Understanding Mis-Stratification in Randomized Controlled Clinical Trials

Stratified randomization is a common practice in randomized, controlled clinical trials. It ensures that key characteristics are evenly distributed across treatment groups and the treatment assignments are balanced within each randomization stratum, enhancing the validity of study results. However, during the course of a trial, mis-stratification can occur—this happens when an incorrect stratification stratum is used during randomization. Let's explore what this means, why it happens, and how it impacts clinical trials.

---

What is Mis-Stratification?

In clinical trials, stratification factors (e.g., age, disease severity, disease subgroup, or background medication use) are used to group participants before randomization. Stratified randomization is used to ensure that equal numbers of subjects with one or more characteristic(s) thought to affect the treatment outcome in efficacy measure will be allocated to each comparison group. Mis-stratification - a type of randomization errors, occurs when:

• An incorrect stratification factor is used for randomization, or
• A participant is placed in the wrong stratum due to an error.

Despite this, the treatment assignment and drug dispensation remain accurate, making it a minor deviation rather than a critical error. When mis-stratification occurs, the random code and the treatment assignment is pulled from an incorrect stratum. 

---

Historical Approach: Intention-to-Treat Principle

Traditionally, clinical trials have adhered to the Intention-to-Treat (ITT) principle, where participants are analyzed according to the group they were originally randomized to, regardless of any errors. This approach maintains the integrity of the randomization process.

In practice, this means using the original stratification data—even if incorrect—in the statistical analysis. A typical Statistical Analysis Plan (SAP) might state:

“All original stratification information used in the randomization procedure will be used for analyses, regardless of whether it was later found to be incorrect. All efficacy analyses will be performed primarily on the ITT Population.”

This approach minimizes bias and reflects the 'real-world' impact of treatment. However, in the mis-stratification situation, using the incorrect stratum information in analyses may be too harsh and too strict unnecessarily.

---

Why Does Mis-Stratification Occur?

Mis-stratification can result from several factors, including:

• Too Many Stratification Factors: More factors increase the complexity and likelihood of error.
• Local vs. Central Lab Results: Differences between local and central lab measurements can lead to misclassification.
• Timing of Measurement: Stratification factors measured at different times (baseline vs. screening) may not align.
• Medication Use: Stratifying by prior or concomitant medication use can be complicated by variations in patient reporting or prescription practices.

These issues highlight potential flaws in protocol design and study quality, emphasizing the need for clear definitions and consistent procedures.

---

Regulatory Perspective: FDA Guidance

The FDA's guidance document, “Adjusting for Covariates in Randomized Clinical Trials for Drugs and Biological Products,” provides clarity on handling mis-stratification:

“Randomization is often stratified by baseline covariates. A covariate adjustment model should generally include strata variables and can also include covariates not used for stratifying randomization. In some cases, incorrect stratification may occur and result in actual and as-randomized baseline strata variables. A covariate adjustment model can use either strata variable definition as long as this is prespecified.  “

This guidance supports the use of either the originally assigned stratification or the actual baseline data in the analysis, provided it is specified before data unblinding. This flexibility helps maintain the study's validity while addressing stratification errors transparently.

---

Impact on Study Results

Mis-stratification is generally considered a minor deviation because its impact on efficacy and safety analyses is minimal. It does not affect treatment assignment or drug dispensation but only the stratum from which the assignment was drawn.

When incorrect stratification occurs, the actual stratification information is collected in the Electronic Data Capture (EDC) system and can be used in sensitivity analyses to evaluate the robustness of the study results.

There is an article "Handling misclassified stratification variables in the analysis of randomised trials with continuous outcomes" where the authors did the simulation study to investigate the impact of the mis-stratification on the statistical analyses. 

---

Minimizing Mis-Stratification in Randomization

Too many mis-stratification errors indicate the poor quality of the clinical trial. To reduce the risk of mis-stratification, consider the following best practices:

• Limit Stratification Factors: Use the minimum necessary factors to reduce complexity.
• Consistent Measurement Timing: Align the timing of stratification factor measurements (e.g., always at baseline).
• Clear Definitions: Ensure stratification criteria are clearly defined, identified or measured, and uniformly applied.
• Training and Quality Checks: Provide thorough training for study personnel and implement rigorous quality checks.

