How to Report Survival Analysis in a Scientific Paper

By Dr. Zubair Khalid, DVM, MS, PhD · 2026-08-22

Key Takeaways

• Core Reporting Elements: Survival analyses necessitate reporting the Kaplan-Meier estimator for time-to-event data, the log-rank test for group comparisons, and the Cox proportional hazards model for multivariable adjustment. This includes presenting numbers at risk, confidence intervals, and detailed censoring information to ensure transparency and reproducibility.
• Precise Definitions are Paramount: Clearly define the "time origin" (e.g., date of diagnosis, treatment initiation) and the "event of interest" (e.g., disease recurrence, death from any cause) with specific criteria. Ambiguous or inconsistent definitions, especially for composite endpoints or competing events, are critical omissions that invalidate interpretation.
• Kaplan-Meier Curves and Numbers at Risk: Report Kaplan-Meier curves visually depicting survival probabilities over time, crucially including the "numbers at risk" at each time interval below the x-axis. This allows readers to assess the precision of survival estimates, which decreases as the number at risk diminishes.
• Cox Model Assumptions and Effect Measures: For Cox proportional hazards models, explicitly test and report the proportional hazards assumption using methods like Schoenfeld residuals or log-minus-log plots. Report hazard ratios with 95% confidence intervals, and if the assumption is violated, consider time-varying coefficients or stratified models, or report restricted mean survival time as an alternative effect measure.
• Adherence to Reporting Guidelines: Select and adhere to study-design-specific reporting guidelines, such as CONSORT for randomized controlled trials or STROBE for observational studies, as recommended by the EQUATOR Network. These guidelines mandate specific reporting elements beyond statistical methods, ensuring comprehensive study reporting.

Quick Answer

• Report survival analyses with three core elements: the Kaplan-Meier estimator for time-to-event data, the log-rank test for group comparisons, and the Cox proportional hazards model for multivariable adjustment.
• Structure your methods and results sections around a reproducible template that includes numbers at risk, confidence intervals, and censoring details.
• A critical limitation is that survival analysis requires complete follow-up data and clear definitions of events, censoring, and time origins, which many manuscripts omit.

At a Glance

Understanding Survival Analysis Reporting Standards

Survival analysis occupies a distinct position in biomedical research because it addresses time-to-event data, where the outcome of interest is not simply whether an event occurs but also when it occurs. The statistical methods used in survival analysis, including Kaplan-Meier estimation, log-rank testing, and Cox proportional hazards regression, require specific reporting practices to ensure that readers can evaluate the validity of the findings and that other researchers can reproduce the analyses.

The need for standardized reporting in survival analysis stems from the complexity of the data structure. Unlike conventional analyses where each participant contributes a single measurement, survival data involve follow-up times, event indicators, and censoring information. Censoring occurs when the event of interest has not been observed for a participant by the end of the study period, when a participant is lost to follow-up, or when a participant withdraws from the study. The presence of censoring requires specialized statistical methods that account for incomplete follow-up information.

The EQUATOR Network serves as a central repository for reporting guidelines across multiple study designs and provides researchers with access to the appropriate standards for their specific study type. For survival analyses, the relevant reporting guidelines depend on the overall study design. A randomized controlled trial reporting survival outcomes should follow the CONSORT statement, while an observational study would follow the STROBE statement. These guidelines provide structured frameworks for reporting the essential elements of the study, including participant flow, outcome definitions, and statistical methods [<a href="#ref-1">1</a>].

The Committee on Publication Ethics Core Practices establish the broader ethical framework within which survival analysis reporting occurs. These practices address the integrity of the research record, including the accurate reporting of methods and results, proper attribution of authorship, and the management of data and conflicts of interest [<a href="#ref-2">2</a>]. When reporting survival analyses, researchers must ensure that their descriptions of methods and results are accurate and complete, as incomplete or misleading reporting can undermine the integrity of the scientific record.

Defining the Time Origin and Event of Interest

The time origin, also known as time zero, is the starting point from which follow-up time is measured. The choice of time origin is a critical decision that affects the interpretation of all subsequent survival estimates. Common time origins include the date of diagnosis, the date of treatment initiation, the date of randomization, or the date of surgery. The time origin must be clearly defined in the methods section and applied consistently across all participants.

