Sample Size Calculation for Diagnostic Test Accuracy Studies

By Dr. Zubair Khalid, DVM, MS, PhD · 2026-08-27

Quick Answer

• Sample size for diagnostic accuracy studies depends on the expected sensitivity and specificity, the desired precision of those estimates, and the prevalence of the target condition in the study population.
• Use the precision-based approach with the Wilson score interval for sensitivity and specificity, then adjust the final sample size upward to account for the proportion of participants expected to have the condition.
• The calculated sample size is a minimum estimate, and researchers must also consider missing data, indeterminate test results, and the need for subgroup analyses when planning recruitment.

At a Glance

Study Design Context for Diagnostic Accuracy Research

Diagnostic test accuracy studies evaluate how well an index test identifies a target condition compared with a reference standard. The index test is the experimental diagnostic method under investigation, and the reference standard is the best available method for determining the true disease status. These studies produce two fundamental estimates: sensitivity, the proportion of individuals with the condition who test positive, and specificity, the proportion of individuals without the condition who test negative.

Sample size planning for diagnostic accuracy studies differs from sample size planning for comparative trials. In a randomized trial, the sample size is driven by the need to detect a difference between groups. In a diagnostic accuracy study, the sample size is driven by the need to estimate sensitivity and specificity with acceptable precision. The precision of these estimates depends on the number of participants with the condition for sensitivity and the number without the condition for specificity. Because the prevalence of the condition determines how many participants fall into each category, prevalence is a central factor in the sample size calculation.

The research question must be defined before any sample size calculation. The researcher must specify the target condition, the index test, the reference standard, and the intended clinical role of the test. A test intended to rule out a serious disease may require high sensitivity with narrow confidence intervals. A test intended to confirm a diagnosis before treatment may require high specificity. The sample size calculation should reflect the clinical consequence of misclassification for each accuracy parameter.

The study population also affects the generalizability of the accuracy estimates. Diagnostic accuracy can vary across clinical settings, disease severity, and demographic groups. If the study population is restricted to a narrow group, the sensitivity and specificity estimates may not apply to the broader population where the test will be used. The sample size calculation should consider whether subgroup analyses are planned, because each subgroup requires its own precision target.

Core Principles of Sample Size Estimation

Precision-Based Estimation

The precision-based approach to sample size calculation for diagnostic accuracy studies uses the width of the confidence interval around the estimated sensitivity and specificity. The researcher specifies a target precision, which is the maximum acceptable half-width of the confidence interval. For example, a target precision of 5 percentage points means the 95% confidence interval for sensitivity should extend no more than 5 percentage points above or below the point estimate.

The standard error of a proportion is the square root of the product of the proportion and its complement divided by the sample size. For sensitivity, the proportion is the expected sensitivity and the sample size is the number of participants with the condition. For specificity, the proportion is the expected specificity and the sample size is the number of participants without the condition. The required sample size for each estimate is calculated separately, and the larger of the two is used for the study.

The expected sensitivity and specificity values used in the calculation should come from prior studies, pilot data, or the clinical judgment of the research team. These values are estimates, and the actual sensitivity and specificity observed in the study may differ. If the observed values are lower than expected, the confidence intervals will be wider than planned. If the observed values are higher than expected, the confidence intervals will be narrower than planned.

Prevalence and the Two-Part Sample Size

The prevalence of the target condition determines how many participants must be enrolled to obtain the required number of participants with and without the condition. If the expected prevalence is 50%, the total sample size is approximately twice the sample size needed for each accuracy estimate. If the expected prevalence is 20%, the total sample size is approximately five times the sample size needed for the sensitivity estimate, because only one in five enrolled participants is expected to have the condition.

The prevalence estimate should be based on the study population and the clinical setting. Prevalence in a referral center may be higher than prevalence in a primary care setting. The researcher should use the prevalence that is most relevant to the intended use of the test. If the prevalence is uncertain, the researcher should consider a range of prevalence values and calculate the sample size for each.

Confidence Level and Power

The confidence level is the probability that the confidence interval contains the true value of the sensitivity or specificity. The standard confidence level is 95%, which corresponds to a z-value of 1.96. A higher confidence level, such as 99%, requires a larger sample size. A lower confidence level, such as 90%, requires a smaller sample size.

Power is the probability that the study will detect a statistically significant result if one exists. In diagnostic accuracy studies, power is often used to determine whether the sensitivity or specificity is above a minimum acceptable threshold. For example, a researcher might want to demonstrate that sensitivity is at least 80% with 90% power. This approach requires a one-sided hypothesis test and a different sample size formula than the precision-based approach.

