Abstract
Background/Aims
Data integrity in multicenter and longitudinal studies requires implementation of standardized reproducible methods throughout the data collection, analysis, and reporting process. This requirement is heightened when results are shared with participants that may influence health care decisions. A quality assurance plan provides a framework for ongoing monitoring and mitigation strategies when errors occur.
Methods
The Diabetes Control and Complications Trial (1983–1993) and its follow-up study, the Epidemiology of Diabetes Interventions and Complications (1994–present), have characterized risk factors and long-term complications in a type 1 diabetes cohort followed for over 40 years. An ancillary study to assess bone mineral density was implemented across 27 sites, using one of two dual x-ray absorptiometry scanner types. Centrally generated reports were distributed to participants by the sites. A query from a site about results that were incongruent with a single participant’s clinical history prompted reevaluation of this scan, revealing a systematic error in the reading of hip scans from one of the two scanner types. A mitigation plan was implemented to correct and communicate the errors to ensure participant safety, particularly among those originally identified as having low bone mineral density scores for whom antiresorptive treatment may have been initiated based on these results.
Results
The error in the analysis of hip scans from the identified scanner type resulted in lower bone mineral density scores in scans requiring manual deletion of the ischium bone. Hip scans with original T-score ≤ −2.5 (n = 84) acquired on either scanner were reviewed, and reanalyzed if the error was detected. Fourteen scans were susceptible to this error and reanalyzed: nine scans were reclassified from osteoporosis to low bone mineral density, one from low to normal bone mineral density, and four were unchanged. All errors occurred on one scanner type. An integrated communication and intervention plan was implemented. The nine participants whose scans were reclassified from osteoporosis to low bone mineral density were contacted; five were using antiresorptive treatment, all of whom had other risk factors for fracture beyond these scan results. Review of all hip scans with a T-score > −2.5 (n = 371) using this scanner type identified 27 additional hip scans that required reanalysis and potential reclassification: 1 scan was reclassified from osteoporosis to low bone mineral density, 11 from low to normal bone mineral density, and 15 were unchanged.
Conclusion
The impact of an analysis error on participant safety, specifically when the initiation of unnecessary treatment may result, necessitated implementation of a coordinated communication and mitigation plan across all clinical centers to ensure consistent messaging and accurate results are provided to participants and their local care providers. This framework may serve as a resource for other clinical studies.
Introduction
Multicenter and longitudinal studies rely on standardized methods throughout the data collection process to help ensure the consistency and comparability of data across all clinical centers. Centralized analysis of measurements ensures uniform data processing and reporting, and a pre-defined analysis plan guides the eventual interpretation and identification of outcomes that are accurate and meaningful.1–8 Any potential systematic errors in the measurement, analysis, and/or interpretation of the data may adversely affect participant safety, as well as the overall validity and strength of the study to address its primary study objectives. Development of a timely coordinated plan to determine the magnitude of any identified error and subsequent implementation of a structured mitigation process are integral to an effective quality assurance strategy.
The Diabetes Control and Complications Trial (DCCT: 1983–1993) and its follow-up study, the Epidemiology of Diabetes Interventions and Complications (EDIC: 1994-present), provide an expansive background to illustrate the importance of a durable quality assurance plan to ensure data accuracy, completeness, and quality over time. 1 During 2017–2022, the primary focus of the DCCT/EDIC study was to understand the impact of long-standing type 1 diabetes and blood glucose control on severe microvascular and cardiovascular complications, as well as on cognitive and physical functioning in an aging cohort with type 1 diabetes. In conjunction with the core EDIC study, an ancillary study was conducted across all 27 EDIC clinical centers in the United States and Canada to assess bone mineral density (Skeletal Health, 2017–2019) using dual-energy x-ray absorptiometry (DXA), a standardized protocol, and defined analysis procedures.
Within the context of a multicenter study, diligence in following the protocol, evaluating unexpected occurrences, and questioning results that may be incongruent with the clinical history are critical to the integrity of overall study results and to preservation of participant confidence. In addition, instances where research results can potentially impact clinical decision-making require a higher level of scrutiny to ensure participant safety.
Notably, a systematic error in the centralized analysis of hip DXA scans in a small subset of participants was discovered due to the diligence of a local investigator who questioned an individual report of osteoporotic bone mineral density (BMD) that was inconsistent with the participant’s history. This discovery necessitated the implementation of a methodical process to identify, isolate, and address the analysis error. Of primary concern was participant safety and the potential impact of the error on subsequent treatment decisions by the participants’ health care providers, particularly among those at greatest risk for initiation of unnecessary treatment. In addition, the error called into question the validity of the overall study results and prompted a systematic multi-step evaluation to ensure the absence of other errors and confirm the results.
