A randomised trial of Study Drug Y added to metformin in adults with type 2 diabetes in Vietnam
Feasibility and design report · Vietnam · prepared for [SPONSOR]
- Recommended design
- Where the plan comes from
- Sample size and power under local behaviour
- Who the eligibility criteria excluded in the ingested study
- Enrolment and screen-failure forecast
- Retention and adherence to expect
- Timeline to first participant and last visit
- Risks and the limits of this report
- Evidence table
1. Recommended design
Run design B: 524 (370 to 750, 90% sensitivity range) Assumed E1 participants, randomised one to one, at 6 (as stated) Assumed E2 central and 5 (as stated) Assumed E3 provincial hospitals.
Under the local profile, expect full enrolment at 17.4 months (13.6 to 22.7, 80% spread across simulated trials) Assumed E4 after approval and the last visit at 29.4 months (25.6 to 34.7, 80% spread across simulated trials) Assumed E5, with power of 87.4% (85.2% to 89.3%, 95% Monte Carlo) Assumed E6 for an assumed difference in HbA1c change of −0.40 % HbA1c (−0.55 to −0.25, 90% published or brief range) Assumed E7.
If the true effect sits at the weak end of the planning range, −0.25 % HbA1c (as stated) Assumed E8, power falls to 53.2% (50.1% to 56.3%, 95% Monte Carlo) Assumed E9.
With no true effect, the design rejects at a rate of 5.1% (4.7% to 5.5%, 95% Monte Carlo) Assumed E10.
2. Where the plan comes from
| Fact | Value |
|---|---|
| Allocation | one to one |
| SD of HbA1c change | 1.00 % HbA1c (0.80 to 1.30, 90% assumed range, no source yet) Assumed E11 |
| Effect to plan for | −0.40 % HbA1c (−0.55 to −0.25, 90% published or brief range) Assumed E7 |
The effect to plan for is a planning assumption that no cited trial backs yet.
3. Sample size and power under local behaviour
Design A is sized the usual way, on global defaults and published figures. Design B is sized on the local profile. Both were run as simulated trials so the difference local inputs make is visible.
| Design and inputs | Enrolled | Power | Months to full enrolment | Participants retained |
|---|---|---|---|---|
| A, global defaults | 330 (214 to 552, 90% sensitivity range) Assumed E12 | 88.2% (86.1% to 90.1%, 95% Monte Carlo) Assumed E13 | 14.8 months (10.7 to 20.8, 80% spread across simulated trials) Assumed E14 | 80.3% (72.4% to 87.0%, 80% spread across simulated trials) Assumed E15 |
| A, local profile | 330 (214 to 552, 90% sensitivity range) Assumed E12 | 73.7% (70.9% to 76.3%, 95% Monte Carlo) Assumed E16 | 13.6 months (10.5 to 18.1, 80% spread across simulated trials) Assumed E17 | 80.0% (73.3% to 85.5%, 80% spread across simulated trials) Assumed E18 |
| B, local profile | 524 (370 to 750, 90% sensitivity range) Assumed E1 | 87.4% (85.2% to 89.3%, 95% Monte Carlo) Assumed E6 | 17.4 months (13.6 to 22.7, 80% spread across simulated trials) Assumed E4 | 79.8% (73.7% to 84.9%, 80% spread across simulated trials) Assumed E19 |
Design B was sized to reach the brief's target power at the central value of each input. Simulated power averages over the uncertainty in every input, so it sits below that target; a sponsor who needs the target with more certainty should enrol toward the top of the sensitivity range.
The inputs that moved most: the SD of HbA1c change is 1.26 % HbA1c (1.06 to 1.51, 90% prediction for a new study) Learned E20 locally against 1.00 % HbA1c (0.80 to 1.30, 90% assumed range, no source yet) Assumed E11 by default, and screen failure is 35.3% (23.1% to 49.8%, 90% prediction for a new study) Learned E21 against 36.3% (25.0% to 50.0%, 90% published or brief range) Assumed E22.
4. Who the eligibility criteria excluded in the ingested study
| Criterion | Text | Share of screened people excluded |
|---|---|---|
| C1 | HbA1c between 7.5% and 10.5% at screening protocol C1 | 15.7% (9.2% to 25.5%, 90% prediction for a new study) Learned E23 |
| C2 | eGFR below 45 mL/min/1.73 m² protocol C2 | 8.2% (4.5% to 14.5%, 90% prediction for a new study) Learned E24 |
| C3 | Written informed consent protocol C3 | 8.0% (4.4% to 14.0%, 90% prediction for a new study) Learned E25 |
| C4 | Other conditions listed in Appendix 2 protocol C4 | 3.4% (1.7% to 6.8%, 90% prediction for a new study) Learned E26 |
Overall screen failure: 35.3% (23.1% to 49.8%, 90% prediction for a new study) Learned E21.
