SPRY Core Protocol

SPRY is the first Core Protocol outlining the embedding of a trial within the EHR and routine perioperative healthcare delivery for at-risk aged adults. The Core Protocol creates standardised trial elements shared by all applied domains, preventing the continuous development and then dismantling of the expensive and complex clinical trial infrastructure.22 As with other adaptive platform trials, SPRY will assess multiple domains simultaneously using Bayesian statistical analysis and response adaptive randomisation evaluating the treatment effect in predefined strata (eg, vascular, orthopaedic, hepatobiliary surgical interventions).23 24 In the REMAP design, patients can be randomised to one of many treatments within one of many simultaneously deployed domains resulting in multiple possible experimental treatment combinations. The Core Protocol allows for aggregation of the treatment response across different simultaneously investigated domains and the multifactorial evaluation of synergistic or antagonistic combinations within each of the strata.

SPRY trial flow per the Consolidation Standard of Reporting Trials (CONSORT) guidelines are adapted from the traditional linear format into a concentric diagram, demonstrating the perpetual nature of the Core Protocol (figure 1A).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Concentric consort diagram: SPRY Core Protocol (A) and Domain-Specific Appendix SPRY-Metformin overlying the Core Protocol (B). (A) The Core Protocol creates a research platform or infrastructure within clinical care for all enrolled into any SPRY Domain-Specific Appendix. This infrastructure includes virtual screening, informed consent and randomisation at preoperative clinic, automated perioperative electronic health record (EHR) monitoring, and a primary outcome of 90-day hospital-free days. Patient privacy is maintained and protected by the embedded application functioning behind the institutional firewall. (B) The SPRY-Metformin Domain-Specific Appendix functions within the infrastructure of the SPRY Core Protocol. Prior to preoperative clinic, the SPRY-Application screens the scheduled preoperative clinic appointments and generates a list of potential patients for enrolling clinicians. In preoperative clinic recruitment, informed consent and randomisation are completed. Patients undergo baseline testing. Study drug exposure begins and continues through postoperative day 90 (green). The SPRY-Application (light blue) supports patient safety monitoring by generating EHR and email alerts, as needed. As possible, all trial aspects are embedded within the standard of care perioperative course. When 500 patients surpass postoperative day 90, a priori interim analysis is completed. Future enrolment is then guided by the predetermined response adaptive randomisation schemes and predetermined stopping rules. HIPAA, Health Insurance Portability and Accountability Act; POD, postoperative day; REMAP, randomised embedded multifactorial adaptive platform; SPRY, Strategies to Promote ResiliencY.

We provided details herein on the first SPRY Core Protocol Domain (SPRY-Metformin) (see online supplemental appendix 2). Other new domains will be added to the Core Protocol as emerging therapies become available. The TSC will consider the scientific validity of each domain, safety of concurrent therapies and current enrolment rates when deciding to introduce a new domain concurrently or following existing domains. A new domain is introduced as a Domain-Specific Appendix to the SPRY Core Protocol. This Domain-Specific Appendix will be generated outlining potential interactions between multiple domains within the primary statistical analysis of efficacy, if deemed clinically appropriate. If multiple domains have been introduced, response adaptive randomisation will be based on best-performing combinations of therapies within the multiple domains and incorporate potential interactions.

SPRY-Metformin Domain-Specific Appendix

SPRY-Metformin is a multihospital, single healthcare system, placebo-controlled, adaptive, phase III clinical trial that is blinded at the level of the patient, clinician, research team and data analyst. SPRY-Metformin is the first domain to be launched on the SPRY Core Protocol testing the effectiveness of metformin in improving perioperative outcomes (figure 1B). Patients are screened and recruited from preoperative clinic through a custom application communicating with EHR data (figure 2). Study drug is started following randomisation and continued throughout the perioperative period through postoperative day (POD) 90 (figure 3). All patients are prospectively monitored through POD 365 with both automated EHR data collection and longitudinal patient follow-up (see online supplemental appendix 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Virtual and in-person screening and randomisation. ^a<7 or >180 preoperative days. ^bCharlson Comorbidity Index (CCI) required within the 365 days prior to screening. Virtual recruitment is completed by Strategies to Promote ResiliencY (SPRY)-Application (light blue) reviewing a subset of SPRY and SPRY-Metformin enrolment criteria. The SPRY-Application then guides the clinical provider to complete the in-person screening and informed consent. Any discrepancies found between the clinical parameters within SPRY-Application and the patient’s reported health state are manually updated within the electronic health record (EHR) and patients are randomised. eGFR, estimated glomerular filtration rate.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