Herein, we provide details of the reanalysis and quality control measures implemented in response to the identified error and describe the communication strategy developed to inform the EDIC Study Group, the study sponsor, the participants, and the study’s safety monitoring board. We describe the specific direction provided to local clinical staff to guide communication with the participant and their local health care provider and provide an overview of the additional quality assurance measures implemented to verify the accuracy and reproducibility of all the centrally assessed measures involved in this ancillary study.
Methods
EDIC participants with type 1 diabetes
The DCCT was a multicenter randomized clinical trial designed to evaluate the impact of glycemia on the occurrence and progression of microvascular complications associated with type 1 diabetes.9,10 In brief, at baseline, participants were 13–39 years of age, had type 1 diabetes for 1–15 years, and were free of significant complications associated with diabetes. Participants were randomized to receive intensive or conventional diabetes therapy and followed for a mean of 6.5 (3–9) years. Intensive therapy aimed to achieve glycemic control as close to the non-diabetic range as safely possible using multiple daily insulin injections or an insulin infusion pump with dosing guided by frequent daily self-monitored blood glucose. Conventional therapy utilized less intense insulin delivery and monitoring methods and focused on the avoidance of episodes of hypo- or hyperglycemia. Frequent standardized evaluations to assess the occurrence and progression of retinal, renal, neurologic, and cardiovascular changes were conducted. Central analysis centers, each with defined internal quality assurance measures to ensure data consistency across the 29 clinical centers and over time, analyzed the data. Following the conclusion of the DCCT, 96% of the surviving cohort enrolled in the EDIC longitudinal observational study designed to evaluate the longer-term impact of type 1 diabetes and associated risk factors on clinical outcomes. 11
During EDIC follow-up years 24–26 (calendar years 2017–2019), all active EDIC participants were invited to participate in the EDIC Skeletal Health ancillary study. 12 Eligible participants were asked to complete DXA scans of the hip, spine and radius, and lateral spine for vertebral fracture assessment. At the end of data collection, 1058 (92% of all active EDIC participants) enrolled and completed at least one set of DXA scans. In addition, participants at the six clinical centers with access to a high-resolution peripheral quantitative computed tomography (HR-pQCT) scanner underwent radial and tibial scans. 13 The study was approved by the institutional review boards (IRBs) at all participating centers and all participants gave written informed consent.
EDIC controls without type 1 diabetes
During the same timeframe, 128 control subjects without diabetes were recruited at the six clinical centers with access to HR-pQCT scanners. At the end of data collection, 103 (80%) of the control subjects completed at least one set of DXA scans. 12
Dual-energy x-ray absorptiometry
DXA scans were obtained at the clinical centers and analyzed centrally at the EDIC DXA Analysis and Quality Assurance Unit (DXA Analysis Unit). All clinical centers were equipped with a scanner from one of two manufacturers, hereafter referred to as Scanner A and Scanner B. Longitudinal scanner performance was monitored by the DXA Analysis Unit using spine phantom scans. A set of phantoms were also circulated to the local sites for purposes of cross-calibration. DXA technicians at each of the clinical centers were trained and certified by the DXA Analysis Unit using a standardized protocol. 12
Analysis and quality control of DXA scans
Participant scans were reviewed at the DXA Analysis Unit for appropriate acquisition. Scans with poor acquisition (e.g. participant motion during the scan) were considered unacceptable and were not included in the analysis dataset. BMD was reported at five locations: total hip, femoral neck, lumbar spine, distal radius, and ultra-distal radius. At the DXA Analysis Unit, one technical analyst was primarily responsible for analyzing the hip, spine, and radius scans. T-scores were calculated for the total hip, femoral neck, and lumbar spine, indicating how much higher or lower the individual’s bone density was compared to a healthy 30-year-old woman (T-score = 0). 14 The lowest T-score from the total hip, femoral neck, or lumbar spine was used to classify overall bone density into three categories: normal bone density (T-score > −1.0), low bone density (−1.0 ≥ T-score > −2.5), or osteoporosis (T-score ≤ −2.5). A second analyst, with specific training in vertebral fracture assessment, graded the lateral spine scans for presence of vertebral deformities, using the Genant semi-quantitative (SQ) method. 15
Participant report
The EDIC Data Coordinating Center generated a Bone Density and Vertebral Fracture Assessment report for each EDIC participant and control subject based on DXA results received from the DXA Analysis Unit. Participant- and control-specific reports included a description of the testing performed, BMD T-scores at the three anatomical locations (total hip, femoral neck, and lumbar spine), bone density classification, and a vertebral fracture assessment, if available. These reports were sent to the investigators at each of the EDIC clinical centers who then shared the results with their participants and control subjects. The local investigator was responsible for reviewing results before distribution to the participants and their health care providers, particularly for those with evidence of osteoporosis.