The event of interest must also be precisely defined. In survival analysis, the event is the outcome that marks the end of the time-to-event period. For example, in a study of cancer treatment, the event might be disease recurrence, disease-specific death, or death from any cause. The event definition should include the criteria used to determine that an event has occurred, such as imaging findings, pathological confirmation, or clinical assessment.

The distinction between the event of interest and competing events is important. A competing event is an event that prevents the event of interest from occurring. For example, in a study of disease recurrence, death from an unrelated cause is a competing event. The handling of competing events affects the interpretation of survival estimates and should be described in the methods section.

The National Library of Medicine provides access to authoritative biomedical texts that describe the principles of survival analysis and the importance of clear definitions for time origins and events [<a href="#ref-3">3</a>]. These resources emphasize that the validity of survival analysis depends on accurate and consistent definitions of the time scale and the outcome.

Kaplan-Meier Curves and Numbers at Risk

The Kaplan-Meier method is the standard approach for estimating survival probabilities over time. The method produces a step function that decreases at each event time, with the size of the decrease depending on the number of events and the number of participants at risk at that time point. The Kaplan-Meier curve provides a visual representation of the survival experience of the study population.

When reporting Kaplan-Meier curves, the figure should include the number of participants at risk at each time point. The numbers at risk are typically displayed below the x-axis of the curve and indicate how many participants remain in the follow-up cohort at each specified time. This information is essential for interpreting the reliability of the survival estimates at different time points. As the number at risk decreases, the survival estimates become less precise and should be interpreted with greater caution.

The Kaplan-Meier curve should also display censoring events. Censored observations are typically indicated with tick marks on the curve. The inclusion of censoring marks allows readers to see when participants were lost to follow-up or when their follow-up ended without an event. The pattern of censoring can affect the interpretation of the survival curve, and the reporting should make the censoring pattern transparent.

The confidence intervals for the survival estimates should be reported. Confidence intervals provide a range of plausible values for the survival probability at each time point and convey the precision of the estimates. The confidence intervals can be displayed as shaded bands around the Kaplan-Meier curve or reported in the text for specific time points of interest.

Log-Rank Test and Group Comparisons

The log-rank test is the standard statistical test for comparing survival distributions between two or more groups. The test evaluates the null hypothesis that the survival distributions are the same across the groups. The log-rank test is a nonparametric test that does not assume a specific shape for the survival distributions.

When reporting the log-rank test, the manuscript should include the test statistic, the degrees of freedom, and the p-value. The p-value indicates the strength of evidence against the null hypothesis of no difference in survival between the groups. A p-value below the conventional threshold of 0.05 is often interpreted as evidence of a statistically significant difference, although the threshold should be specified in the methods section.

The log-rank test has limitations that should be acknowledged in the reporting. The test gives equal weight to all time points, which means that it is most sensitive to differences that occur early in the follow-up period. The test also assumes that the censoring patterns are similar across the groups being compared. If the censoring patterns differ substantially between groups, the log-rank test may not be appropriate, and alternative tests should be considered.

The reporting of the log-rank test should include the number of events in each group, the expected number of events under the null hypothesis, and the observed number of events. This information allows readers to understand the basis for the test result and to assess the adequacy of the sample size for detecting group differences.

Cox Proportional Hazards Regression

The Cox proportional hazards model is the most widely used method for multivariable survival analysis. The model estimates the hazard ratio, which represents the relative hazard of the event for one group compared to another, while adjusting for other covariates. The hazard ratio is interpreted as the multiplicative effect of the predictor variable on the hazard of the event.

The reporting of Cox models should include the hazard ratio, the 95% confidence interval, and the p-value for each covariate in the model. The hazard ratio should be reported with a clear description of the comparison being made. For example, a hazard ratio of 1.5 for a treatment variable indicates that the hazard of the event is 1.5 times higher in the treatment group compared to the control group.

The proportional hazards assumption is a key assumption of the Cox model. The assumption states that the hazard ratio between groups is constant over time. The proportional hazards assumption should be tested and the results reported. Common methods for testing the assumption include the Schoenfeld residuals test and the examination of log-minus-log plots. If the proportional hazards assumption is violated, alternative methods such as time-varying coefficients or stratified Cox models should be considered.

