GLP-1 trial data capture: lessons for small sponsors

Whatever your protocol requires for a valid weight, such as a calibrated scale, light clothing, time of day or fasting status, should be a field or a completion instruction, not an assumption. A dropdown for the scale used and a simple fasted yes/no cost a second at the site and give the statistician something to check if a site's results look different. Where weights are also taken at home, capture a separate flag so the two sources can be told apart.

Capture calculates BMI from height and weight as a read-only field during entry, and offers other calculated fields such as BSA and waist-hip ratio. The principle applies to any system: staff enter measured values, the system derives the result, and nobody re-keys a calculator output. Add a dedicated height field at screening and decide whether it is re-measured at later visits. Our vital signs eCRF template and DEXA body composition eCRF show how these forms are laid out.

If a site can enter either kilograms or pounds, store the unit with the value or fix the unit on the form and say so in the field label. Add range-high and range-low edit checks at plausible limits so that a typo (91.4 entered as 914 or 9.14) raises an auto-query as the value is entered, not at the monitoring visit. Keep the limits wide enough that real values never trigger them, since a query on every entry trains staff to ignore queries.

Do not rely on a single "treatment compliance" yes/no. Give dosing its own form with the planned dose, the actual dose, the date and time, and a repeating row per administration if the protocol needs that level of detail. When a dose is missed, delayed or changed, add a reason. Self-administered injections are a special case: decide in advance whether the record is site-observed, participant-reported or from a diary, and mark it.

An ePRO diary for tolerability, appetite or injection-site reactions is only interpretable if you can place each entry relative to dosing. Put the date of the last dose in the same data model (or at least in an exportable dataset joined by participant and date), keep the number of questions small so completion stays high and use the same wording across visits. Capture's ePRO offers ten question types and participants complete questionnaires from a link without installing an app. See GI symptoms diary examples and our guide to digital patient diaries, and aim to read compliance by week as your study runs (ePRO compliance and completion rates).

Participants stop, pause, switch dose or begin other treatments. How those events are handled in the analysis is a statistical decision, discussed in the ICH E9(R1) addendum on estimands, and your statistician owns it. Your job in the EDC is to make sure the raw facts exist: a date, a reason from a defined list and a link to any new medication. Retrofitting that from free text at database lock is slow and unreliable.

Site-measured, participant-reported and device-derived values behave differently. If weights arrive from a connected scale or from a wearable, keep that as a separate field or flag rather than blending it into the site column. Capture connects wearables by native OAuth sync, with the participant connecting their own account, and the FDA's final guidance on digital health technologies for remote data acquisition (22 December 2023) is the document to read on fitness for purpose and usability. See EDC for wearable device studies and the FDA DHT guidance explainer.

Capture's edit checks are range-high, range-low and custom value rules on a field, raising an auto-query when a value breaks a rule. A large drop between two visits is a different question, and a monitor or data manager review listing is a more honest tool than an automatic query that might reflect real change. Decide before go-live which anomalies get a rule and which get a recurring review. See edit checks software.

Ask your statistician what they need: wide or long format, a data dictionary, SDTM domains. Capture exports CSV and Excel with an automatic data dictionary, and SDTM datasets as SAS XPT files with Define-XML. If your protocol follows the ICH M11 structured template, which reached Step 4 on 19 November 2025, aligning your visit schedule and forms to it makes the build easier to audit. Our ICH M11 guide explains the template.