Statistical methods
The five primary DXA BMD locations (total hip, femoral neck, lumbar spine, distal radius, and ultra-distal radius) were assessed as quantitative outcomes, while the presence versus absence of a vertebral fracture was assessed as a binary outcome.
The process flow of activities is presented in Figure 1. Steps 1 through 4 required rereview of scans at the DXA Analysis Unit and descriptive analyses were used to summarize disease reclassification. Once participants were reclassified, as part of the quality control analysis (Figure 1, Step 5), the five BMD locations were reanalyzed in a random sample of 100 participants, stratified by scanner type (Scanners A and B). Intra-class correlation coefficients were calculated for each anatomic location to assess the reproducibility of BMD measurements between the original technical analyst and the secondary independent analyst. 16 Bland–Altman plots were also produced for the total hip and femoral neck. Lateral spine scans were also reanalyzed in another random sample of 100 participants, stratified by scanner type (Scanners A and B), to reassess vertebral fractures. Kappa statistics were used to evaluate the agreement between the original analyst and the secondary independent analyst in assessing the presence or absence of vertebral fractures.8,17 All analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA).
Detection and mitigation of systematic analysis error.
Results
At the time of these analyses, DXA scans were collected from 1049 EDIC participants and 69 controls without diabetes, representing 99% and 67% of the final sample cohorts, respectively: 401 scans (36%) were collected using Scanner A and 717 scans (64%) using Scanner B (Figure 2). The mean (SD) age of participants was 59 (7) years, mean duration of type 1 diabetes was 38 (5), and 48% were women. 12
Flow chart of hip DXA scans. Boxes outlined in bold represent hip scans requiring reanalysis. None of the scans obtained from Scanner B were affected by the systematic analysis error and no further action was required. *There were two unreadable scans.
Detection of systematic analysis error in hip DXA scans
Toward the end of the data collection period, study staff at 1 of the 27 clinical centers raised concerns about one of their participant’s results. The participant’s clinical history was void of prior positive risk factor evaluations and fracture history, yet the results indicated BMD in the osteoporotic range. Review of this participant’s scans revealed an error in the analysis protocol used for the hip scan obtained from Scanner A. Specifically, the hip scan obtained on this participant required manual deletion of the ischium bone during analysis, but this deletion was performed incorrectly. The deleted area was defined as “tissue” rather than “neutral.” This error resulted in artificially lower BMD T-scores for the total hip and femoral neck, compared to values expected using the correct analysis protocol. The clinical implication of this error was an incorrect assessment of osteoporosis and the potential for unwarranted initiation of antiresorptive medication based on the initial BMD T-scores. The research implication was the introduction of measurement error into the BMD results, the primary outcome of this ancillary study.
Review of hip and spine scans in the osteoporotic range
To address the concern for participant safety, the first step was to identify and review hip and spine scans on all participants (n = 84) who had a BMD T-score in the osteoporotic range (T-score ≤ −2.5). Although the error appeared to be limited to hip scans, both hip and spine scans were reviewed to be certain that other errors were not occurring. Review by a second analyst at the DXA Analysis Unit confirmed that errors were limited to those hip scans requiring manual deletion of the ischium bone that were obtained on Scanner A. Importantly, manual deletion of the ischium bone was not required on scans obtained using Scanner B. The review identified 14 out of 30 “osteoporotic” hip scans from Scanner A that required reanalysis due to incorrect deletion of the ischium bone, including the initially identified hip scan. After reanalysis, nine participants were reclassified from osteoporosis to low BMD, one from low to normal BMD, and four had no change in classification. Of the nine participants reclassified from osteoporosis to low BMD, two had vertebral fractures, which were considered a clinical sign of osteoporosis independent of BMD T-score (Figure 1).