---

Conclusion

While mis-stratification is not ideal, its impact on clinical trial results is usually minimal. By adhering to the Intention-to-Treat principle and following regulatory guidance, researchers can maintain the integrity of their analyses. As clinical trial designs become more complex, understanding and managing mis-stratification will continue to be crucial for maintaining study quality and reliability.

Survivorship Bias and its Occurrences in Clinical Trials

Survivorship bias or survival bias is the logical error of concentrating on entities that passed a selection process while overlooking those that did not. This can lead to incorrect conclusions because of incomplete data. Survivorship bias is a selection bias that occurs when an individual only considers the surviving observation without considering those data points that didn’t “survive” in the event. It can lead to incorrect or misleading conclusions.

 
Most famous example of survivorship bias came from the World War II time. During World War II, the high rate of planes being shot down posed a significant challenge. To address this, a team was tasked with improving the protection and armor of the aircraft to reduce their vulnerability. The team analyzed the planes that returned from missions, documenting the locations of damage and creating a damage map. They observed heavy damage on the wings and tail sections, leading them to reinforce these areas. However, despite these modifications, the rate of plane losses did not significantly decrease.

The breakthrough came when the team realized a critical oversight: they were only examining the planes that had successfully returned, meaning the damage they observed was survivable and not the kind that would cause a plane to be lost. This insight shifted their perspective, highlighting that the most vulnerable areas were likely those that hadn’t been hit on the returning planes—areas where damage would be fatal. By focusing on reinforcing these previously overlooked sections, the team was able to better protect the aircraft and pilots, ultimately saving lives. This story underscores the importance of considering not only the visible data but also the hidden or missing information, as what you don’t know can be just as critical as what you do know.

Survivorship Bias in Clinical Trials. 

Survivorship bias occurs when we focus only on the "survivors", "completers", "responders", or successful outcomes in a dataset while ignoring those that didn’t make it. This can lead to overly optimistic or inaccurate conclusions because the full picture isn’t being considered. In the context of clinical trials, survivorship bias can distort our understanding of a treatment’s effectiveness or safety by excluding data from participants who dropped out, didn’t respond to the treatment, or experienced adverse events.

Survivorship Bias are very common in clinical trials even though the term "survivorship bias" may not be explicitly used and may be described under "selection bias".

How Does Survivorship Bias Manifest in Clinical Trials?

1. Dropout Rates and Missing Data
   Clinical trials often experience participant dropouts due to side effects, lack of efficacy, or personal reasons. If researchers only analyze data from participants who completed the trial, they may overestimate the treatment’s effectiveness or underestimate its risks. For example, a drug might appear highly effective because only the patients who benefited from it stayed in the trial, while those who didn’t respond or experienced severe side effects left.
2. Selective Reporting
   Researchers or sponsors may inadvertently (or intentionally) focus on positive outcomes while downplaying or omitting negative results. This can create a skewed perception of a treatment’s success. For instance, if a trial reports only the patients who improved and ignores those who didn’t, the treatment may seem more promising than it actually is.
3. Long-Term Follow-Up Gaps
   Many clinical trials focus on short-term outcomes, which can miss long-term effects. Patients who experience adverse events or relapse after the trial ends may not be included in the final analysis, leading to an incomplete understanding of the treatment’s safety and efficacy.
4. Population Selection
   Clinical trials often exclude certain populations, such as older adults, pregnant women, or individuals with comorbidities. While this is sometimes necessary for safety or feasibility, it can create a biased sample that doesn’t reflect the real-world population. The "survivors" in this case are the participants who met the strict inclusion criteria, potentially limiting the generalizability of the results.

