Concept Architecture
Concept
Theoretically, Sequential Analysis is a statistical methodology in which data are evaluated repeatedly as they are collected, allowing a study to stop early when sufficient evidence has accumulated. It is founded on sequential probability theory and decision theory and differs from conventional fixed-sample analysis by permitting interim decisions regarding continuation, modification or termination of a study. In health economics, sequential analysis is used in adaptive clinical trials, evidence generation and value of information studies to improve research efficiency while maintaining statistical validity.
Mathematically, sequential analysis evaluates a predefined test statistic against stopping boundaries after each interim analysis or recruitment stage. The boundaries are constructed to control the overall Type I and Type II error rates despite repeated testing. Common approaches include the Sequential Probability Ratio Test, group sequential methods and alpha-spending functions.
In practice, sequential analysis is implemented by specifying interim analysis schedules, stopping rules and statistical boundaries before data collection begins. Health economists use sequential analysis to support adaptive trial designs, optimise evidence generation and reduce unnecessary research costs when convincing evidence is obtained before the planned maximum sample size.
Purpose
Used to evaluate accumulating evidence during data collection, enabling efficient study designs that may stop early while preserving statistical validity.
Mathematical Formulae
Primary Formula
Sequential Probability Ratio:
?? = L(H?) � L(H?)
where:
?? = sequential likelihood ratio after n observations
L(H?) = likelihood under the alternative hypothesis
L(H?) = likelihood under the null hypothesis
Decision rules:
Accept H? if ?? � A
Accept H? if ?? � B
Continue sampling if B < ?? < A
Supporting Formulae
Upper boundary:
A = (1 ? ?) � �
Lower boundary:
B = ? � (1 ? �)
where:
� = Type I error probability
? = Type II error probability
Related Mathematical Methods
- Sequential Probability Ratio Test
- Group Sequential Design
- Adaptive Clinical Trial
- Interim Analysis
- Bayesian Decision Theory
- Optimal Sample Size
- Value of Information Analysis
Example
A clinical trial evaluating a new oncology treatment plans interim analyses after every 100 participants. Following the third interim analysis, the observed treatment benefit exceeds the predefined efficacy boundary while maintaining the overall Type I error rate. Recruitment is stopped early because sufficient evidence has accumulated, reducing research costs and allowing earlier implementation of the intervention.
Excel Implementation
| Function | Example Formula | Health Economics Application |
|---|---|---|
| LN | =LN(Likelihood_H1/Likelihood_H0) | Calculate the logarithm of the sequential likelihood ratio. |
| IF | =IF(B2>=UpperBoundary,"Stop for efficacy",IF(B2<=LowerBoundary,"Stop for futility","Continue")) | Apply sequential stopping rules. |
| SUM | =SUM(DataRange) | Update cumulative evidence at each interim analysis. |
| COUNT | =COUNT(DataRange) | Track the accumulated sample size throughout the trial. |
VBA (Optional)
VBA can automate interim analyses, evaluate stopping boundaries and update sequential decision rules during adaptive clinical trials.
Sources
- Wald A. Sequential Analysis. Wiley.
- Jennison C, Turnbull BW. Group Sequential Methods with Applications to Clinical Trials. Chapman & Hall/CRC.
- Whitehead J. The Design and Analysis of Sequential Clinical Trials. Wiley.
- Briggs A, Claxton K, Sculpher M. Decision Modelling for Health Economic Evaluation. Oxford University Press.
- NICE. NICE Health Technology Evaluations: The Manual.
Related Concepts (2)
Library
Publications
1
Value of Information Analytical Methods: Report 2 of the ISPOR Value of Information Analysis Emerging Good Practices Task Force — Rothery, Strong, Koffijberg, Basu, Ghabri, Knies, Murray, Sanders Schmidler, Steuten & Fenwick, Vol. 23, No. 3 ed., 2020 (Value in Health)
The methods companion to the ISPOR VOI series, giving detailed algorithms and software guidance for computing EVPI, EVPPI, EVSI and the expected net benefit of sampling, with recommendations for selecting methods by decision-problem features.
Journal ArticleView source →
Frequently Asked Questions (6)
What is sequential analysis?
A statistical approach evaluating data as it accumulates, allowing early stopping once a predefined level of evidence is reached, rather than a fixed sample size.
Source: Wald 1947
What advantage does sequential analysis offer?
By examining the data as they accumulate rather than only at a preset final sample size, sequential analysis can stop a study as soon as the evidence is decisive, whether for benefit, harm, or futility. This can spare patients from a treatment already shown to be worse and save the cost of continuing beyond the point of a clear answer. The saving in time and resources is its main advantage. It requires statistical adjustment, since repeated looks at the data inflate the chance of a false positive if uncorrected. Whitehead (1997) describes this approach.
Source: Whitehead 1997
How does sequential analysis work?
Sequential analysis works by defining stopping rules in advance and assessing the accumulating data against them after each observation or interim look: if the evidence crosses a boundary favouring one conclusion, sampling stops with that conclusion; otherwise it continues. Wald's sequential probability ratio test, for example, compares the likelihood of the data under competing hypotheses and stops when the ratio crosses preset limits. The boundaries are set to control the error rates, so the procedure balances early stopping against the risk of a wrong conclusion.
Source: Wald 1947
Why is sequential analysis used?
Sequential analysis is used because it can reach a conclusion with fewer observations than a fixed-sample design when the evidence is strong, saving time and resources and, in trials, exposing fewer participants than necessary. By monitoring data as it accumulates and stopping once enough evidence is gathered, it avoids collecting more data than needed. This efficiency is valuable where observations are costly or where early stopping is ethically or practically desirable, so sequential analysis suits settings in which data arrive over time and prompt decisions are wanted.
Source: Ades, Lu & Claxton 2004
How does sequential analysis differ from fixed-sample analysis?
Sequential analysis evaluates data as it accumulates and can stop early once a stopping rule is met, so the sample size is not fixed in advance, whereas fixed-sample analysis collects a predetermined number of observations and analyses them once at the end. Sequential designs can conclude sooner when the effect is clear but require preset stopping boundaries that control error rates, since repeated looks otherwise inflate error. So the approaches differ in whether the sample size is fixed and in how repeated interim assessment is handled.
Source: Wald 1947
What are the limitations of sequential analysis?
Sequential analysis requires stopping boundaries designed in advance to control error rates, since repeatedly testing accumulating data without adjustment inflates the chance of a false positive, and its design and analysis are more complex than fixed-sample methods. Estimates from studies stopped early can be biased, tending to overstate effects, and the maximum sample size may be uncertain at the outset. These limitations mean sequential analysis is applied with properly constructed stopping rules and careful interpretation of results from early stopping, weighing its efficiency against its added complexity.
Source: Wald 1947
Trust Record
Verified by Dr Darrin Baines
British health economist
Professional identity: darrinbaines.org
Verification date: 31 Oct 2025
Content version: 1.0.0
Canonical Identity
- Persistent URI
- https://healtheconomics.wiki/concept/sequential-analysis
- Term code
- HE-EM-VI-001
Stable URI · Machine-readable · Resolvable · CC BY 4.0