The choice between the precision-based and power-based approaches depends on the research question. If the goal is to estimate sensitivity and specificity with a certain level of precision, the precision-based approach is appropriate. If the goal is to test whether sensitivity or specificity exceeds a minimum threshold, the power-based approach is appropriate. Some studies use both approaches and report the larger sample size.

Finite Population Correction

The finite population correction adjusts the sample size when the study population is small relative to the total population. The correction reduces the required sample size because sampling from a finite population produces more precise estimates than sampling from an infinite population. The correction factor is the square root of the ratio of the population size minus the sample size to the population size minus one.

The finite population correction is rarely needed in diagnostic accuracy studies because the target population is usually large. However, if the study is conducted in a small, well-defined population, such as all patients with a rare condition at a single institution, the correction may be appropriate. The researcher should apply the correction only when the sample size is a substantial fraction of the population size.

Sample Size Formulas for Sensitivity and Specificity

Precision-Based Formula

The precision-based sample size formula for a single proportion is:

n = (z^2 p (1 - p)) / d^2

In this formula, n is the required sample size, z is the z-value corresponding to the desired confidence level, p is the expected proportion, and d is the desired half-width of the confidence interval. For a 95% confidence level, z is 1.96. For a 90% confidence level, z is 1.645.

For sensitivity, p is the expected sensitivity and n is the number of participants with the condition. For specificity, p is the expected specificity and n is the number of participants without the condition. The total sample size is then calculated by dividing each n by the expected prevalence for sensitivity and by one minus the prevalence for specificity.

The formula assumes a normal approximation to the binomial distribution. This approximation is valid when the expected number of events is sufficiently large. A common rule is that the expected number of events and the expected number of non-events should both be at least 5. If the expected sensitivity or specificity is very high, such as above 95%, the normal approximation may be less accurate and the sample size may need to be increased.

Example Calculation for Sensitivity

Consider a study evaluating a new diagnostic test for a condition with an expected sensitivity of 90% and an expected specificity of 85%. The researcher wants a 95% confidence interval with a half-width of 5 percentage points for both estimates. The expected prevalence of the condition in the study population is 30%.

For sensitivity, the sample size is calculated as:

n_sensitivity = (1.96² 0.90 0.10) / 0.05² = 138.3

The sample size for sensitivity is rounded up to 139 participants with the condition. Because the prevalence is 30%, the total number of participants needed is 139 / 0.30 = 463.3, rounded up to 464.

For specificity, the sample size is calculated as:

n_specificity = (1.96² 0.85 0.15) / 0.05² = 195.9

The sample size for specificity is rounded up to 196 participants without the condition. Because the prevalence is 30%, the proportion without the condition is 70%, and the total number of participants needed is 196 / 0.70 = 280.

The larger total sample size is 464 participants. This sample size ensures that the study will have enough participants with the condition to estimate sensitivity with the target precision. The study will also have more than enough participants without the condition to estimate specificity with the target precision.

Exact example for low prevalence

Consider the same sensitivity and specificity targets but a prevalence of 10%. The sample size for sensitivity is still 139 participants with the condition. The total number of participants needed is 139 / 0.10 = 1,390. The sample size for specificity is still 196 participants without the condition. The total number of participants needed is 196 / 0.90 = 217.8, rounded up to 218.

The larger total sample size is 1,390 participants. The low prevalence has a substantial effect on the total sample size because most enrolled participants will not have the condition. The researcher must decide whether the study is feasible with this sample size or whether the precision target should be relaxed.

Power-Based Formula for a Minimum Threshold

The power-based formula for testing whether sensitivity exceeds a minimum threshold is:

n = (z_alpha sqrt(p0 (1 - p0)) + z_beta sqrt(p1 (1 - p1)))² / (p1 - p0)²

In this formula, p0 is the minimum acceptable sensitivity, p1 is the expected sensitivity, z_alpha is the z-value for the significance level, and z_beta is the z-value for the desired power. For a one-sided test at the 5% significance level, z_alpha is 1.645. For 80% power, z_beta is 0.84. For 90% power, z_beta is 1.28.

The power-based approach is used when the researcher wants to demonstrate that the test meets a minimum performance threshold. For example, a researcher might want to show that sensitivity is at least 80% with 90% power, expecting the true sensitivity to be 90%. The sample size is calculated for the number of participants with the condition, and the total sample size is then adjusted for prevalence.

The power-based approach is more complex than the precision-based approach because it requires specifying both the minimum acceptable value and the expected value. The difference between these two values is the effect size. A larger effect size requires a smaller sample size. A smaller effect size requires a larger sample size.