The selection of variables for inclusion in the Cox model should be described in the methods section. The variable selection approach, whether based on clinical relevance, statistical significance, or a combination of both, should be stated. The reporting should also describe how continuous variables were handled, including whether they were categorized and the basis for the categorization.

Numbers at Risk and Confidence Intervals

The numbers at risk are a critical component of survival analysis reporting. The numbers at risk indicate the number of participants who are still being followed and have not yet experienced the event at each time point. The numbers at risk should be reported at regular intervals throughout the follow-up period, typically at the same time points used for the x-axis of the Kaplan-Meier curve.

The numbers at risk provide context for interpreting the survival estimates. As the numbers at risk decrease, the survival estimates become less reliable and the confidence intervals widen. The reporting of the numbers at risk allows readers to assess the precision of the survival estimates at each time point and to identify time points where the estimates are based on small numbers of participants.

Confidence intervals should be reported for the survival estimates and the hazard ratios. The confidence interval provides a range of plausible values for the true effect and reflects the precision of the estimate. A wide confidence interval indicates that the estimate is imprecise, while a narrow confidence interval indicates greater precision. The confidence intervals should be reported with the point estimates in the text and in the figures.

The reporting of confidence intervals is particularly important for the hazard ratios from Cox models. The confidence interval for the hazard ratio provides information about the precision of the effect estimate and whether the effect is statistically significant. A confidence interval that includes 1.0 indicates that the effect is not statistically significant at the 5% level.

Methods Section Template

The methods section of a manuscript reporting survival analysis should describe the statistical methods in sufficient detail to allow replication. The following template provides a structured approach to reporting the methods for survival analysis.

The first paragraph should describe the study design and the participants. This includes the inclusion and exclusion criteria, the source of the participants, and the study period. The description should be sufficiently detailed to allow readers to understand the population to which the results apply.

The second paragraph should define the time origin and the event. The time origin should be described with reference to a specific clinical event, such as the date of diagnosis or the date of treatment initiation. The event should be defined with specific criteria for determining that the event has occurred.

The third paragraph should describe the statistical methods. This includes the Kaplan-Meier method for estimating survival curves, the log-rank test for comparing groups, and the Cox proportional hazards regression for multivariable analysis. The methods should describe how the proportional hazards assumption was tested and how the covariates were selected.

The fourth paragraph should describe the software and the version used for the analysis. The software should be identified by name and version, and the specific procedures or functions used should be described. This information is essential for reproducibility.

Results Section Template

The results section should present the findings of the survival analysis in a clear and structured manner. The following template provides a structured approach for reporting the results of survival analysis.

Participant Characteristics and Follow-Up

The first paragraph should describe the participant characteristics and the follow-up. This includes the number of participants, the number of events, the median follow-up time, and the number of participants censored. The follow-up time should be reported with the range or the interquartile range.

Kaplan-Meier Estimates

The second paragraph should describe the Kaplan-Meier estimates. This includes the survival probabilities at specific time points of interest, such as the 1-year, 3-year, or 5-year survival rates. The survival probabilities should be reported with their confidence intervals.

Group Comparisons

The third paragraph should describe the results of the log-rank test. This includes the test statistic, the degrees of freedom, and the p-value. The survival curves for the groups should be described, including any differences in the shape of the curves.

Cox Regression

The fourth paragraph should describe the results of the Cox regression analysis. This includes the hazard ratios, the confidence intervals, and the p-values for each covariate in the model. The results should be presented in a table with the hazard ratios and confidence intervals for each variable.

Table for Reporting Cox Model Results

Common Failure Patterns in Survival Analysis Reporting

Several common failure patterns recur in manuscripts reporting survival analysis. Recognizing these patterns can help researchers avoid them in their own reporting and can help reviewers identify problems in manuscripts under review.

Failure to Define the Time Origin

A common failure is the omission of a clear definition of the time origin. Without a clear definition of the time origin, the survival times are ambiguous, and the results cannot be interpreted or reproduced. The time origin should be defined in the methods section and applied consistently across all participants.

Failure to Report Censoring

Another common failure is the omission of censoring information. Censoring is an inherent feature of survival data, and the reporting should describe the reasons for censoring and the number of participants censored. The failure to report censoring can lead to misinterpretation of the survival estimates.