Mitigation plan
A working group consisting of representatives from the DXA Analysis Unit, the Data and Clinical Coordinating Centers, and the Data Quality Assurance committee defined and implemented a three-step mitigation plan to address participant safety and data integrity issues with direct accountability to the EDIC Executive Committee. The mitigation plan included (1) notification of the error, and follow-up with the 10 participants incorrectly classified as osteoporotic by BMD based on review of selected hip scans described above; (2) review of the remaining hip scans obtained on Scanner A to identify and reanalyze any with incorrect deletions of the ischium bone; and (3) masked reanalysis of a random sample of 100 DXA scans (hip, spine, radius) and masked reanalysis of a separate random sample of 100 lateral spine scans to assess inter-rater agreement, both stratified by scanner type (Scanners A and B).
Notification and follow-up of analysis error and corrected results
Once the affected scans resulting in reclassification were identified (n = 10), expedited communication was sent to all clinical centers and the EDIC Observational Study Monitoring Board (OSMB). Subsequently, targeted communications were sent to the six clinical centers with the impacted scans that outlined the expected communication to the participants and their local health care providers, as appropriate, and to each center’s IRB. Each of these centers received the corrected results reports and were advised to contact each affected participant to inform them of the error, determine whether antiresorptive medication was started solely based on the original EDIC scan result, determine whether this medication was continued or changed post-notification, and identify the existence of other risk factors for decreased bone health. Among the 10 participants with reclassified results, 1 had not yet been notified of the original result and 4 were receiving antiresorptive medication; following notification, treatment was continued for all 4 participants due to the presence of other known risk factors.
Review of remaining hip scans
The DXA Analysis Unit reviewed all of the remaining hip scans (n = 371) obtained from Scanner A to identify any that had incorrect manual deletion of the ischium bone and required reanalysis. Of the 371 scans, additional 27 hip scans were identified that required reanalysis. This resulted in 1 participant reclassified from osteoporosis to low BMD, 11 from low to normal BMD, and 15 with no change in classification. Of the 11 participants reclassified from low to normal BMD, 2 also had vertebral fractures. A similar series of communications were sent to the clinical centers and the EDIC OSMB. The 7 clinical centers with the 12 affected participants were advised to follow-up with each affected participant and document treatment initiation, adverse events, and treatment changes made based on the reclassification. Among these 12 participants, none had initiated antiresorptive medication based on the EDIC DXA results.
Feedback from IRBs and affected participants
Clinical centers with one or more participants identified as having received inaccurate DXA results were instructed to inform the IRBs at their respective institutions. Seven IRBs were contacted: three IRBs acknowledged receipt of the information and did not require further action and four classified this as an adverse event and required additional documentation. One of these IRBs required additional detailed information about the internal quality control processes that was provided by the DXA Analysis Unit and the EDIC Data Coordinating Center. Questions raised by the IRBs and OSMB throughout the entire process focused on the error detection, frequency, and prevention; the expected detection of the error in clinical research compared to clinical practice; and a description of a defined quality assurance plan and adherence to the defined methods.
Participants whose results were reclassified were evaluated to determine the impact of this error on their overall participation in the EDIC study. No participants reported a desire to withdraw from the EDIC study or the Skeletal Health ancillary study. Responses indicated disappointment that the error had been made, appreciation for notification of the revised results, and relief that the results were better than originally reported.
Masked reanalysis in random samples of participant scans
Once the urgent participant safety issues were addressed, additional measures were implemented to verify overall data quality. For each scanner type (A and B), a random sample of 50 participants were identified by the Data Coordinating Center from those with a complete set of DXA scans (hip, spine, radius). A second analyst analyzed these randomly selected scans (50 per scanner type) without knowledge of the original analysis results. The same five primary DXA BMD locations were analyzed: total hip, femoral neck, lumbar spine, distal radius, and ultra-distal radius. The intra-class correlation between the two analysts across all five anatomic locations was ≥0.93; no additional systematic errors were identified (Table 1, Figure 3).
Intra-class correlation coefficients among five anatomic locations to assess inter-rater reliability between two technical analysts, by scanner type.
Data are intra-class correlation coefficients for each anatomic location representing the reproducibility of BMD measurements between the original technical analyst and the secondary independent analyst.
Bland–Altman plots for total hip and femoral neck, by scanner type. The difference between BMD measurements analyzed by the original technical analyst and the secondary independent analyst is displayed on the vertical axis against the mean of the two measures on the horizonal axis.