Real-World Consequences of Survivorship Bias

Survivorship bias in clinical trials can have serious implications for patients, healthcare providers, and policymakers. For example:

• Overestimation of Treatment Efficacy: If a drug appears more effective than it truly is, patients may be prescribed a treatment that doesn’t work for them, wasting time and resources.
• Underestimation of Risks: Ignoring data from participants who dropped out due to side effects can lead to an incomplete understanding of a treatment’s safety profile.
• Misguided Policy Decisions: Policymakers relying on biased trial results may approve treatments that aren’t as effective or safe as they seem, potentially putting public health at risk.
  

How Can We Address Survivorship Bias?

1. Intent-to-Treat Analysis or Treatment Policy Strategy in Estimand Framework
   One of the most effective ways to mitigate survivorship bias is to use an intent-to-treat (ITT) analysis or treatment policy strategy in handling the intercurrent event, which includes all participants who were randomized in the trial, regardless of whether they completed it. This approach provides a more realistic picture of the treatment’s effectiveness in real-world conditions.
2. Avoid performing the analyses only on completers
3. Encourage the patients with intercurrent events to remain in the study to minimize the dropouts
4. Transparency in Reporting
   Researchers should report all outcomes, including dropouts, adverse events, and negative results. Journals and regulatory agencies can encourage this by requiring comprehensive data disclosure.
5. Long-Term Follow-Up
   Extending the follow-up period can help capture long-term outcomes and provide a more complete understanding of a treatment’s benefits and risks.
6. Diverse Participant Populations
   Including a broader range of participants, such as older adults and individuals with comorbidities, can improve the generalizability of trial results and reduce bias.
7. Independent Oversight
   Independent review boards and data monitoring committees can help ensure that trials are conducted and analyzed objectively, minimizing the risk of bias.

Conclusion

Survivorship bias is a subtle but significant issue in clinical trials that can distort our understanding of medical treatments. By focusing only on the "survivors", "completers", "responders", or successful outcomes, we risk overlooking critical data that could impact patient care and public health. Addressing this bias requires a commitment to transparency, rigorous analysis, and inclusive research practices. As we continue to advance medical science, it’s essential to remember that what we see isn’t always the full picture—and that the missing pieces may hold the key to better, safer, and more effective treatments.

By being aware of survivorship bias and taking steps to mitigate it, we can ensure that clinical trials provide the most accurate and reliable evidence possible, ultimately improving outcomes for patients everywhere.

Further Reading:

• Elston (2021) Survivorship bias. Journal of the American Academy of Dermatology
• Francesco Pasqualetti (2023) The impact of survivorship bias in glioblastoma research. Critical Reviews in Oncology/Hematology
• Mark É. Czeisler (2021) Uncovering survivorship bias in longitudinal mental health surveys during the COVID-19 pandemic. Epidemiology and Psychiatric Sciences
• Conner (2019) Survivorship Bias in Analyses of Immune Checkpoint Inhibitor Trials. JAMA oncology
• Selection Bias: Selection Bias: The Art of Choosing the Right Data
• WHO (2019) WHO consolidated guidelines on drug-resistant tuberculosis treatment

Intention-to-Treat Principle versus Treatment Policy Estimand: Different Names, but Same Meaning?

ICH E9 "Statistical Principles for Clinical Trials" was finalized in February 1998. The E9 guidelines established the Intention-to-Treat principle for the design and analysis of clinical trials. With the intention-to-treatment principle, we are required to include all study participants (full analysis set) in the analyses. Here are the definitions for 'full analysis set' and 'intention-to-treat principle' from ICH E9. 

In 90's, it took a while for the people to understand and accept the concept of the intention-to-treat principle. We also see that the intention-to-treat principle was misused, over-used, or undercut by the use of practical intention-to-treat and modified intention to treat. I had a presentation (in 2004) about the misuse/overuse of intention-to-treat and modified intention-to-treat. What I said then is still applicable today. 

