What selective collection changes
For a sponsor planning a large post-approval outcomes trial, the first decision is which observations remain necessary for the trial objectives and safety evaluation. E19 permits consideration of reduced collection of non-serious adverse events, laboratory monitoring, ECGs and imaging, physical examinations and vital signs, and changes in concomitant therapies after baseline. It does not make an approved product automatically eligible. The safety evidence must support the proposed population, dose, duration and implementation, and sponsors should obtain prospective regulatory agreement. See sections 1.3, 2.2, 2.5 and 2.7 of the final ICH E19 guideline.
Selective collection concerns what investigators record in case report forms (CRFs) and report to sponsors. Investigators still monitor and care for participants according to prevailing standards of care and document adverse events in medical records. Applicable local and regional reporting duties continue. A data-management instruction to omit a routine CRF item cannot override a clinical need, a protocol requirement, or a retained safety-event investigation.
FDA describes its E19 guidance as nonbinding recommendations unless specific statutes or regulations are cited. Therefore, distinguish recommendations in the guideline from duties established by applicable law and the authorized protocol. This article provides an operational interpretation and a hypothetical planning example, rather than a product-specific regulatory determination. See the introductory status statement in FDA E19.
Establish eligibility from the actual fourteen factors
E19 applies primarily to post-approval interventional trials. In rare cases, sufficient safety data from completed trials may justify selective collection in phase 3 before approval. Gene-therapy and rare/orphan-disease trials are outside its scope. Section 2.2 lists fourteen contributing factors; the presence or absence of any one is not determinative. It does not prescribe a universal patient count or a pass/fail score.
The following table groups the actual factors for review. It is an editorial summary of ICH E19 section 2.2, not a substitute for the guideline or a new validation standard. Record the evidence and remaining uncertainty for each factor, rather than counting checked boxes.
| Factors in E19 section 2.2 | Evidence to review |
|---|---|
| 1. Regulatory status | Whether the product has marketing authorization. Approval is one factor, not the entire justification. |
| 2–3. Mechanism and class knowledge | Understanding of the mechanism and off-target effects, and knowledge of safety for drugs in the same pharmacological class. |
| 4–6. Clinical safety database | Number exposed, consistency of findings across comprehensively assessed trials, and intensity of prior monitoring, including its frequency and duration. |
| 7–9. Dose, duration and product | Whether planned dose and dosing frequency exceed prior experience; whether previous exposure supports the proposed duration; comparability of product and route. |
| 10. Population similarity | Demographics, illnesses, concomitant therapies and other susceptibility factors in the planned population compared with prior trials. |
| 11–12. Clinical pharmacology | Characterization of drug interactions, metabolism and excretion. |
| 13–14. Nonclinical and post-approval evidence | Characterized nonclinical toxicology and the quantity and quality of post-approval safety information. |
Population comparability deserves an explicit decision. Section 2.2 says selective collection would not be feasible if the proposed population is more susceptible to adverse drug effects than the population used to characterize safety. This is a comparative judgment, not a blanket exclusion of every older participant or participant with organ impairment. Section 3.2 separately describes comprehensive collection in subsets where additional information matters. Section 2.6 also warns that comprehensive collection may still be warranted when moving from a severe disease setting, such as late-stage cancer or heart failure, to a less severe one, such as migraine or hypertension. See ICH E19 sections 2.2, 2.6 and 3.2.
Define the data that should generally remain
Section 2.4 identifies six types of data that should generally be collected. Preserve their distinctions in the protocol and CRF specifications: a non-serious event may still be an event of special interest or lead to study-drug discontinuation. A rule that captures only serious events would miss these other retained categories.
| Data type | What the collection specification should resolve |
|---|---|
| Serious adverse events | Event capture and supporting information, with reporting handled under applicable requirements and any prospectively agreed endpoint arrangements. |
| Important medical events | Recognition of events that may jeopardize the participant or require intervention to prevent a serious outcome, even without immediate hospitalization. |
| Medication error/overdose | Capture of medication errors and intentional or unintentional overdose. |
| Adverse events leading to study-drug discontinuation | Recording the event that led to discontinuation. Any additional capture of temporary interruptions or dose reductions should be specified separately. |
| Pregnancy and lactation exposure and outcomes | Definition of the exposure and outcome information and associated follow-up in the study plan. |
| Adverse events of special interest | Protocol-defined events critical to safety evaluation, including relevant laboratory abnormalities; these may be serious or non-serious. |
The list is explicitly non-exhaustive. Section 2.4 also allows consideration of selective collection of some listed data where regional or local rules permit it and the protocol adequately justifies it. Such a reduction requires a specific rationale and agreement; it cannot be inferred from the existence of E19. Under section 2.5, relevant medical records, laboratory data and follow-up may be needed to characterize retained events even when those data are otherwise outside routine CRF collection. See ICH E19 sections 2.4–2.5.