5. Enrolment and screen-failure forecast
Screened per site per month, all sites: 7.27 screened per site per month (4.41 to 12.0, 90% prediction for a new study) Learned E27, against a global default of 6.00 screened per site per month (2.00 to 12.0, 90% assumed range, no source yet) Assumed E28.
Values by site type come from fewer than three sites each and would identify those hospitals, so this report uses values from all sites together.
Back-test on the ingested study: fitted on records up to the day half the target was enrolled, the model predicted 11.0 months (9.59 to 12.7, 80% spread across simulated trials) Learned E29 to full enrolment. It took 10.2 months (observed, exact) Learned E30.
6. Retention and adherence to expect
| Measure | All sites |
|---|---|
| Monthly dropout rate | 0.0184 per month (0.0108 to 0.0315, 90% prediction for a new study) Learned E31 |
| Interim visits attended | 87.5% (79.4% to 92.7%, 90% prediction for a new study) Learned E32 |
Expect 279 (206 to 372, 80% spread across simulated trials) Assumed E33 protocol deviations over the study and 0.51 per participant (0.37 to 0.67, 80% spread across simulated trials) Assumed E34 missed interim visits per participant.
| Other local measure | Value |
|---|---|
| Tablets taken as prescribed | 81.9% (71.7% to 88.9%, 90% prediction for a new study) Learned E35 |
| Days from visit to data entry, provincial hospitals | 7 days (5 to 11, 90% prediction for a new study) Learned E36 |
| Days to close a data query, provincial hospitals | 8 days (5 to 13, 90% prediction for a new study) Learned E37 |
| Patients also using traditional medicine | 31.0% (20.0% to 45.0%, 90% elicited range) Elicited E38 |
7. Timeline to first participant and last visit
Approval and ethics: 120 days (80 to 200, 90% elicited range) Elicited E39. Site activation: 220 days (120 to 365, 90% published or brief range) Assumed E40. That value is a global default because no local study shows it, and it is the weakest input in this plan.
From approval: first participant at 4.28 months (2.85 to 6.60, 80% spread across simulated trials) Assumed E41, full enrolment at 17.4 months (13.6 to 22.7, 80% spread across simulated trials) Assumed E4, last visit at 29.4 months (25.6 to 34.7, 80% spread across simulated trials) Assumed E5.
8. Risks and the limits of this report
| Theme | What reviewers raised |
|---|---|
| Consent | The consent form should explain secondary use of the data in plain language. ethics_letter_national.en.md#C-EN-02 |
| Safety reporting | Hypoglycaemia was rare and mild. debrief_notes.en.md#C-DB-05 |
| Visit schedule | Patients from rural districts missed interim visits during the harvest season. debrief_notes.en.md#C-DB-03 |
| Data entry and queries | Coordinators entered data in batches at the end of each month. debrief_notes.en.md#C-DB-04 |
| Recruitment | Enrolment was slower than planned because many referred patients had HbA1c below the eligibility range. debrief_notes.en.md#C-DB-01 |
One study informs this report. It cannot show how much studies differ, so each learned value's interval for a new study is widened by a between-study spread that clinicians estimated. The treatment effect is the brief's assumption, not a prediction of whether the drug works.
Replay: the profile was re-fitted on the first two-thirds of the ingested study by calendar date and the last third predicted before it was revealed. Dropouts: predicted 18 (13 to 25, 80% spread across simulated trials) Learned E42, observed 13 (observed, exact) Learned E43. Interim visits attended: predicted 212 (202 to 222, 80% spread across simulated trials) Learned E44, observed 215 (observed, exact) Learned E45.