The strict definition of intention-to-treat can be traced back to the book chapter by Fisher, LD et al. Intention to treat in clinical trials in Statistical Issues in Drug Research and Development. Edited by Peace KE (1990). The intention-to-treat was defined as:

Includes all randomized patients in the groups to which they were randomly assigned, regardless of their adherence with the entry criteria, regardless of the treatment they actually received, and regardless of subsequent withdrawal from treatment or deviation from the protocol

The intention-to-treat principle includes all randomized subjects in the analyses and ignores what happens to the subjects after the randomization (whether or not the subject discontinued the study drug, took prohibited or rescue therapies, crossed over the alternate treatment,...), which is obviously not the best option in estimating the treatment effect in some situations.  This leads to the development of Addendum to ICH E9 "ICH E9 (R1) Estimands and Sensitivity Analysis in Clinical Trials". ICH E9 (R1) explained the issues with the intention-to-treat principle and introduced the new concept of estimands (including treatment policy estimand) and intercurrent events. 

This addendum clarifies and extends ICH E9 in respect of the following topics. Firstly, ICH E9 introduced the Intention-To-Treat (ITT) principle in connection with the effect of a treatment policy in a randomised controlled trial, whereby subjects are followed, assessed and analysed irrespective of their compliance to the planned course of treatment, indicating that preservation of randomisation provides a secure foundation for statistical tests. Multiple consequences arising from the ITT principle can be distinguished. Firstly, that the trial analysis should include all subjects relevant for the research question. Secondly, that subjects should be included in the analysis as randomised. Taken directly from the definition of the ITT principle (see ICH E9 Glossary), a third consequence is that subjects should be followed-up and assessed regardless of adherence to the planned course of treatment and that those assessments should be used in the analysis. It remains undisputed that randomisation is a cornerstone of controlled clinical trials and that analysis should aim at exploiting the advantages of randomisation to the greatest extent possible. However, the question remains whether estimating an effect in accordance with the ITT principle always represents the treatment effect of greatest relevance to regulatory and clinical decision making. The framework outlined in this addendum gives a basis for describing different treatment effects and some points to consider for the design and analysis of trials to give estimates of these treatment effects that are reliable for decision making. Secondly, issues considered generally under data handling and “missing data” (see Glossary) are re-visited. Two important distinctions are made. 

With the intention-to-treat principle, subjects who discontinued the study drug prematurely should continue to be followed up and the data after dose discontinuation should continue to be collected. However, in practice for many studies, the data collection was stopped for subjects who discontinued the study drug, or the data collected after subjects' discontinuation of study drug were collected, but not used in the analyses. To some extent, the intention-to-treat principle was not fully followed. That is why the FDA has issued its guidance "Data Retention When Subjects Withdraw from FDA-RegulatedClinical Trials" to encourage the data collection after the subjects withdraw from the study. As discussed in the guidance:

The validity of a clinical study would also be compromised by the exclusion of data collected during the study. There is long-standing concern with the removal of data, particularly when removal is non-random, a situation called “informative censoring.” FDA has long advised “intent-to-treat” analyses (analyzing data related to all subjects the investigator intended to treat), and a variety of approaches for interpretation and imputation of missing data have been developed to maintain study validity. Complete removal of data, possibly in a non-random or informative way, raises great concerns about the validity of the study. 

The addendum to ICH E9 introduced the concept of estimands and intercurrent events. Those events that occurred after the randomization were previously ignored even though the analyses were under the intention-to-treat principle. With the addendum, Those events that occurred after the randomization would be called 'intercurrent events'. Here is the official definition of the intercurrent events:

Intercurrent Events:

Events occurring after treatment initiation that affect either the interpretation or the existence of the measurements associated with the clinical question of interest. It is necessary to address intercurrent events when describing the clinical question of interest in order to precisely define the treatment effect that is to be estimated.

Estimands can be classified based on the strategies of handling the intercurrent events. One way to handle the intercurrent events is the 'treatment policy' strategy - therefore, we have a treatment policy estimand. The treatment policy estimand under the addendum is almost identical to the intention-to-treatment principle under the original ICH E9. 