Baseline data considerations remain determined by the trial objectives under section 2.3. Baseline information supports enrollment eligibility and subgroup evaluation of efficacy and safety. E19 does not impose the same laboratory panel, ECG or examination schedule on every trial; it leaves those study-specific considerations unchanged.
For serious adverse events serving as efficacy or safety endpoints, arrangements under which events would not be subject to unblinding and expedited reporting should be agreed prospectively with regulators and described in the protocol. The events should be collected and monitored regularly through an independent data monitoring committee (IDMC). Endpoint status alone creates no reporting exemption. E19 says such events could be adjudicated; it does not mandate an endpoint committee for every trial. See ICH E19 sections 2.4 and 3.1.
Reduce routine collection while retaining event exceptions
| Section 2.5 category | Potential reduction when justified | Boundary to preserve |
|---|---|---|
| Non-serious adverse events | Omit routine collection or reduce its frequency. | Retain protocol-required events, including non-serious AESIs and events leading to drug discontinuation. |
| Laboratory monitoring, ECGs and imaging | Omit particular routine assessments or reduce their frequency. | Preserve measures needed for care, endpoints, targeted safety questions and investigation of retained events. |
| Physical examinations and vital signs | Reduce or omit routine post-baseline collection. | Preserve clinically needed assessment and protocol-defined measures. |
| Changes in concomitant therapies | Reduce collection of changes after baseline medication use is documented. | Collect relevant therapy information when needed to characterize retained events or meet other protocol objectives. |
The blood-glucose example in section 2.1 illustrates the distinction. When labeling recommends glucose monitoring for a drug with a characterized safety profile, clinical monitoring continues. CRF recording or sponsor reporting may be omitted within the agreed approach, but glucose levels and hypoglycemia are recorded when stipulated by the protocol, deemed clinically relevant, or associated with serious adverse events. The exception is broader than hypoglycemic coma. See ICH E19 section 2.1.
Translate this into a site instruction that names both the omitted routine item and the circumstances that restore capture. Avoid an instruction such as “do not enter mild events” without the AESI, discontinuation, clinical-relevance and supporting-information checks. A blank field should have an understandable collection context rather than being treated as evidence that no event occurred.
Choose a collection model sites can execute
E19 section 3 offers four flexible approaches. CRFs should be designed for the chosen approach and investigators appropriately trained. The operational questions in this table are planning prompts derived from those approaches, not additional regulatory requirements. See ICH E19 section 3.
| Approach | Operational question | Analysis consequence |
|---|---|---|
| Selective collection for all participants | Which reduced data types and retained exceptions apply throughout the trial? | Frequency summaries still need to describe the collection method. |
| Comprehensive collection in specific subsets | Which participant characteristics determine comprehensive collection, and how will sites identify the subset? | Keep the ascertainment differences visible in summaries. |
| Comprehensive collection initially, then selective collection | What prespecified period or transition determines the change, and which items continue? | Identify the different collection periods and their interpretation. |
| Comprehensive collection in a representative subset | How are sites or participants selected, and how are selection bias and inadequate diversity avoided? | Describe the selection method and collection approach; representativeness must be supported. |
E19 describes random selection of sites or participants as examples for the representative-subset approach. It does not prescribe a universal subset percentage or guarantee unbiased estimates simply because random selection is used. Selection, follow-up and the resulting population still need review.
Worked plan for a hypothetical post-approval outcomes trial
Consider a hypothetical outcomes trial of an approved drug in a population comparable to prior comprehensively monitored trials. Assume the sponsor has assembled evidence across all fourteen factors and is seeking regional agreement. These assumptions illustrate a planning process; they do not establish eligibility for any actual product. The proposed model is selective collection throughout, with targeted measures retained for known safety concerns and endpoint evaluation.