In addition, a masked grading of another random sample of 100 lateral spine scans was performed to reevaluate vertebral fractures. Briefly, for SQ grading, each vertebral level that could be evaluated was given an SQ score classified as no (0), mild (1), moderate (2), or severe (3) vertebral deformity. “Any deformity” was defined as a score of 1–3 at one or more vertebral levels. “Any fracture” was defined as a score of 2–3 at one or more levels. The random sample was stratified by scanner type (Scanners A and B) and by any/no deformity in order to over-sample scans initially graded as having “any deformity” (19% of scans). First, 50 scans were sampled from participants with all vertebral levels graded as SQ = 0, split evenly between the two scanner types (A and B). An additional 50 scans were selected from participants with at least one vertebral level graded SQ > 0, evenly split between the two scanner types (A and B). This sample of 100 scans was graded by the second analyst without knowledge of the original analysis. The kappa statistic (95% confidence interval) between the two analysts after adjustment for scanner type was 0.45 (0.20, 0.69) and 0.46 (0.30, 0.63) for “any fracture” and “any deformity,” respectively. Although agreement between the analysts was low, it confirmed the absence of any further systematic errors in the analyses.
The final data set incorporated the corrected values from 41 hip scans (total hip and femoral neck BMD). The Data Quality Assurance Committee monitored ongoing quality assurance activities prior to and subsequent to the detection of the systematic analysis error. The EDIC Executive Committee monitored the impact of the errors, guided the communication strategy to all of the clinical centers, evaluated the impact of the corrections on the bone health of the study participants, advised communications with the local IRBs, and maintained communication with the EDIC OSMB throughout the process.
Discussion
The discovery of a systematic error in the measurement of a study outcome where these results were also reported to participants in the DCCT/EDIC study provided an excellent opportunity to evaluate the implementation and effectiveness of a mitigation plan in a multicenter clinical study.
This ancillary study was conducted across 27 EDIC clinical centers using standardized collection and transmission procedures. Local technicians were trained and certified to perform study procedures; quality assurance measures were implemented routinely to ensure consistency in data collection across the clinical centers, scanner types, and among the technical staff. The Central Analysis Unit provided uniform training and ongoing monitoring of the quality of local scan acquisition and the performance of the scanners. Centrally, one primary analyst analyzed BMD scans and a second analyst graded vertebral fracture assessment scans.
Participant results reports were based on centrally generated data distributed to the clinical centers. The clinical centers evaluated the participant-specific results and shared these with the participants. The decision regarding further evaluation or treatment based on these results was deferred to the participant’s health care provider. An inquiry from one of the participating EDIC clinical center investigators questioning a participant’s results prompted the discovery of a systematic error in the analysis of hip scans obtained from one of the scanners. Of note, at preceding EDIC Study Group Meetings, the study teams were provided with expert presentations on BMD, risk factors for osteoporosis, imaging modalities, and current diagnostic and treatment guidelines. This enabled local study teams to serve as knowledgeable resources and advocates for the participants. To ensure participant safety, an organized study-wide approach and mitigation plan was developed. Close collaboration between the Central Analysis Unit, the Data Coordinating Centers, and study leadership was necessary to ensure that an integrated plan and strategies were implemented that focused on protection of participants and preservation of overall data integrity. Consistent and frequent centralized communication with the clinical centers ensured that all affected staff and participants received consistent guidance and accurate results.
The skeletal health research results from this ancillary study were generated based on a standardized research protocol. However, given the universal availability and use of DXA scans to assess bone health in routine clinical care, it was important to share these results with study participants and their local health care provider as appropriate. The research results alone were insufficient to prompt clinical intervention or guide treatment decisions. However, the analysis error had the potential to influence clinical care if the decision for treatment was based solely on the research results. The implications of a data error on participant trust in the study and study personnel merited attention. While disconcerting to some, the affected participants were generally appreciative of the information and guidance received.
Governing bodies, such as local IRBs and study monitoring committees, are appropriately concerned when errors occur during the conduct of a study especially if the errors may affect participant safety. As expected, these bodies sought further clarification and additional information about central analysis and quality measures to aid in the interpretation of the data and carefully considered the mitigation strategies to provide a comprehensive assessment of participant risk.
The success of this mitigation plan relied on several factors: close collaboration between the Central Analysis Unit and Data Coordinating Center; consensus on the specifics of the mitigation plan; frequent, consistent, and transparent communication between study leadership and the clinical centers; and the collaborative research partnership between the clinical investigators and the study participants.