Failure to Report Confidence Intervals

The failure to report confidence intervals is a common problem in survival analysis reporting. Confidence intervals provide essential information about the precision of the estimates, and their omission makes it difficult to assess the reliability of the findings. The confidence intervals should be reported for the survival estimates and the hazard ratios.

Failure to Test the Proportional Hazards Assumption

The proportional hazards assumption is a key assumption of the Cox model, and the failure to test this assumption is a common problem. The authors should describe the methods used to test the assumption and report the results of the testing. If the assumption is violated, the authors should describe the alternative methods used.

Failure to Report the Software and Analysis Parameters

The failure to report the software and the analysis parameters is a common problem that affects the reproducibility of the analysis. The authors should identify the software and the version used, and describe the specific functions and parameters used for the analysis.

Reproducibility and Data Sharing

Reproducibility is a core principle of scientific research, and the reporting of survival analysis should support the reproduction of the analysis by other researchers. The reporting should include sufficient detail about the data, the methods, and the software to allow another researcher to reproduce the analysis.

The National Institutes of Health Data Management and Sharing Policy describes the expectations for data management and sharing for NIH-funded research [<a href="#ref-4">4</a>]. The policy requires that researchers plan for the management and sharing of data, and that the data be shared in a manner that is consistent with the policy. The data sharing plan should describe the data that will be shared, the repository where the data will be deposited, and the timeline for sharing.

The reporting of survival analysis should include a description of the data that were used for the analysis. This includes the source of the data, the inclusion and exclusion criteria, and the handling of missing data. The description should be detailed enough to allow another researcher to understand the data and to reproduce the analysis.

The NIH Grants and Funding pages provide information about the expectations for data management and sharing in NIH-funded research [<a href="#ref-5">5</a>]. The data management and sharing plan should be described in the grant application, and the plan should be updated as the research progresses. The plan should describe the data that will be collected, the standards for the data, and the repository where the data will be deposited.

Author Identity and ORCID

The reporting of survival analysis should include the identification of the authors and their affiliations. The use of ORCID identifiers provides a persistent and unique identifier for each researcher, which helps to disambiguate authors with similar names and to link the researcher to their publications and other research outputs [<a href="#ref-6">6</a>].

The ORCID for Researchers page describes the benefits of the ORCID identifier for researchers [<a href="#ref-6">6</a>]. The ORCID identifier can be used to link the researcher to their publications, their grants, and their other research activities. The use of ORCID identifiers in the manuscript reporting is recommended to ensure the accurate attribution of the research.

The author list should be determined according to the authorship criteria described in the Committee on Publication Ethics Core Practices [<a href="#ref-2">2</a>]. The authorship should be based on the substantial contributions to the conception, design, data collection, analysis, and interpretation of the study. The authors should have approved the final version of the manuscript and should be accountable for the content.

Publication Ethics and Reporting Integrity

The reporting of survival analysis should adhere to the publication ethics standards described in the Committee on Publication Ethics Core Practices [<a href="#ref-2">2</a>]. The Core Practices address the responsibilities of the authors, the reviewers, and the editors in the publication process.

The authors should ensure that the data and the results are reported accurately and completely. The authors should not fabricate or falsify the data, and should not omit the data that is relevant to the interpretation of the findings. The authors should also disclose any conflicts of interest that could influence the interpretation of the results.

The authors should also ensure that the manuscript is original and that it has not been published elsewhere. The authors should not duplicate the publication of the same data in multiple manuscripts, unless the duplication is justified and the prior publication is disclosed.

The reporting of survival analysis should also adhere to the reporting guidelines that are relevant to the study design. The EQUATOR Network provides a comprehensive database of reporting guidelines for diverse study designs [<a href="#ref-1">1</a>]. The authors should select the appropriate reporting guideline for their study design and follow the guideline in the preparation of the manuscript.

Practical Implementation Steps

The following steps provide a practical approach for reporting survival analysis in a scientific paper.

Step 1: Define the Time Origin and Event

The first step is to define the time origin and the event. The time origin should be a clear and reproducible clinical event, and the event should be defined with specific criteria. The definitions should be recorded in the study protocol and applied consistently.