SPRY-Metformin timeline. ^aIf patients are discharged on the day of the surgical intervention, laboratory sample 4 will be omitted. If hospital discharge occurs prior to postoperative day 3, laboratory sample 4 occur immediately prior to discharge. ^bLongitudinal testing at contact point 6 testing is dependent on participation in the motor subgroup (table 1). Patients are recruited, consented by providers, randomised, undergo baseline venous blood sampling and are provided study drug at preoperative clinic (contact point 1). In the 7–180 preoperative days, patients undergo baseline testing (table 1), and both patient safety and study drug compliance are monitored via phone interview (contact point 2). Three venous blood samples are coupled with clinical blood draws throughout the operative hospital admission (contact point 3). A final venous sample is collected in standard of care postoperative clinic (contact point 4). At postoperative days 30 and 90, patients are contacted to monitor both patient safety and study drug compliance, collect postoperative outcomes (box 2) and complete additional outcome testing (table 1).

Patient and public involvement

Patients were not invited to comment on the study design or result interpretation for the SPRY Core Protocol or SPRY-Metformin.

Digital embedding

We developed Java-based (Oracle Corporation, Redwood Shores, California, USA) custom software, the SPRY-Application, which interfaces with the research team and EHR data. The digital embedding of the SPRY-Application serves multiple purposes. First, protecting the privacy of trial patients. Second, automating patient screening, enrolment and randomisation while synchronising research activities within perioperative standard of care clinical encounters. Third, accessing the robust EHR data generated as a part of routine patient care.

At UPMC, a two-factor authentication system safeguards all private patient information accessed through a single Citrix Workspace (Fort Lauderdale, Florida, USA) in accordance with Health Insurance Portability and Accountability Act. Like all protected data and programmes within the healthcare system, the SPRY-Application resides behind this institutional firewall. Here, the SPRY-Application is distinct from, but communicates with, EHR data. The SPRY-Application accesses the clinical research data repository within UPMC Clinical Analytics and is managed by Biostatistical and Data Management Core in the Department of Critical Care Medicine at UPMC. The data repository abstracts structured, raw data from the inpatient (CERNER, Kansas City, Missouri, USA) and outpatient (Epic Systems, Madison, Wisconsin, USA) EHR and generates accessible data tables. The data extraction process parallels the methodology used traditionally for retrospective EHR data collection and research25–27; however, these data are updated in real time.

Potential trial participant identification begins with the SPRY-Application screening. The SPRY-Application reviews SPRY-specific, EHR data tables for each patient with a scheduled appointment at enrolling preoperative SPRY-Metformin clinics (figure 3). The EHR of each scheduled patient is reviewed, generating a list of patients meeting a subset of inclusion and exclusion criteria. This list of potential SPRY-Metformin candidates is then automatically distributed to the study team and clinicians via institutional email for review.

In preoperative clinics, patients are offered the opportunity to participate in SPRY-Metformin. The SPRY-Application guides the clinician through the stepwise informed consent process (see online supplemental appendix 4). Then, pertinent clinical biorepository EHR data auto-populates screening information within the SPRY-Application for review and confirmation with the patient (figure 4). Any identified discrepancies between patient report and the EHR auto-populated SPRY-Application data prompt the clinician to update the EHR (figure 1A). This both minimises trial data entry and maintains the accuracy of the EHR.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Randomised, embedded, multifactorial, adaptative platform (REMAP) Strategies to Promote ResiliencY (SPRY) administrative organisation. The Trial Steering Committee receives trial updates from the Statistical Monitoring Committee as well as recommendations from the Data and Safety Monitoring Board to oversee all trial conduct.