9. Evidence table
| Ref | Number | Value | Interval | Spread from | Source | Based on | Provenance |
|---|---|---|---|---|---|---|---|
| E1 | Total enrolment, design B | 524 | 370 to 750, 90% sensitivity range | none | Assumed | target power and alpha from the brief; range from the outcome assumption's interval | analytic sample size (continuous) · outcome from profile v1 · bp-outcome-change-sd · dropout from local profile v1 |
| E2 | Central hospital sites | 6 | as stated | none | Assumed | the sponsor's brief | brief · sites.central_hospital |
| E3 | Provincial hospital sites | 5 | as stated | none | Assumed | the sponsor's brief | brief · sites.provincial_hospital |
| E4 | Months to full enrolment, design B, local profile | 17.4 months | 13.6 to 22.7, 80% spread across simulated trials | none | Assumed | 1,000 simulated trials, local profile v1 | run-c2c8f49fe438 · seed 7 · profile v1 · code 0.1.0+g089e218.src.52797a48 |
| E5 | Months to last visit, design B, local profile | 29.4 months | 25.6 to 34.7, 80% spread across simulated trials | none | Assumed | 1,000 simulated trials, local profile v1 | run-c2c8f49fe438 · seed 7 · profile v1 · code 0.1.0+g089e218.src.52797a48 |
| E6 | Power, design B, local profile | 87.4% | 85.2% to 89.3%, 95% Monte Carlo | none | Assumed | 1,000 simulated trials, local profile v1 | run-c2c8f49fe438 · seed 7 · profile v1 · code 0.1.0+g089e218.src.52797a48 |
| E7 | Assumed difference in HbA1c change | −0.40 % HbA1c | −0.55 to −0.25, 90% published or brief range | none | Assumed | Synthetic. Stands in for a published trial of the same drug class; F3.2 requires a real citation. Not yet checked against its source. | brief · assumed_effect |
| E8 | Assumed difference in HbA1c change, weak end of the range | −0.25 % HbA1c | as stated | none | Assumed | the end of the brief's planning range nearest to no effect | brief · assumed_effect range |
| E9 | Power, design B, effect at the weak end | 53.2% | 50.1% to 56.3%, 95% Monte Carlo | none | Assumed | 1,000 simulated trials, local profile v1 | run-a31d7f0e67eb · seed 7 · profile v1 · code 0.1.0+g089e218.src.52797a48 |
| E10 | Type I error, design B, no true effect | 5.1% | 4.7% to 5.5%, 95% Monte Carlo | none | Assumed | 10,000 simulated trials, local profile v1 | run-f8b301d10a6a · seed 7 · profile v1 · code 0.1.0+g089e218.src.52797a48 |
| E11 | SD of HbA1c change, global default or published | 1.00 % HbA1c | 0.80 to 1.30, 90% assumed range, no source yet | none | Assumed | no published source yet | global default · outcome_change_sd |
| E12 | Total enrolment, design A | 330 | 214 to 552, 90% sensitivity range | none | Assumed | target power and alpha from the brief; range from the outcome assumption's interval | analytic sample size (continuous) · outcome from global default · outcome_change_sd · dropout from global defaults |
| E13 | Power, design A, global defaults | 88.2% | 86.1% to 90.1%, 95% Monte Carlo | none | Assumed | 1,000 simulated trials, global defaults | run-e50f82e6f538 · seed 7 · profile global-defaults · code 0.1.0+g089e218.src.52797a48 |
| E14 | Months to full enrolment, design A, global defaults | 14.8 months | 10.7 to 20.8, 80% spread across simulated trials | none | Assumed | 1,000 simulated trials, global defaults | run-e50f82e6f538 · seed 7 · profile global-defaults · code 0.1.0+g089e218.src.52797a48 |
| E15 | Participants retained, design A, global defaults | 80.3% | 72.4% to 87.0%, 80% spread across simulated trials | none | Assumed | 1,000 simulated trials, global defaults | run-e50f82e6f538 · seed 7 · profile global-defaults · code 0.1.0+g089e218.src.52797a48 |
| E16 | Power, design A, local profile | 73.7% | 70.9% to 76.3%, 95% Monte Carlo | none | Assumed | 1,000 simulated trials, local profile v1 | run-3d0da899393e · seed 7 · profile v1 · code 0.1.0+g089e218.src.52797a48 |
| E17 | Months to full enrolment, design A, local profile | 13.6 months | 10.5 to 18.1, 80% spread across simulated trials | none | Assumed | 1,000 simulated trials, local profile v1 | run-3d0da899393e · seed 7 · profile v1 · code 0.1.0+g089e218.src.52797a48 |