Step 2: Perform the Kaplan-Meier Analysis

The second step is to perform the Kaplan-Meier analysis. The analysis should produce the survival curves, the numbers at risk, and the confidence intervals. The results should be reviewed to ensure that the curves are consistent with the data.

Step 3: Perform the Log-Rank Test

The third step is to perform the log-rank test to compare the survival distributions between the groups. The test should be performed for the primary comparison and for the subgroup comparisons. The results should be recorded with the test statistic and the p-value.

Step 4: Perform the Cox Regression

The fourth step is to perform the Cox regression analysis. The model should include the covariates that are relevant to the research question. The proportional hazards assumption should be tested, and the results should be recorded.

Step 5: Prepare the Figures and Tables

The fifth step is to prepare the figures and the tables. The Kaplan-Meier curve should include the numbers at risk and the censoring marks. The tables should include the hazard ratios, the confidence intervals, and the p-values.

Step 6: Write the Methods and Results Sections

The sixth step is to write the methods and the results sections. The methods section should describe the statistical methods in detail, and the results section should present the findings in a structured manner.

Step 7: Review the Reporting Against the Guidelines

The seventh step is to review the reporting against the relevant reporting guidelines. The EQUATOR Network provides access to the reporting guidelines for the study design [<a href="#ref-1">1</a>]. The review should ensure that all the essential elements are reported.

Records and Measurements

The reporting of survival analysis should include the records of the data and the measurements that are used for the analysis. The records should include the following:

• The participant-level data with the time to event and the event status
• The censoring information with the reasons for censoring
• The covariate data for the variables included in the Cox model
• The software code used for the analysis
• The output of the analysis, including the survival estimates, the hazard ratios, and the confidence intervals

The records should be maintained in a manner that allows the analysis to be reproduced. The data should be stored in a secure and accessible repository, and the software code should be documented and versioned.

Quality Controls and Verification

The quality of the survival analysis reporting should be verified through a series of checks. The checks should be performed before the submission of the manuscript and should be documented.

Check the Time Origin and Event Definitions

The first check is to verify that the time origin and the event definitions are clear and consistent. The definitions should be applied consistently across all participants, and the definitions should be described in the methods section.

Check the Numbers at Risk

The second check is to verify that the numbers at risk are correct. The numbers at risk should be consistent with the data, and the numbers should be reported at the appropriate time intervals.

Check the Confidence Intervals

The third check is to verify that the confidence intervals are reported for the survival estimates and the hazard ratios. The confidence intervals should be consistent with the point estimates and the sample size.

Check the Proportional Hazards Assumption

The fourth check is to verify that the proportional hazards assumption has been tested and that the results are reported. The check should include the review of the Schoenfeld residuals or the log-minus-log plots.

Check the Software and the Parameters

The fifth check is to verify that the software and the parameters are reported. The software should be identified by name and version, and the parameters should be described.

Limitations and Interpretation

The reporting of survival analysis should include a discussion of the limitations of the analysis. The limitations should be described in the discussion section of the manuscript and should be considered in the interpretation of the findings.

Censoring and Follow-Up

The censoring and the follow-up are important limitations of the survival analysis. The censoring can introduce bias if the censoring is related to the event. The follow-up time should be sufficient to observe the events of interest, and the follow-up should be similar across the groups.

Sample Size and Power

The sample size and the power are important limitations of the survival analysis. The sample size should be sufficient to detect the effect of interest, and the power should be reported. The power is the probability of detecting the effect if the effect exists.

Confounding and Bias

The confounding and the bias are important limitations of the survival analysis. The Cox model can adjust for the measured confounders, but the unmeasured confounders can still bias the results. The bias can be introduced by the selection of the participants, the measurement of the variables, and the handling of the missing data.

Generalizability

The generalizability of the findings is an important limitation of the survival analysis. The findings may not be generalizable to the populations that are different from the study population. The generalizability should be considered in the interpretation of the results.

Professional Escalation Criteria

The reporting of survival analysis should include the criteria for the professional escalation of the findings. The escalation criteria should be defined in the study protocol and should be applied when the findings meet the criteria.

Escalation for the Safety Concerns

The findings should be escalated if the safety concerns are identified. The safety concerns should be reported to the appropriate authorities, and the study should be reviewed.