Treatment policy strategy

The occurrence of the intercurrent event is considered irrelevant in defining the treatment effect of interest: the value for the variable of interest is used regardless of whether or not the intercurrent event occurs. For example, when specifying how to address use of additional medication as an intercurrent event, the values of the variable of interest are used whether or not the patient takes additional medication.

If applied in relation to whether or not a patient continues treatment, and whether or not a patient experiences changes in other treatments (e.g. background or concomitant treatments), the intercurrent event is considered to be part of the treatments being compared. In that case, this reflects the comparison described in the ICH E9 Glossary (under ITT Principle) as the effect of a treatment policy.

The intention-to-treat and treatment policy estimand are two different names with the same meaning. If we have to differentiate them, we can say that the intention-to-treatment principle is more focused on which subjects should be included in the analyses while the treatment policy estimand is more focused on which data points should be included in the analyses. If a randomized subject has an intercurrent event (for example, discontinued the study treatment), the subject is still included in the intention-to-treatment population for analysis, but will the measures after the subject's discontinuation of the study treatment be included in the analyses? With the treatment policy estimand, these measures after the subject's discontinuation of the study treatment will need to be included in the analyses. 

Here is a thread discussing the difference between the Intention-to-treat principle and the treatment policy estimand in resident360.nejm.com. 

We have started to see that the ICH E9 addendum and the concept of estimands are gradually adopted, especially in EU countries. The adoption of the ICH E9 in the US is much slower than in EU countries. The concept of estimands and intercurrent events is still considered as the words invented by statisticians. It will take a while for non-statisticians to understand the concept and to accept these new terms. A presentation "Regulator’s experience with estimands" by Andreas Brandt from EMA summarized the challenges for the adoption and implementation of the ICH E9 Addendum. We will anticipate the difficulties ahead for non-statisticians and clinicians to accept the concept of estimand and intercurrent events. This is reflected in a paper by Min & Bain "Estimands in diabetes clinical trials"

During 2019 several type 2 diabetes trials results using the term estimand were published. This word will be unfamiliar to many clinicians (and to spellcheck) but given that regulatory bodies have endorsed its use, this word is likely to become a staple of medical jargon in the future.

ICH E9 Addendum described five different strategies for handling the intercurrent events: treatment policy strategy, hypothetical strategy, composite variable strategy, while on treatment strategy, and principle stratum strategy. However, in practice, the treatment policy estimand is used the vast majority of the studies where the estimand concept is mentioned. There are a few studies using the principle stratum strategy. The other three strategies (hypothetical strategy, composite variable strategy, while on treatment strategy) are rarely used in practice perhaps because they are relatively new, are uncertain with the regulatory acceptance, and because there is no available method to estimate the treatment difference for some estimands.  

If the vast majority of the estimand application is treatment policy strategy which is almost identical to the traditional intention-to-treat principle, we will question if it is worth revamping the entire ICH E9 to come up with an addendum for estimand and intercurrent event concept.  

Should We Follow ICH E9 Addendum to Include the Estimands in all Clinical Trial Protocols?

ICH E9 "Statistical Principles for Clinical Trials" was issued in 1998 - more than 20 years ago. While the principles specified in ICH E9 are still being followed, a call for a revision or addendum has been there for many years. In 2017, the draft version of ICH E9 (R1) “Addendum on Estimands and Sensitivity Analysis in Clinical Trials to the Guideline on Statistical Principles for Clinical Trials” was released and at the end of 2019, ICH E9 (R1) was finalized. The E9 (R1) guidelines are now gradually been adopted by various regulatory agencies. In terms of the implementation, EMA seems to be ahead of the US requiring the sponsors to include the concept of Estimands in the regulatory submissions. 