| E18 | Participants retained, design A, local profile | 80.0% | 73.3% to 85.5%, 80% spread across simulated trials | none | Assumed | 1,000 simulated trials, local profile v1 | run-3d0da899393e · seed 7 · profile v1 · code 0.1.0+g089e218.src.52797a48 |
| E19 | Participants retained, design B, local profile | 79.8% | 73.7% to 84.9%, 80% spread across simulated trials | none | Assumed | 1,000 simulated trials, local profile v1 | run-c2c8f49fe438 · seed 7 · profile v1 · code 0.1.0+g089e218.src.52797a48 |
| E20 | SD of HbA1c change, local | 1.26 % HbA1c | 1.06 to 1.51, 90% prediction for a new study | elicited | Learned | 241 patients, 3 sites | profile v1 · bp-outcome-change-sd |
| E21 | Screen-failure rate, local | 35.3% | 23.1% to 49.8%, 90% prediction for a new study | elicited | Learned | 464 patients, 3 sites | profile v1 · bp-screen-failure-proportion |
| E22 | Screen-failure rate, global default | 36.3% | 25.0% to 50.0%, 90% published or brief range | none | Assumed | Tufts CSDD via Applied Clinical Trials, 'Can recruitment and retention get any worse?' (PRD Appendix A). The range is assumed. | global default · screen_failure_proportion |
| E23 | Screened out by C1 | 15.7% | 9.2% to 25.5%, 90% prediction for a new study | elicited | Learned | 464 patients, 3 sites | profile v1 · bp-screen-failure-C1 |
| E24 | Screened out by C2 | 8.2% | 4.5% to 14.5%, 90% prediction for a new study | elicited | Learned | 464 patients, 3 sites | profile v1 · bp-screen-failure-C2 |
| E25 | Screened out by C3 | 8.0% | 4.4% to 14.0%, 90% prediction for a new study | elicited | Learned | 464 patients, 3 sites | profile v1 · bp-screen-failure-C3 |
| E26 | Screened out by C4 | 3.4% | 1.7% to 6.8%, 90% prediction for a new study | elicited | Learned | 464 patients, 3 sites | profile v1 · bp-screen-failure-C4 |
| E27 | Screened per site per month, all sites | 7.27 screened per site per month | 4.41 to 12.0, 90% prediction for a new study | elicited | Learned | 464 patients, 3 sites | profile v1 · bp-screening-rate-per-site-per-month |
| E28 | Screened per site per month, global default | 6.00 screened per site per month | 2.00 to 12.0, 90% assumed range, no source yet | none | Assumed | no published source yet | global default · screening_rate_per_site_per_month |
| E29 | Months to full enrolment from mid-enrolment, predicted | 11.0 months | 9.59 to 12.7, 80% spread across simulated trials | none | Learned | fitted on data dated before the cut | export exp-71acbbf231 · backtest |
| E30 | Months to full enrolment from mid-enrolment, observed | 10.2 months | observed, exact | none | Learned | revealed after the prediction | export exp-71acbbf231 · backtest |
| E31 | Monthly dropout rate, all sites | 0.0184 per month | 0.0108 to 0.0315, 90% prediction for a new study | elicited | Learned | 300 patients, 3 sites | profile v1 · bp-dropout-hazard-per-month |
| E32 | Interim visits attended, all sites | 87.5% | 79.4% to 92.7%, 90% prediction for a new study | elicited | Learned | 293 patients, 3 sites | profile v1 · bp-visit-attendance-probability |
| E33 | Protocol deviations, whole study, design B, local profile | 279 | 206 to 372, 80% spread across simulated trials | none | Assumed | 1,000 simulated trials, local profile v1 | run-c2c8f49fe438 · seed 7 · profile v1 · code 0.1.0+g089e218.src.52797a48 |
| E34 | Interim visits missed per participant, design B, local profile | 0.51 per participant | 0.37 to 0.67, 80% spread across simulated trials | none | Assumed | 1,000 simulated trials, local profile v1 | run-c2c8f49fe438 · seed 7 · profile v1 · code 0.1.0+g089e218.src.52797a48 |
| E35 | Tablets taken as prescribed | 81.9% | 71.7% to 88.9%, 90% prediction for a new study | elicited | Learned | 293 patients, 3 sites | profile v1 · bp-adherence-proportion |
| E36 | Days from visit to data entry, provincial hospitals | 7 days | 5 to 11, 90% prediction for a new study | elicited | Learned | 300 patients, 3 sites | profile v1 · bp-data-entry-lag-days |
| E37 | Days to close a data query, provincial hospitals | 8 days | 5 to 13, 90% prediction for a new study | elicited | Learned | 207 patients, 3 sites | profile v1 · bp-query-turnaround-days |
| E38 | Patients also using traditional medicine | 31.0% | 20.0% to 45.0%, 90% elicited range | elicited | Elicited | 3 clinicians | profile v1 · bp-traditional-medicine-use-proportion |