Escalation for the Unexpected Findings

The findings should be escalated if the findings are unexpected. The unexpected findings should be reviewed by the study team, and the findings should be reported to the appropriate authorities.

Escalation for the Data Quality Issues

The findings should be escalated if the data quality issues are identified. The data quality issues should be reviewed, and the data should be corrected or the analysis should be repeated.

A Decision Framework for Choosing Between Survival Analysis Reporting Approaches

Beyond knowing what to report, researchers often struggle with deciding which survival analysis approach fits their data structure and research question. This section provides a practical decision framework that helps you select the correct reporting strategy before you write a single sentence of your manuscript.

Step 1: Classify Your Data Structure

The first decision point concerns whether your data contain competing risks. A competing risk is an event that alters the probability of experiencing the event of interest. For example, in a study of time to disease recurrence, death from an unrelated cause is a competing risk because patients who die cannot subsequently experience recurrence.

Ask yourself three questions to classify your data:

1. Can a participant experience more than one type of event during follow-up?
2. Does the occurrence of one event type preclude the occurrence of the event of interest?
3. Are you interested in the cumulative incidence of the event in the presence of other events?

If the answer to all three questions is yes, you are analyzing competing risks data. The standard Kaplan-Meier estimator and the Cox model treat competing events as censored observations, which can overestimate the cumulative incidence of the event of interest. In this situation, you should report the cumulative incidence function and consider the Fine-Gray subdistribution hazard model instead of the standard Cox model.

If the answer to any question is no, the standard Kaplan-Meier and Cox approaches are appropriate, and you can proceed with the reporting templates described in the main sections of this article.

Step 2: Assess the Proportional Hazards Assumption

The second decision point occurs after you have fit a Cox model. The proportional hazards assumption states that the hazard ratio between groups remains constant over time. If this assumption is violated, the hazard ratio from a standard Cox model is a misleading summary of the treatment effect.

Use the following assessment sequence:

1. Examine the Schoenfeld residuals for each covariate. A non-significant test suggests the assumption holds.
2. Plot the log-minus-log survival curves for categorical covariates. Parallel curves indicate the assumption is satisfied.
3. If the assumption is violated, decide between two reporting alternatives: a time-varying coefficient model that allows the hazard ratio to change at specified time points, or a stratified Cox model that allows the baseline hazard to differ across strata.

The decision should be documented in the methods section with the test results and the rationale for the chosen alternative.

Step 3: Determine the Appropriate Effect Measure

The third decision point concerns the effect measure you will report. The hazard ratio is the standard effect measure from a Cox model, but it is not always the most interpretable or clinically meaningful measure.

Consider reporting the restricted mean survival time instead of the hazard ratio when the proportional hazards assumption is violated. The restricted mean survival time is the average survival time up to a specified time point, such as 5 years. This measure does not require the proportional hazards assumption and provides a direct interpretation in units of time.

The decision between the hazard ratio and the restricted mean survival time should be based on the primary research question. If the question is about the relative rate of events over time, report the hazard ratio. If the question is about the average survival time within a clinically relevant window, report the restricted mean survival time.

Step 4: Match the Reporting Guideline to the Study Design

The final decision point involves selecting the appropriate reporting guideline. The EQUATOR Network provides a searchable database of reporting guidelines, and the correct guideline depends on the overall study design, not the statistical method [<a href="#ref-1">1</a>].

For a randomized controlled trial reporting survival outcomes, use the CONSORT statement. For an observational cohort study, use the STROBE statement. For a prognostic factor study, use the TRIPOD statement. Each guideline has specific requirements for reporting survival analyses that go beyond the statistical details.

The decision framework should be applied before data analysis begins. Document the decisions in the study protocol and reference the protocol in the manuscript methods section.

Common Failure Patterns in the Decision Process

Three recurring errors undermine the validity of survival analysis reporting.

The first error is applying the standard Cox model to competing risks data without acknowledging the competing events. This produces overestimated cumulative incidence and biased hazard ratios. The manuscript should report the cumulative incidence function and the subdistribution hazard model when competing risks are present.

The second error is reporting the hazard ratio without testing the proportional hazards assumption. The assumption test should be described in the methods section and the results reported in the results section. If the assumption is violated, the alternative model should be reported.