Patients meeting all inclusion and no exclusion criteria are allocated to a treatment regimen based on the established randomisation tables uploaded to the SPRY-Application. Automatically, the SPRY-Application then generates the study drug and laboratory prescription and synchronises all research activities (eg, blood and stool samples) within pertinent, scheduled perioperative standard of care clinical encounters. Throughout patient enrolment, the SPRY-Application monitors for biorepository updates to the Cerner Admission-Discharge-Transfer tables and informs the research team of hospital admissions and discharges for enrolled patients. Inevitable in-trial schedule changes can be manually updated within the SPRY-Application user interface and therefore adjusts the research activity timeline, updating research personnel and distributing additional study drug, as needed, via the mail.

Cultural embedding

SPRY-Metformin is designed with the intent to rely heavily on bedside clinicians for many aspects of trial execution. Healthcare system staff within high volume surgical clinics are busy with existing patient care responsibilities. We have attempted to minimise the burden of research in two ways. First, whenever possible, the protocol is fused within existing care activities. Second, we focus on engaging, educating and motivating the entire clinical team.

For example, as each new site is identified, prior to site initiation, the research team informs the clinical team about the potential benefits of a REMAP trial design and a self-learning healthcare system. Simultaneously, the clinical team educates the research team on their patients’ experiences and the clinic or unit-specific workflow. Both teams generate clinic or unit-specific protocols and SPRY-Application user manuals.

Metformin rationale

We hypothesise that pharmacological perioperative optimisation will improve surgical outcomes for an aged, frail patient population. Metformin, the most commonly prescribed non-insulin medication for type 2 diabetics, is one such therapy.15–17 In multiple studies, metformin has an excellent safety profile, is well tolerated and consistently delays the ageing process and minimises deleterious cellular inflammation5 through effects on cellular respiration,28 muscle function29 and the microbiome.30 Metformin advantageously modulates the body’s response to physical stress through its systemic anti-inflammatory properties31 32 and appear to be independent of blood glucose control.32

Study drug

Both the duration and the dose of study drug exposure will be evaluated. Patients are stratified based on the anticipated perioperative duration: short (7–28 days), intermediate (29–90 days) or long (91–180 days). Within each duration window, patients are randomised to one of three doses of metformin extended release 500, 1000 or 1500 mg, or placebo. Study drug is initiated the day following randomisation and continued through postoperative day 90 without planned interruption perioperatively (figure 2).

EHR embedded safety alerts

Surgical stress can cause fluctuation in organ function perioperatively. As a part of routine clinical care, patients at the greatest risk of physiological derailment and significant postoperative complications are admitted for postoperative monitoring. In real time, the SPRY-Application generates ‘pop-up’ style inpatient EHR alerts prompting the bedside nurse to hold study drug administration in the setting of current (ie, estimated glomerular filtration rate <45 or serum lactate ≥4) or potential future (ie, ordered contrasted imaging studies) organ dysfunction (figure 2). Simultaneously, the SPRY-Application generates an institutional email notifying the research team.

Patient sample biorepository

An additional long-term goal of SPRY-Metformin is to understand the molecular mechanisms by which metformin might attenuate the inflammatory response and improve outcomes after surgical stress.

