An interim analysis is a pre-planned statistical evaluation of study data conducted before the scheduled end of the trial. Its purpose is to enable decisions on whether to continue, modify, or terminate a clinical trial early based on accumulating data. Because interim analyses can substantially affect the statistical error rate, they are subject to strict methodological and regulatory requirements that must be specified in the study protocol and the Statistical Analysis Plan.
Purpose and areas of application
Interim analyses serve various purposes within a clinical trial. The most common use cases are:
- Futility analyses: Assessment of whether, based on the data to date, the study still has a realistic chance of achieving its primary endpoint. If the result is negative, the study may be terminated early for ethical and economic reasons—saving resources and protecting participants from an ineffective intervention.
- Efficacy analyses: Assessment of whether a significant treatment effect has already been demonstrated that justifies early termination in favor of the treatment group (early stopping for efficacy). This approach is particularly relevant in oncology trials with survival endpoints.
- Safety analyses: Ongoing monitoring of the safety profile, particularly for new investigational products or especially vulnerable patient populations. The Data Safety Monitoring Board performs such analyses at regular intervals.
- Sample size reassessment: Adjustment of the sample size based on observed variance or effect size without unblinding (blinded sample size re-estimation). This variant is less critical from a regulatory perspective because no unblinded between-group comparisons are performed.
The decision to conduct an interim analysis must already be specified in the study protocol and the Statistical Analysis Plan. The number, timing, and decision rules must be defined in advance and are binding. Unannounced interim analyses are not acceptable from a regulatory perspective and may lead to rejection of a marketing authorization application.
Statistical methods and alpha spending
The central issue with any interim analysis is inflation of the type I error (false-positive findings) when the same data are analyzed multiple times. Without statistical adjustment, the probability of a false-positive result increases substantially with each additional analysis. Standard methods to control the overall alpha level include:
- O’Brien–Fleming boundary: Stringent stopping boundaries for early analyses that approach the nominal alpha level as the study nears completion. This method is widely used and recognized by the EMA and the BfArM.
- Pocock boundary: A constant stopping boundary for all analyses, but a more stringent nominal significance level for the final analysis. This often results in a less intuitive final p-value threshold.
- Alpha-spending function (Lan–DeMets): A flexible approach that allocates the alpha budget continuously and also accounts for uneven information accrual. Particularly suitable for adaptive designs and studies with variable recruitment rates.
- Haybittle–Peto boundary: A very conservative early stopping boundary (p < 0.001 for all interim analyses) that minimizes alpha consumption for the final analysis.
The selected method and all stopping boundaries must be documented in the Statistical Analysis Plan before the first interim analysis is conducted. Deviations require a formal amendment.
Organizational requirements and independence
Interim analyses that could lead to unblinding may only be conducted by an independent body. The Data Safety Monitoring Board (DSMB) or Data Monitoring Committee (DMC) evaluates the results and provides recommendations to the sponsor without the actual study team having access to the unblinded data. This independence is a core requirement of ICH E6(R3) and EU Regulation No. 536/2014.
As a rule, the sponsor receives only a recommendation—continue, adapt, or stop—but not the raw data from the unblinded analysis. All responsibilities, communication pathways, and decision rules of the DSMB must be set out in writing in a DSMB charter in advance. If such a charter is missing or incomplete, this is considered a critical finding during regulatory inspections.
Relevance for clinical trials
Interim analyses are particularly important in phase III trials with long durations and in oncology trials with hard endpoints such as overall survival. They make it possible to stop trials early if patients are put at risk, or to make successful therapies available sooner. Inadequate planning—such as the absence of an alpha-spending function or an independent DSMB—may lead to rejection of a marketing authorization application. Full-service CROs such as mediconomics support sponsors with the methodologically sound planning of interim analyses, the preparation of DSMB charters, and the statistical implementation of alpha-spending functions in the Statistical Analysis Plan.
Frequently Asked Questions (FAQ)
Can an interim analysis be added to the study plan retrospectively?
No. Interim analyses must be specified in advance in the study protocol and the Statistical Analysis Plan. Retrospective additions require a formal amendment that must be approved by the responsible ethics committees and competent authorities. Unannounced interim analyses can jeopardize the approvability of the study.
What happens to the significance level after an interim analysis?
The overall alpha level (typically 0.05 for two-sided tests) is allocated between the interim analyses and the final analysis via an alpha-spending function. The nominal significance level for the final analysis is therefore lower than 0.05 if alpha has already been spent at the interim analysis. The exact allocation depends on the method selected.
Who is allowed to view the results of an unblinded interim analysis?
Only the members of the Data Safety Monitoring Board and an independent statistician. The study team, the sponsor, and the investigator must not be informed of unblinded interim results, as this would compromise study integrity. Any disclosure must be clearly defined in the DSMB charter.
Regulatory References
- ICH E9(R1) – Addendum on Estimands and Sensitivity Analysis (2019): requirements for pre-planned analyses
- ICH E6(R3) – Good Clinical Practice (2023): DSMB independence, documentation requirements
- EMA Guideline on Data Monitoring Committees (EMEA/CHMP/EWP/5872/03 Corr, 2005)
- EU Regulation No. 536/2014 (CTR): amendment requirement for protocol changes
- ICH E8(R1) – General Considerations for Clinical Studies (2021): study design and planning flexibility