The third error is selecting a reporting guideline based on the outcome instead of the study design. A randomized trial reporting survival outcomes should follow CONSORT, not STROBE. The guideline selection should be stated in the methods section.

Records and Measurements for the Decision Framework

The following records should be maintained to support the decision framework:

• The study protocol with the pre-specified decision rules for the data structure classification
• The output of the proportional hazards assumption tests, including the Schoenfeld residual plots and the log-minus-log plots
• The cumulative incidence function estimates and the subdistribution hazard ratios for competing risks analyses
• The restricted mean survival time estimates and their confidence intervals
• The reporting guideline selected and the checklist items completed

These records should be stored with the analysis code and the participant-level data to support the reproducibility of the analysis.

Professional Escalation Criteria

The decision framework includes specific criteria for escalating concerns to a statistician or a senior researcher. Escalation is required when the data structure is ambiguous and the competing risks classification is uncertain. Escalation is also required when the proportional hazards assumption tests produce conflicting results across different methods. Finally, escalation is required when the restricted mean survival time and the hazard ratio produce conflicting conclusions about the treatment effect.

The escalation should be documented in the analysis records with the date, the reason for the escalation, and the resolution. The documentation should be available for the manuscript submission and for the peer review process.

Frequently Asked Questions

What is the difference between the Kaplan-Meier curve and the Cox model?

The Kaplan-Meier curve is a nonparametric method that estimates the survival probability over time without adjusting for covariates. The Cox model is a semiparametric method that estimates the hazard ratio for the effect of covariates on the survival time. The Kaplan-Meier curve is used to describe the survival experience of the study population, while the Cox model is used to assess the effect of covariates on the survival time.

How do I report the numbers at risk in a Kaplan-Meier figure?

The numbers at risk should be reported below the x-axis of the Kaplan-Meier figure at regular time intervals. The numbers at risk indicate the number of participants who are still at risk of the event at each time point. The numbers at risk should be reported at the same time points as the x-axis of the figure.

What is the proportional hazards assumption and how do I test it?

The proportional hazards assumption states that the hazard ratio between groups is constant over time. The assumption can be tested using the Schoenfeld residuals or the log-minus-log plots. If the assumption is violated, the Cox model may not be appropriate, and alternative methods should be considered.

How do I report the confidence intervals for the hazard ratios?

The confidence intervals for the hazard ratios should be reported in the table of the Cox model results. The confidence intervals should be reported alongside the hazard ratios and the p-values. The confidence intervals provide a range of plausible values for the true hazard ratio.

What is the difference between the hazard ratio and the relative risk?

The hazard ratio is the ratio of the hazard of the event between two groups, where the hazard is the instantaneous rate of the event at a given time. The relative risk is the ratio of the probability of the event between two groups over a fixed time period. The hazard ratio is a time-dependent measure, while the relative risk is a time-fixed measure.

How do I handle the competing events in the survival analysis?

The competing events are the events that prevent the event of interest from occurring. The competing events can be handled using the cumulative incidence function or the cause-specific hazard model. The choice of the method depends on the research question and the data.

What should I include in the methods section for the survival analysis?

The methods section should include the description of the study design, the participants, the time origin, the event definition, the statistical methods, and the software used for the analysis. The methods should be described in sufficient detail to allow the analysis to be reproduced.

How do I report the censoring in the survival analysis?

The censoring should be reported in the methods section and the results section. The methods section should describe the reasons for censoring, and the results section should report the number of censored participants and the censoring pattern. The censoring should be indicated in the Kaplan-Meier figure with the tick marks.

Related Bioinformatics Guides

• Genomic Data Analysis Tools: A Comparative Guide for Researchers
• How to Interpret Gene Set Enrichment Analysis Results
• Metabolomics Data Analysis in R: A Practical Workflow
• Microbiome Data Analysis in R: A Practical Guide for Compositional Data
• Foundation Models in Genetics: Opportunities and Challenges

Related Clinical & Scientific Guides

• A Practical Guide to Detecting Antimicrobial Resistance Genes in Shotgun Metagenomic Data
• Computational Immunology: Modeling the Immune System
• How to Set Hard Filters for Germline Variant Calling: A Practical Guide to GATK Best Practices

References and Further Reading

This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.