Simulations and sample size generation

Clinical trial simulations are used to optimise clinical trial design (best thresholds for early success, dose dropping and futility stopping) to determine the sample size needed within this trial to obtain at least 80% power for a clinically meaningful treatment effect and a one-sided 2.5% type I error under the null distributions, and to quantify additional operating characteristics of the SPRY-Metformin trial. Using pertinent retrospective UPMC EHR data, virtual patient datasets were created based on the observed distributions of the primary endpoint, 90-day HFD, within each stratum. Clinical trial simulations randomised patients to study drug and numerous trials were virtually executed, including all interim analysis and randomisation adaptations. For simulated patients randomised to placebo, we assumed the primary outcome to be distributed similar to the observed 90-day HFD distribution per surgical strata within the UPMC EHR data. For simulated patients randomised to metformin, the distributions of 90-day HFD within UPMC HER data per surgical strata were shifted toward higher values of 90-day HFD being more likely based on a common percent reduction in 90-day hospital days (90 (90-day HFD)). For examples of how the distributions were shifted, see online supplemental appendix 2, figure 4.1 and table 4.2. The minimum clinically meaningful effect size was assumed to be a common percent decrease of 15% in 90-day mean hospital days for the highest treatment dose. This was chosen because it is sensitive to absolute differences in hospital days and treatments may have a larger absolute benefit for those procedures that are expected to result in more hospital days (see online supplemental appendix 2, table 4.2).34

Trial behaviours, such as power and type I error, were summarised as the proportion of simulated trials that were successful under the alternative and null scenarios, respectively. Patients will therefore be adaptively randomised to placebo or three doses of metformin, for a maximum sample size of 2500 patients enrolled. The trial has at least 84% power to detect a treatment effect of at least a 15% reduction in mean hospital days for a minimum of one of the doses under the assumption that the dose has an equally effective percent reduction in mean hospital days across all three preoperative metformin durations. If a dose is not effective for the short preoperative duration, the trial has at least 77% power to detect a treatment effect of at least a 15% reduction in mean hospital days for at least one of the doses. Under the assumption that no doses are effective, there is an overall one-sided type I error of 2.5%.

The motor subgroup will enrol up to one-third of SPRY-Metformin trial patients. The microbiome and muscle biopsy subgroups are exploratory pilot substudies with 1000 and 200 patients to be enrolled.

Response adaptative randomisation and interim analysis

Initially, SPRY-Metformin will randomise a maximum of 2000 patients. Within each of the three preoperative durations, patients will initially be randomised to placebo and the three doses of metformin. Interim analysis is completed each time 500 patients have been randomised across all preoperative durations and followed up for 90 POD. At each interim analysis, the trial can be stopped early for demonstrating efficacy, response adaptive randomisation will be adjusted to preferentially randomise patients to the best-performing treatment group or dose(s) can be dropped for futility.

Analysis plan

The primary analysis method of 90-day HFD within SPRY-Metformin is a Bayesian ordinal logistic regression model that accounts for differences in the expected 90-day HFD distribution depending on surgical strata and allows for borrowing of information across pre-op durations and doses of metformin.47 Within this model, the effect of each dose of metformin for each preoperative duration relative to placebo is characterised as a constant log-OR shift in the 90-day HFD distribution. The primary intention-to-treat analysis will include those who have been randomised. All missing data will be imputed based on the median observed 90-day HFD value for each treatment arm and preoperative duration. Sensitivity analysis will explore a per protocol analysis and alternative imputation strategies that do not make missing at random assumptions. Exploratory analyses will investigate the heterogeneity of treatment effects across key patient subgroups including patient age and frailty as well as operative stress and surgical strata.

Superiority of a metformin dose to placebo within SPRY-Metformin is determined based on the posterior probability that the pooled log-OR effect of that dose across all enrolling preoperative durations relative to placebo is less than 0, indicating a shift in the 90-day HFD distribution toward more HFD under treatment compared with placebo. Success is declared at an interim or at the final analysis if the posterior probability of superiority for any dose of metformin is greater than the predefined interim-specific success threshold. The thresholds are based on an O’Brien Fleming spending function assuming a maximum sample size of 2500.48

SPRY-Metformin secondary outcomes will be analysed using regression models that account for expected differences in surgical strata of the patient.

Additional domains and additional interventions within domains will be added to the SPRY Core Protocol. Treatment effects and treatment-by-treatment interactions can be added for each additional perioperative therapy.