Purpose and scope of the addendum to ICH E9:

• Provides a framework for describing with precision a treatment effect of interest
• Precision in describing a treatment effect of interest is facilitated by constructing the “estimand”
• Estimand: A precise description of the treatment effect reflecting the clinical question posed by the trial objective. It summarises at a population-level what the outcomes would be in the same patients under different treatment conditions being compared
• Clarity requires a thoughtful envisioning of “intercurrent events” such as discontinuation of assigned treatment, use of additional or alternative treatment, and terminal events such as death
• Intercurrent Events: Events occurring after treatment initiation that affect either the interpretation of the existence of the measurements associated with the clinical question of interest
• It is necessary to address intercurrent events when describing the clinical question of interest in order to precisely define the treatment effect that is to be estimated
• Addendum introduces strategies to reflect different questions of interest that might be posed
• Attributes used to construct the estimand are also introduced in the addendum
• Addendum clarifies the definition and the role of sensitivity analysis
• Sensitivity Analysis: A series of analyses conducted with the intent to explore the robustness of inferences from the main estimator to deviations from its underlying modeling assumptions and limitations in the data

Estimand attributes:

• Treatment: The treatment condition of interest and, as appropriate, the alternative treatment condition to which comparison will be made
• Population: Patients targeted by the clinical question
• Variable (or endpoint): Obtained for each patient and required to address the clinical question
• Population-level summary: Provides a basis for comparison between treatment conditions for the variable
• Handling of intercurrent events

While FDA has not mandated the implementation of the ICH E9 (R1), there seems to be a trend in the industry that ICH E9 (R1) is gradually being adopted and the concept ‘estimands’ is being mentioned in the study protocols and statistical analysis plans (SAPs). 

Looking at the clinical trial protocol and SAP templates developed by TransCelerate Biopharma, both contained a section about estimands and the estimands was listed together with endpoints.

In my previous post "Should Clinical Trial Protocol be Made Public While the Trial is still ongoing?", I mentioned that the protocols for three phase III studies for Covid-19 vaccine were all made public because of the demand for transparency. I examed all three protocols and they had described the 'estimands'. The concept of 'intercurrent events', 'principal stratum strategy', 'treatment policy' was also mentioned. 

In Phase III study protocol by Moderna "A Phase 3, Randomized, Stratified, Observer-Blind, Placebo-Controlled Study to Evaluate the Efficacy, Safety, and Immunogenicity of mRNA-1273 SARS-CoV-2 Vaccine in Adults Aged 18 Years and Older":

In the estimand of the primary analysis on the primary endpoint, a treatment policy strategy will be used to address the intercurrent events of 1) withdrawal from the study or death unrelated to COVID-19, where the time to COVID-19 will be censored at the date of withdrawal from the study or death; 2) early COVID-19, where the time to COVID-19 will be censored at the time of early infection. Principal stratum strategy will be used to address the other 2 types of intercurrent events in the primary analysis based on the PP Set. The details of intercurrent event description and estimand strategies are presented in Section 11.4.1.

In Phase III study protocol by AstraZeneca / Oxford "A Phase III Randomized, Double-blind, Placebo-controlled Multicenter Study in Adults to Determine the Safety, Efficacy, and Immunogenicity of AZD1222, a Non-replicating ChAdOx1 VectorVaccine, for the Prevention of COVID-19":

The primary estimand will be used for the analysis of the primary efficacy endpoint. It will be based on participants in the full analysis set, defined as all randomized participants who received at least 1 dose of study intervention excluding those participants who are seropositive at baseline, analyzed according to their randomized treatment. For participants with multiple events, only the first occurrence will be used for the primary efficacy endpoint analysis. The set of intercurrent events for this estimand consists of participants who withdraw from the study prior to having met the primary efficacy endpoint. The intercurrent events will be handled using the treatment policy strategy and the absence of data following these participants’ withdrawal will be treated as missing (ie, counted as not having met the criteria). Participants who withdraw before 15 days post second dose or who have a case prior to 15 days post second dose will be excluded from primary endpoint analysis.
Additional estimands will be specified for the primary efficacy endpoint to carry out sensitivity analyses for assessing the robustness of results. These sensitivity analyses will explore different methods for handling intercurrent events and different assumptions for missing data. Estimands will also be specified for the analysis of secondary endpoints. Full details will be provided in the SAP.