| E39 | Days for approval and ethics | 120 days | 80 to 200, 90% elicited range | elicited | Elicited | 3 clinicians | profile v1 · bp-approval-and-ethics-days |
| E40 | Days to activate a site | 220 days | 120 to 365, 90% published or brief range | none | Assumed | 31.4 weeks to site activation, Tufts CSDD site initiation benchmark, 2018 (PRD Appendix A). The range is assumed. | global default · site_activation_days |
| E41 | Months to first participant, design B, local profile | 4.28 months | 2.85 to 6.60, 80% spread across simulated trials | none | Assumed | 1,000 simulated trials, local profile v1 | run-c2c8f49fe438 · seed 7 · profile v1 · code 0.1.0+g089e218.src.52797a48 |
| E42 | Dropouts in the last third, predicted | 18 | 13 to 25, 80% spread across simulated trials | none | Learned | fitted on data dated before the cut | export exp-71acbbf231 · replay.dropouts |
| E43 | Dropouts in the last third, observed | 13 | observed, exact | none | Learned | revealed after the prediction | export exp-71acbbf231 · replay.dropouts |
| E44 | Interim visits in the last third, predicted | 212 | 202 to 222, 80% spread across simulated trials | none | Learned | fitted on data dated before the cut | export exp-71acbbf231 · replay.visits |
| E45 | Interim visits in the last third, observed | 215 | observed, exact | none | Learned | revealed after the prediction | export exp-71acbbf231 · replay.visits |
| E46 | Protocol deviations, whole study, design A, global defaults | 184 | 109 to 313, 80% spread across simulated trials | none | Assumed | 1,000 simulated trials, global defaults | run-e50f82e6f538 · seed 7 · profile global-defaults · code 0.1.0+g089e218.src.52797a48 |
| E47 | Protocol deviations, whole study, design A, local profile | 175 | 130 to 234, 80% spread across simulated trials | none | Assumed | 1,000 simulated trials, local profile v1 | run-3d0da899393e · seed 7 · profile v1 · code 0.1.0+g089e218.src.52797a48 |
| E48 | Months to last visit, design A, global defaults | 26.8 months | 22.7 to 32.8, 80% spread across simulated trials | none | Assumed | 1,000 simulated trials, global defaults | run-e50f82e6f538 · seed 7 · profile global-defaults · code 0.1.0+g089e218.src.52797a48 |
| E49 | Months to last visit, design A, local profile | 25.6 months | 22.5 to 30.1, 80% spread across simulated trials | none | Assumed | 1,000 simulated trials, local profile v1 | run-3d0da899393e · seed 7 · profile v1 · code 0.1.0+g089e218.src.52797a48 |
| E50 | Months to first participant, design A, global defaults | 4.33 months | 2.92 to 6.73, 80% spread across simulated trials | none | Assumed | 1,000 simulated trials, global defaults | run-e50f82e6f538 · seed 7 · profile global-defaults · code 0.1.0+g089e218.src.52797a48 |
| E51 | Months to first participant, design A, local profile | 4.28 months | 2.85 to 6.54, 80% spread across simulated trials | none | Assumed | 1,000 simulated trials, local profile v1 | run-3d0da899393e · seed 7 · profile v1 · code 0.1.0+g089e218.src.52797a48 |
| E52 | Interim visits attended, global default | 92.0% | 85.0% to 97.0%, 90% assumed range, no source yet | none | Assumed | no published source yet | global default · visit_attendance_probability |
| E53 | Monthly dropout rate, global default | 0.0177 per month | 0.0088 to 0.0353, 90% published or brief range | none | Assumed | 19.1% dropout, Tufts CSDD via Applied Clinical Trials (PRD Appendix A), converted to a monthly hazard assuming 12 months of follow-up. The conversion and the range are assumed. | global default · dropout_hazard_per_month |
| E54 | Interim visits missed per participant, design A, global defaults | 0.34 per participant | 0.21 to 0.51, 80% spread across simulated trials | none | Assumed | 1,000 simulated trials, global defaults | run-e50f82e6f538 · seed 7 · profile global-defaults · code 0.1.0+g089e218.src.52797a48 |
| E55 | Interim visits missed per participant, design A, local profile | 0.51 per participant | 0.38 to 0.67, 80% spread across simulated trials | none | Assumed | 1,000 simulated trials, local profile v1 | run-3d0da899393e · seed 7 · profile v1 · code 0.1.0+g089e218.src.52797a48 |
Sign-off
Not yet signed. Export is blocked until a named person signs.