroup studies, more than three-quarters used active comparators and were randomized. In addition, nearly half of all postmarket studies (83/179 [46.4%]) explicitly described examining devices for different indications than those originally approved by the FDA (completed: 9/26 [34.6%], ongoing: 74/153 [48.4%]).

We identified 226 primary effectiveness end points among these 209 studies: 44 (19.5%) end points among 30 pivotal studies, 27 (11.9%) end points among 26 completed postmarket studies, and 155 (68.6%) end points among 153 ongoing postmarket studies (Figure). Nearly 80% (35/44) of pivotal study end points were clinical outcomes, in contrast to 57.1% of postmarket study end points (completed: 14/27 [51.9%], ongoing: 90/155 [58.1%]; P = .02) (Table 4 and eAppendix 2 in the Supplement). Median duration of end point follow-up was 3.0 months (IQR, 3.0-12.0) for pivotal studies, 9.0 months (IQR, 0.3-12.0) for completed postmarket studies, and 12.0 months (IQR, 7.0-24.0) for ongoing postmarket studies (P = .002). However, we found no difference in median duration of end point follow-up for implantable device studies (pivotal: 12.0 months [IQR, 4.0-12.0], completed postmarket: 10.5 months [IQR, 0.3-21.0], ongoing postmarket: 12.0 months [IQR, 8.0-24.0]; P = .07).

Our characterization of the clinical studies examining high-risk therapeutic medical devices initially approved via the FDA PMA pathway between 2010 and 2011 demonstrates that the amount and quality of evidence generated over the total product life cycle varies widely. Some devices are currently being evaluated through ongoing studies that, if completed, will provide evidence on clinical outcomes for large numbers of patients with planned follow-up of a year or longer. However, most devices have been or will be evaluated through only a few studies, which often focus on surrogate markers of disease in small numbers of patients followed up over short periods of time and study indications that differ from the original FDA-approved indication.

Premarket clinical studies of high-risk therapeutic devices were limited in number and quality. Nearly all devices were cleared on the basis of 2 studies: 1 nonpivotal and 1 pivotal study. Nonpivotal studies are typically conducted to assess device feasibility, enrolling a limited number of patients to examine device performance and guide premarket development (eg, design modifications) and clinical use (eg, anatomical restrictions).²⁶ Nonpivotal studies may also include internationally based studies initiated prior to FDA approval; in our study, all incomplete nonpivotal premarket studies were internationally based. In addition, to support market approval, the FDA requires at least 1 pivotal study with substantial evidence of device safety and effectiveness.²³ We found that pivotal studies generally enrolled fewer than 300 patients and were often designed without blinding, comparators, or primary end point follow-up exceeding 1 year. Our findings are consistent with previous studies of premarket evidentiary standards focused on devices used for cardiovascular diseases, rare conditions, and patients who are children or have unmet medical needs^4,5,27,28 and confirm that premarket studies provide limited data to address important clinical questions that often arise after approval, including those related to long-term device performance, new indications or iterations, and safety and effectiveness in real-world populations.^27,29-31

Prior studies have not examined total product life cycle evidence generation for high-risk therapeutic devices, instead focusing solely on the FDA PAS program or orthopedic prostheses, which often receive market clearance via the 510(k) regulatory pathway intended for moderate-risk devices.^9,32 We found that postmarket studies, like premarket studies, were often small, unblinded, and without comparators. In addition, postmarket studies—including those examining implantable devices—were also generally limited to 1 year of primary end point follow-up, and nearly half focused on surrogate markers of disease. However, approximately 13% of identified postmarket studies were completed between 3 and 5 years after FDA approval. Postmarket evidence may be generated from ongoing observational studies and registries before completing primary effectiveness end point follow-up, as well as afterwards from longer-term follow-up of safety end points. However, the potential for this postmarket evidence to inform practice remains unclear, even under the presumption that all ongoing studies will be completed, given that clinicians often rapidly adopt new devices after market introduction^33,34 and shortcomings of the medical device literature related to selective publication and selective outcome reporting.³⁵ Furthermore, it’s not clear how this evidence will inform regulatory decisions, if at all, such as whether to recall a product. Interestingly, completed and ongoing postmarket studies examining recalled and nonrecalled devices were similar in size.

The FDA has adopted a total product life cycle approach to device evaluation with the understanding that “[a]t the time of device approval, certain safety and effectiveness questions may not be fully resolved […] because controlled clinical studies do not fully represent the benefit-risk profile of a device when used in real-world clinical practice.”¹³ Although the FDA may not require a PAS for every newly approved device, the agency often requires a postmarket study to complement premarket understanding of device safety and effectiveness. However, by law, the FDA may only require the “least burdensome” postmarket data necessary to address unresolved clinical questions about devices,¹⁴ limiting its capacity to mandate additional studies for the purpose of generating evidence to inform regulatory and clinical decision making. Furthermore, the FDA has not imposed penalties against manufacturers failing to comply with postmarket study requirements mandated through its PAS program.⁹ Our findings of limited premarket evidence generation and few FDA-required postmarket studies highlight the need for continued study, either through manufacturer-initiated or investigator-initiated studies, to advance postmarket understanding of device safety and effectiveness. Approximately 85% of the postmarket studies we identified were not initiated in response to FDA requirements, and 40% were conducted without manufacturer support. To ensure generation of additional robust, objective evidence to inform the use of high-risk devices in clinical practice, government agencies may consider taking on a more principal role in supporting postmarket research, as they have done for several commonly used pharmaceutical products.