In Phase III study protocol by Pfizer/BioNtech "A Phase 1/2/3, Placebo-Controlled, Randomized, Observer-Blind, Dose-Finding Study to Evaluate the Safety, Tolerability, Immunogenicity, and Efficacy of SARS-COV-2 RNA Vaccine Candidates Against COVID-19 in Healthy Individuals": 

It looks like that the concept of 'estimands' has not been widely accepted by the medical community - it is indeed a new concept and a new term for clinical trialists to digest. Using a most recent paper in the New England Journal of Medicine (Rabe et al 2020 "Triple Inhaled Therapy at Two Glucocorticoid
Doses in Moderate-to-Very-Severe COPD"), the main body of the article had no mention of the concept of 'estimands' even though the attached protocol and SAP contained a section about 'estimands':

In the phase III study by J&J (Janssen) "A Randomized, Double-blind, Placebo-controlled Phase 3 Study to Assess the Efficacy and Safety of Ad26.COV2.S for the Prevention of SARS-CoV-2-mediated COVID-19 in Adults Aged 18 Years and Older", there was no mention of estimands. 

Additional References: 

• US FDA and Health Canada Joint Regional Consultation on the International Council for Harmonisation (ICH)
• University of Pennsylvania 11th annual conference on statistical issues in clinical trials: Estimands, missing data and sensitivity analysis (morning panel session)
• Estimand Framework: What it is and Why You Need it
• Estimands in clinical trials – broadening the perspective
• PSI, EFSPI & ASA BIOP Webinar: Estimands
• Estimand Framework in Oncology Drug Development: Challenges and Opportunities

Are we become slaves of the Intent to Treat Principle?

The intent to treat (or intention to treat) principle was invented by the statistician about 30 years ago. It took a while for the clinical trial community to accept the this concept. Nowadays, the intent to treat principle has been well accepted by the people well beyond the statisticians. However, I don't think everybody really understand the concept even though he (or she) may mention the intent to treat pricinple every time he (or she) can. I have been really bothered by the comments from regulatory reviewers to suggest us to define an intent to treatment population for studies without randomization and without placebo or active control (for example, a dose escalation study). We seem to become slaves of the intent-to-treat.

In a lot of situations, the intent to treat principle is misunderstood. The intent-to-treat concept is tied with randomization for treatment allocation. No randomization, no Intent-to-treat.
The intent-to-treat concept is really for the large scale, confirmatory, pivitol studies. For very ealier stage studies (for example, the dose escalation studies) with very few subjects, there is no need to follow the intent-to-treat principle.

Intent to treat population includes all randomized patients in the groups to which they were randomly assigned, regardless of their adherence with the entry criteria, regardless of the treatment they actually received, and regardless of subsequent withdrawal from treatment or deviation from the protocol. Stricly according to the intent to treat principle, if a subject is randomized, but never receive study medication, the subject would be included in the statistical analysis; if a subject is randomized to drug A, but wrontly takes the drug B, the subject would be analyzed in treatment group A, not B (so called as randomized, not as treated); if a subject is randomized, but with no outcome measures, the subject would be included in the analysis with subject considered as treatment failure.

Intention to treat analyses are done to avoid the effects of crossover and drop-out, which may break the randomization to the treatment groups in a study. Intention to treat analysis provides information about the potential effects of treatment policy rather than on the potential effects of specific treatment.

To apply the intent to treat principle, an appropriate method for handling the missing data needs to be specified. A popular practical approach (not idea approach from statistical standpoint) is last value carried forward.

Intent to treat principle is not needed for all clinical trials and should not be interpreted as "include all enrolled subjects" or "all subjects who signed informed consent". The intent to treat is from the randomization standpoint, it has nothing to do with "study subject has intention to be treated in the clinical trial".

References:
1. ICH guidance E9 http://www.fda.gov/cder/guidance/ICH_E9-fnl.pdf
2. My presentation on ITT vs mITT http://webspace.webring.com/people/eu/um_3826/ITT_mITT_JSM2004.ppt
3. Wikipedian http://en.wikipedia.org/wiki/Intention_to_treat_analysis