Conclusion
Valid research depends on the generation of accurate, reliable, and reproducible results. Consistent methods for data collection, processing, and reporting are essential while quality control measures implemented and monitored over time allow the identification of systematic errors in the data acquisition and management process. Within the context of a multicenter study, diligence in following the protocol, evaluating unexpected occurrences, and questioning results that may be incongruent with the clinical history are critical to the integrity of overall study results and to preservation of participant confidence. Instances where research results can potentially impact clinical decision-making require a higher level of scrutiny to ensure participant safety and underscore the importance of a timely, unified, and consistent mitigation strategy.
We have described the impact of a single questionable test result. The development of a systematic mitigation plan to identify and quantify the issue, address the ramifications of the systematic error, and address the impact of the results on participant safety as described herein can help ensure that data integrity, regulatory requirements, and participant safety are appropriately considered and protected. The methods employed to detect and mitigate a systematic error encountered in this study may provide a useful template for other research studies to consider when faced with management of an error that has potential to threaten the integrity of study-wide results and participant safety.
Supplemental Material
sj-pdf-1-ctj-10.1177_17407745251328257 – Supplemental material for Identification and mitigation of a systematic analysis error in a multicenter dual-energy x-ray absorptiometry study
Supplemental material, sj-pdf-1-ctj-10.1177_17407745251328257 for Identification and mitigation of a systematic analysis error in a multicenter dual-energy x-ray absorptiometry study by Gayle M Lorenzi, Barbara H Braffett, Ionut Bebu, Victoria R Trapani, Jye-Yu C Backlund, Kaleigh Farrell, Rose A Gubitosi-Klug and Ann V Schwartz in Clinical Trials
Footnotes
Acknowledgements
A complete list of members in the DCCT/EDIC Research Group is presented in the Supplementary Material. The DCCT/EDIC Research Group owes its scientific success and public health contributions to the dedication and commitment of the DCCT/EDIC participants. The authors acknowledge the members of the EDIC Skeletal Health Working Group: Valerie Arends, Jye-Yu C. Backlund, Annette Barnie, Ionut Bebu, Barbara H. Braffett, Andrew Burghardt, Ian De Boer, Kaleigh Farrell, Naina Sinha Gregory, Rose Gubitosi-Klug, Galateia Kazakia, David J. Kenny, John M. Lachin, Ming-Hui Lin, Thomas Link, Mishaela Rubin, Ann V. Schwartz, Victoria R. Trapani, and Amisha Wallia.
Author contributions
G.M.L. and A.V.S. wrote the manuscript. B.H.B., I.B., J.Y.B., and V.R.T. conducted the statistical analysis. All authors reviewed the manuscript for critical content, and approved the final version.
Industry contributions
Industry contributors have had no role in the DCCT/EDIC study but have provided free or discounted supplies or equipment to support participants’ adherence to the study: Abbott Diabetes Care (Alameda, CA), Animas (Westchester, PA), Bayer Diabetes Care (North America Headquarters, Tarrytown, NY), Becton, Dickinson and Company (Franklin Lakes, NJ), Eli Lilly (Indianapolis, IN), Extend Nutrition (St. Louis, MO), Insulet Corporation (Bedford, MA), LifeScan (Milpitas, CA), Medtronic Diabetes (Minneapolis, MN), Nipro Home Diagnostics (Ft. Lauderdale, FL), Nova Diabetes Care (Billerica, MA), Omron (Shelton, CT), Perrigo Diabetes Care (Allegan, MI), Roche Diabetes Care (Indianapolis, IN), and Sanofi-Aventis (Bridgewater, NJ).
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The DCCT/EDIC has been supported by cooperative agreement grants (1982–1993, 2012–2017, 2017–2022), and contracts (1982–2012) with the Division of Diabetes Endocrinology and Metabolic Diseases of the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK; current grant numbers U01 DK094176 and U01 DK094157), and through support by the National Eye Institute, the National Institute of Neurologic Disorders and Stroke, the General Clinical Research Centers Program (1993–2007), and Clinical Translational Science Center Program (2006–present), Bethesda, Maryland, USA. The sponsor of this study is represented by the NIDDK Project Scientist who serves as part of the DCCT/EDIC Research Group and plays a part in the study design and conduct as well as the review and approval of manuscripts. The NIDDK Project Scientist was not a member of the writing group of this paper. The opinions expressed are those of the investigators and do not necessarily reflect the views of the funding agencies.
Guarantor statement
A.V.S. and B.H.B. are the guarantors of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Trial registration
clinicaltrials.gov NCT00360815 and NCT00360893.
Data sharing
Supplemental material
Supplemental material for this article is available online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