The plan below is an author-developed operational synthesis of E19 sections 2–4. Its proposed owners and escalation arrangements are illustrative. It deliberately supplies no laboratory cutoffs, universal reporting deadlines, sample-size thresholds or enrollment rules; those require study-specific justification.
| Decision | Proposed collection rule | Exception or action | Responsible function in this example |
|---|---|---|---|
| Eligibility rationale | Document prior safety knowledge and its applicability to the planned dose, exposure and population. | Reassess if enrollment, product use or new evidence undermines comparability. | Clinical development and safety leads, with regional regulatory review. |
| Baseline | Retain baseline data required by the objectives and eligibility criteria. | Resolve incomplete required assessments under the actual protocol; E19 creates no automatic enrollment-invalidity rule. | Investigator and clinical operations. |
| Retained event categories | Generally retain all six section 2.4 categories in the agreed specification. | Obtain relevant records, laboratory information and follow-up to characterize an event; document any permitted, justified regional reduction. | Investigator, safety and data management. |
| Routine non-serious events | Omit routine CRF entry outside the retained categories. | Capture a non-serious AESI or event leading to study-drug discontinuation. | Investigator, using trained CRF rules. |
| Routine tests and medication changes | Reduce specified routine items while keeping targeted safety and endpoint measures. | Record clinically relevant findings and information needed to characterize retained events. Clinical care and source documentation continue. | Clinical lead and data management. |
| Endpoint SAEs | Collect endpoint events through the agreed process. | Apply special unblinding/reporting arrangements only when prospectively agreed; retain regular IDMC review. | Safety, endpoint operations and IDMC. |
| Regional differences | Maintain an explicit record of accepted collection methods for each region. | Update forms, training and analysis plans if an authority requires additional collection. | Regulatory lead and clinical operations. |
| Emerging concern | Route new concerns for prompt safety review. | Consider targeted intensified monitoring or comprehensive collection, then document decisions and implement necessary changes. | Safety lead, IDMC within its remit, and sponsor decision makers. |
Before launch, walk site staff through cases that cross the routine-data boundary: a minor event that causes discontinuation, an otherwise routine laboratory result needed to investigate an AESI, and an SAE that is also an endpoint. Check that the configured forms can retain the event and its supporting information. This is a practical way to test the specification, rather than an E19-mandated validation exercise.
Plan escalation and preserve the collection history
Section 4 recognizes that emerging concerns may require intensified safety monitoring or reversion to comprehensive collection, with logistical consequences for sites, protocols and forms. It does not supply universal numeric triggers. The following are proposed planning scenarios to tailor to the product, monitoring arrangements and applicable requirements. See ICH E19 section 4.
A new product or class safety concern: the safety lead reviews whether the omitted data would help characterize it and whether targeted collection or broader reversion is appropriate.
A clinically concerning pattern identified during committee review: the IDMC communicates through its established remit, and the sponsor evaluates participant protection and changes to collection.
A change in the enrolled population or exposure: clinical and safety leads reassess whether the original characterization evidence still supports selective collection.
A regional requirement for additional data: the regulatory and operations teams determine the authorized scope and coordinate protocol, form, training and analysis changes.
For this hypothetical plan, record the decision maker, rationale, affected data types and participants, and the effective collection period. Coordinate urgent participant protection with applicable processes for implementing study changes. Data management should preserve which collection regime applied to each participant and period so analysis teams can distinguish an intentional omission from a failure to collect a required item.
Reversion cannot recreate observations that were never collected. E19 specifically identifies retrospective questions about concomitant medications, laboratory parameters and blood pressure as a disadvantage. A historical value may sometimes exist in medical records, but its availability and suitability cannot be assumed merely because comprehensive CRF collection resumes.
Keep ascertainment differences visible in the analysis
Under section 4, when collection methods differ across participants or over time, data obtained under comprehensive and selective approaches cannot be pooled. Methods for aggregation should be specified in the protocol and associated analysis plans, and summaries should identify where collection affects interpretation. Uniform selective collection still needs a description of its scope. See ICH E19 section 4.
A hypothetical arithmetic example shows the problem. Suppose a comprehensively assessed subset has 20 participants with a routine non-serious event among 200 assessed participants. Another 800 participants have no collection of that event type. The observed proportion in the assessed subset is 20/200, or 10%. Reporting 20/1,000, or 2%, as a trial-wide event frequency would incorrectly treat uncollected events as absent. These numbers are illustrative, not trial findings.
The statistical analysis plan (SAP) should identify collection populations and periods, explain denominators and planned summaries, and distinguish items required but missing from items deliberately outside scope. Separate the safety-frequency question from the efficacy endpoint question: reducing unrelated routine safety data does not establish that endpoint follow-up is complete or settle how treatment discontinuation is handled.