The “right” number and “appropriate” design of premarket and postmarket studies for high-risk therapeutic devices should vary based on expected benefit and risk, therapeutic alternatives, and anticipated challenges of real-world use, including physician learning curves and facility expertise. For any given device, conducting numerous large studies with long periods of follow-up may not be a feasible or efficient use of resources. However, pending legislative efforts will only further reduce premarket evidence requirements for medical devices in order to expedite patient access to new technologies.³⁶ Although the FDA has begun developing postmarket safety surveillance methods, used primarily for pharmaceuticals and biologics, which leverage routinely collected electronic health information through a distributed data model under its Mini-Sentinel initiative,³⁷ the validity of these methods remains uncertain, and this approach cannot be used for surveillance of medical devices until there is widespread adoption of unique device identifiers.³⁸ Moreover, safety surveillance efforts have uncertain applications for generation of comparative effectiveness evidence or insights into long-term effectiveness of medical devices. Postmarket assessments of both medical device safety and effectiveness in real-world practice, through clinical trials or registries, or analysis of health systems data, may provide complementary evidence to guide regulatory and clinical decision making.

Our study has several limitations that deserve consideration. First, we may not have identified all clinical studies of devices in our sample despite the inclusive nature of our search algorithm, and our findings may thus underrepresent the clinical evidence generated. This is more likely true of nonpivotal premarket clinical studies, as these could have taken place prior to the ClinicalTrials.gov registration requirements that took effect in late 2007. Nevertheless, all pivotal studies were identified, and these studies represent the most robust evidence available during premarket evaluation. Conversely, by including all studies registered on ClinicalTrials.gov, our study may overrepresent the clinical evidence generated, particularly in the postmarket period; approximately one-third of clinical trials remain unpublished even years after study completion,³⁹ and only one-fifth of completed trials registered on ClinicalTrials.gov report their results,⁴⁰ such that the results of many studies we identified may never be disseminated to inform clinical practice.

Second, we cannot account for postmarket studies not registered on ClinicalTrials.gov, such as medical record reviews or case studies, though the strength of evidence derived from these studies is often limited. In addition, internationally based studies may also be less likely to be registered on ClinicalTrials.gov, although more than half of the postmarket studies we identified were conducted entirely outside of the United States. Similarly, observational studies and patient registries of medical devices are required to be registered on ClinicalTrials.gov under the FDA Amendments Act and composed nearly half of the postmarket studies we identified. However, non–product-specific registries (ie, disease registries) were unlikely to have been identified and may contribute to device evaluation over the total product life cycle.

Third, our analysis was cross-sectional and our search was completed in October 2014, allowing between 3 and 5 years for studies to be initiated and completed after FDA approval. It is likely that there will be additional clinical studies examining these devices, and some of the studies we identified as ongoing will be completed or already have been completed. However, we expect that most major postmarket clinical studies of devices are likely to be initiated within 5 years of approval given their relatively short market life, and our findings therefore likely reflect the best evidence available and anticipated to inform clinical practice. Furthermore, our study was focused on evidence generated for high-risk therapeutic devices receiving PMA approval. Our findings do not apply to devices receiving market clearance via the 510(k) or Humanitarian Device Exemption regulatory pathways, which are used less frequently for high-risk devices, nor to diagnostic devices receiving PMA approval.

Among high-risk therapeutic devices approved via the FDA PMA pathway between 2010 and 2011, total product life cycle evidence generation varied in both the number and quality of premarket and postmarket studies, with approximately 13% of initiated postmarket studies completed between 3 and 5 years after FDA approval.

Corresponding Author: Joseph S. Ross, MD, MHS, Section of General Internal Medicine, Yale University School of Medicine, PO Box 208093, New Haven, CT 06520-8093 (joseph.ross@yale.edu).

Author Contributions: Mr Rathi and Dr Ross had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Rathi, Masoudi, Ross.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Rathi, Ross.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Rathi.

Obtained funding: Rathi, Krumholz.

Study supervision: Ross.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Krumholz reported that he chairs a scientific advisory board for UnitedHealthcare. Dr Masoudi reported having received a contract (through the University of Colorado) from the American College of Cardiology. Dr Ross reported having received grants from the Pew Charitable Trusts. No other disclosures were reported.

Funding/Support: This project was not supported by any external grants or funds. The authors assume full responsibility for the accuracy and completeness of the ideas presented. Mr Rathi is supported by the Yale University School of Medicine Office of Student Research. Drs Krumholz and Ross receive support through Yale University from Medtronic and Johnson & Johnson to develop methods of clinical trial data sharing, from the Centers of Medicare & Medicaid Services to develop and maintain performance measures that are used for public reporting, and from the Food and Drug Administration to develop methods for postmarket surveillance of medical devices. Dr Krumholz is supported by a National Heart, Lung, and Blood Institute Cardiovascular Outcomes Center Award (1U01HL105270-02). Dr Ross is supported by the National Institute on Aging (K08 AG032886) and by the American Federation for Aging Research through the Paul B. Beeson Career Development Award Program.

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.