How an adverse event becomes trial data
An adverse event becomes trial data through collection, clinical assessment and coding. The resulting tables count affected participants, while additional records preserve the circumstances, severity and course of individual events.
Collection and clinical assessment
An electronic case report form (eCRF) records medical events observed by an investigator or reported by a participant, regardless of suspected relationship to study drugs or procedures. Events that start or worsen after trial entry are recorded as adverse events, while pre-existing conditions remain baseline or medical history. Collection can include screening after consent and before dosing; treatment-emergent events occur after study medication starts, with a residual-effect period after the last dose accounting for delayed elimination.
An invented trial case illustrates the process: Judy, aged 51 with recently diagnosed type 2 diabetes, entered a double-blind comparison of Wonderdrug and metformin, fainted three and a half weeks after starting treatment, sprained her ankle in the fall and received hospital treatment for low blood glucose. At follow-up, the investigator elicited the circumstances of the faint and ankle injury and noticed a rash the participant had forgotten to mention. Tiredness that predated enrolment remained medical history rather than a new adverse event.
Each event carries separate assessments of seriousness, intensity and possible relationship to treatment. Typical intensity definitions distinguish mild symptoms that are easily tolerated without interference or treatment, moderate symptoms that may interfere with activities but improve with simple measures, and severe symptoms that incapacitate the participant and prevent normal activities. Intensity and seriousness are separate assessments: in the example the ankle sprain was moderate in intensity but serious because it required an overnight hospital stay.
The example's eCRF contained three events, with the following fields:
| Verbatim event | Start | End | Intensity | Drug-related | Therapy | Action with trial medication | Serious | Outcome |
|---|---|---|---|---|---|---|---|---|
| Rash on left lower arm | 19 April 2017 | — | Mild | Possibly | None | None | No | Ongoing |
| Fainting due to hypoglycaemia and standing up quickly | 22 April 2017 | 22 April 2017 | Moderate | Unlikely | Glucagon injection; fruit juice | None | No | Resolved |
| Sprained ankle | 22 April 2017 | — | Moderate | No | Ibuprofen 400 mg three times daily; ice packs | None | Yes | Ongoing |
Sponsors may prespecify adverse events of special interest (AESIs) because of known or suspected drug risks, risks associated with the study population, related drugs or study objectives. The protocol defines them and investigators collect more detail than for other events. Hypoglycaemia was an AESI in the example, with a dedicated eCRF page for accompanying symptoms, assistance required and glucose at the event, completed using the hospital report.
From individual records to participant counts
The illustrative records used a participant number linked to randomised treatment rather than the patient's name, while participant and investigator remained blinded. After unblinding and aggregation, a safety table showed the numbers and percentages of participants with each adverse event by MedDRA preferred term (PT) and system organ class (SOC). The standard summary counted each participant once per PT even if the event recurred, and once per SOC even if several different PTs in that SOC occurred.
The invented study's table for the kinds of event Judy experienced was:
| SOC or PT | Wonderdrug, N = 652, n (%) | Comparator, N = 326, n (%) |
|---|---|---|
| Metabolism and nutrition disorders | 156 (23.9) | 63 (19.3) |
| Hypoglycaemia | 122 (18.7) | 46 (14.1) |
| Skin and subcutaneous tissue disorders | 9 (1.4) | 6 (1.8) |
| Rash | 3 (0.5) | 2 (0.6) |
| Injury, poisoning and procedural complications | 7 (1.1) | 5 (1.5) |
| Joint injury | 2 (0.3) | 0 |
Hypoglycaemia was the notable difference among these events: 18.7% versus 14.1%, with a slightly higher proportion also affected in its SOC. Such a difference does not by itself establish drug causation, because factors other than medication affect hypoglycaemia.
Recurrence, timing and additional AESI data
Participant counts estimate how many people experience an event, but not the number of episodes each experiences or when those episodes occur. Every recurrence is recorded in the database even though the standard participant summary counts that person only once. Additional AESI data can therefore describe severity, action taken, recurrence and onset.
The invented trial's additional hypoglycaemia data were:
| Characteristic | Wonderdrug, N = 652 | Comparator, N = 326 |
|---|---|---|
| Participants with hypoglycaemic event, n (%) | 132 (20.2) | 49 (15.0) |
| Worst episode required assistance, n (%) | 5 (0.8) | 6 (1.8) |
| Worst episode symptomatic, glucose <3.0 mmol/L, n (%) | 19 (2.9) | 24 (7.4) |
| Worst episode symptomatic, glucose ≥3.0 and <3.9 mmol/L, n (%) | 108 (16.6) | 19 (5.8) |
| Therapy given, n (%) | 19 (2.9) | 28 (8.6) |
| Study medication discontinued, n (%) | 3 (0.5) | 4 (1.2) |
| Hospitalisation required, n (%) | 1 (0.2) | 2 (0.6) |
| One episode, n (%) | 105 (16.1) | 38 (11.7) |
| Two or three episodes, n (%) | 18 (2.8) | 9 (2.8) |
| Four or five episodes, n (%) | 3 (0.5) | 1 (0.3) |
| More than five episodes, n (%) | 6 (0.9) | 1 (0.3) |
| First episode ≤7 days, n/N at risk (%) | 11/652 (1.7) | 2/326 (0.6) |
| First episode >7 to ≤28 days, n/N at risk (%) | 57/638 (8.9) | 14/324 (4.3) |
| First episode >28 to ≤56 days, n/N at risk (%) | 39/572 (6.8) | 21/307 (6.8) |
| First episode >56 days, n/N at risk (%) | 25/541 (4.6) | 12/279 (4.3) |
| Exposure-adjusted incidence rate per 100 patient-years | 56.2 | 54.7 |
Query-level hypoglycaemia counts exceeded those for the single hypoglycaemia PT because the query included related terms. Although hypoglycaemia was more frequent with Wonderdrug, lower proportions had episodes requiring assistance, symptomatic glucose below 3.0 mmol/L, therapy, discontinuation or hospitalisation. These findings did not identify one treatment as clearly safer for hypoglycaemia.
Very few participants in either group had more than three episodes, while first episodes appeared more likely with Wonderdrug during the first four weeks and new-case rates were similar thereafter. The similar exposure-adjusted rates suggested that some of the crude-incidence difference might reflect different exposure durations, potentially from earlier comparator discontinuations.
What a single trial can establish
Adverse-event analysis seeks to identify adverse drug reactions so that benefit and risk can be weighed. Trials are usually sized for a primary efficacy endpoint rather than differences in individual safety events. Testing hundreds of adverse events separately creates multiple-testing problems and chance findings, so accurate descriptions of notable group differences do not amount to firm causal conclusions from a single trial.
An investigator's “drug-related” assessment means that causation cannot reasonably be excluded, taking account of timing, mechanism and the known safety profile. Early in development that profile is incomplete, and blinding prevents investigators from knowing the treatment received; events labelled drug-related can therefore occur even in a placebo arm. Assessment of serious events and discontinuations uses all available information beyond that initial relatedness judgement.