For the wider endpoint specification, use the site’s guide to statistical analysis plans. Keep the selective-collection rules traceable to the protocol, forms and reporting descriptions.
Check regional implementation before finalizing the plan
ICH finalized E19 on 27 September 2022. The implementation dates below identify guidance status, not automatic acceptance of a particular trial. Regional requirements and prospective agreement remain relevant.
| Jurisdiction | Verified implementation status | Planning implication |
|---|---|---|
| United States | FDA final E19 Guidance for Industry, December 2022. | Recommendations are nonbinding except where statutes or regulations are cited; discuss the actual design with the relevant review team. |
| European Union | EMA Step 5, EMA/782210/2022; the page lists a legal effective date of 16 March 2023. | This is the guideline implementation date, not evidence that every selective design is accepted across all participating countries. |
| Canada | Implementation notice dated 21 December 2023, revised 27 February 2024. | Review the eligible-trial criteria and interim enforcement-discretion provisions together with E19. |
Status sources: FDA guidance page; EMA current-version page; Health Canada implementation notice.
Health Canada requires a well-understood and documented safety profile and excludes gene therapy and rare disease. Its trial criteria cover phase IV, late-stage trials outside authorized conditions of use and, rarely, phase III before authorization. The notice describes interim discretion on enforcement of retaining all adverse-event records for eligible authorized protocols. It preserves regulatory safety reporting, allows inspection of medical-care records and reserves enforcement if actual or potential health risks are identified. This is not an amendment of the regulations. See the Health Canada notice.
E19 section 2.7 recommends obtaining scientific advice before initiation in regions where results will support regulatory purposes. A single selective-collection protocol across regions depends on agreement from all relevant authorities. Resolve differences in collection scope and their operational consequences before assuming one set of CRFs will suffice.
Use trial precedents with their limits
FDA’s C3TI white paper reports that VICTORION-2-PREVENT implemented selective collection at more than 1,100 sites across 40 countries; countries that did not accept the approach did not participate. Its example includes retention of all SAEs, discontinuation events and AESIs, with limited liver assessments in a subset. In EMPACT-MI, eight of 22 countries requested local amendments, such as follow-up creatinine or all-AE collection. These are examples of negotiated scope, not interchangeable templates. See the C3TI white paper.
Pragmatica-Lung provides a different illustration. Its registry defines the high-grade adverse-event secondary outcome as grade 3 or higher unexpected treatment-related events determined by the treating physician, with overall survival as the primary outcome. An outcome definition does not establish every protocol-level safety reporting requirement. Use the registry to understand the registered measure, rather than treating it as a complete CRF or pharmacovigilance specification. See NCT05633602 registered outcomes.
The C3TI paper is a discussion document, not draft or final guidance. Boehringer Ingelheim and Novartis reviewed relevant sections about their programs. Its selected examples illuminate implementation choices but do not measure comparative safety or prove that reduced collection is equally suitable in another trial. FDA’s separate demonstration-project page describes enhanced sponsor–CDER communication to explore appropriate use; participation is not a guarantee of regulatory acceptance. See the SSDC Demonstration Project.
Yamatani and colleagues identified selective-collection features in 44 of 459 medicinal-product trials published in the New England Journal of Medicine from February 2016 through December 2019: 44 divided by 459 is approximately 9.6%. This publication-limited, historical sample predates final E19 and cannot estimate uptake after 2022 or establish why sponsors chose their collection methods. See Yamatani et al., 2022.
The C3TI paper also describes comprehensive collection in the tocilizumab giant-cell-arteritis program despite prior safety experience in 4,009 patients and 14,994 patient-years, followed by approval without safety-label modifications. That observation supports asking what additional collection will contribute. It does not establish that the collected data had no clinical value or that selective collection would necessarily have produced the same benefit–risk assessment. See the C3TI discussion of the tocilizumab example.
Keep participant refusal on its own decision path
Selective collection is a scientifically justified protocol property for specified data, populations or periods. Individual refusal or withdrawal concerns what a particular participant has agreed to provide. E19 does not authorize disregarding that refusal or resolve consent and retention questions. Handle the participant’s actual choices through the consent and withdrawal process, and distinguish a protocol-defined omission from a required observation unavailable after refusal. The separate guide to withdrawal and retained trial data addresses that boundary.