Practical Workflow for Sample Size Determination

Step 1: Define the Research Question and Target Condition

The research question must specify the target condition, the index test, the reference standard, and the intended population. The target condition should be defined with clear diagnostic criteria. The index test should be described in enough detail that another researcher could replicate the test. The reference standard should be the best available method for determining the true disease status.

The intended population should be defined by the clinical setting, the demographic characteristics, and the spectrum of disease severity. The sample size calculation should use the prevalence of the target condition in this specific population. If the prevalence is not known, the researcher should conduct a pilot study or use prevalence estimates from the literature.

Step 2: Specify the Expected Sensitivity and Specificity

The expected sensitivity and specificity should be based on the best available evidence. This evidence may come from previous studies of the index test, studies of similar tests, or the clinical judgment of the research team. The expected values should be realistic and should reflect the performance of the test in the intended population.

If the test is a new technology with no prior data, the researcher may need to conduct a pilot study to estimate sensitivity and specificity. The pilot study should be small but large enough to provide a reasonable estimate. The pilot study results should be reported separately from the main study results.

Step 3: Choose the Precision or Power Target

The researcher must decide whether the study will estimate sensitivity and specificity with a target precision or test whether the estimates exceed a minimum threshold. The precision-based approach is more common and is appropriate for most diagnostic accuracy studies. The power-based approach is appropriate when the test must meet a regulatory or clinical minimum threshold.

The precision target should be clinically meaningful. A half-width of 5 percentage points is common, but a narrower target may be needed if the test is used for a high-stakes decision. A wider target may be acceptable if the test is used for screening and the consequences of misclassification are low.

Step 4: Calculate the Sample Size for Each Estimate

The sample size for sensitivity and specificity should be calculated separately using the appropriate formula. The precision-based formula requires the expected proportion, the z-value for the confidence level, and the desired half-width. The power-based formula requires the minimum acceptable value, the expected value, the significance level, and the power.

The sample size for sensitivity is the number of participants with the condition. The sample size for specificity is the number of participants without the condition. The two sample sizes are usually different because the expected sensitivity and specificity are usually different.

Step 5: Adjust for Prevalence

The total sample size is calculated by dividing the sample size for sensitivity by the expected prevalence and the sample size for specificity by the complement of the prevalence. The larger of the two total sample sizes is the required total sample size.

The prevalence estimate should be reviewed after the calculation. If the prevalence is uncertain, the researcher should calculate the total sample size for a range of prevalence values. The researcher should also consider the feasibility of recruiting the required number of participants within the study timeline and budget.

Step 6: Account for Missing Data and Indeterminate Results

The sample size calculation assumes that every enrolled participant will have a valid test result and a valid reference standard result. In practice, some participants will have missing data, indeterminate test results, or will withdraw from the study. The researcher should increase the total sample size to account for these losses.

The expected rate of missing data and indeterminate results should be based on the literature or a pilot study. A common approach is to increase the sample size by 10 to 20% to account for these losses. The researcher should document the expected rate and the adjustment in the study protocol.

Step 7: Document the Calculation and Assumptions

The sample size calculation should be documented in the study protocol. The documentation should include the expected sensitivity and specificity, the precision or power target, the confidence level, the expected prevalence, and the adjustment for missing data. The documentation should also include the formula used and the resulting sample size.

The documentation should be transparent enough that another researcher can reproduce the calculation. The assumptions should be reviewed by the research team and, if appropriate, by a statistician. The sample size calculation should be finalized before the study begins and should not be changed after the data are collected.

Options and Tradeoffs in Sample Size Planning

Precision versus Power

The precision-based approach and the power-based approach answer different research questions. The precision-based approach estimates the sensitivity and specificity with a target confidence interval width. The power-based approach tests whether the sensitivity or specificity exceeds a minimum threshold. The choice between the two approaches depends on the purpose of the study.

A study that is designed to estimate the performance of a test for clinical use should use the precision-based approach. A study that is designed to demonstrate that a test meets a regulatory or clinical minimum threshold should use the power-based approach. Some studies may use both approaches, and the larger sample size should be used.

Narrow versus Wide Precision

A narrower precision target requires a larger sample size. A half-width of 3 percentage points requires a sample size that is approximately 2.8 times larger than a half-width of 5 percentage points. The researcher should choose the precision target that is clinically meaningful and feasible.

The precision target should be based on the clinical consequences of the test. If the test is used to make a high-stakes decision, such as whether to start a treatment with serious side effects, a narrower precision is appropriate. If the test is used for screening and the consequences of misclassification are low, a wider precision may be acceptable.

Single-Stage versus Multi-Stage Designs

A single-stage design enrolls all participants and then evaluates the test. A multi-stage design enrolls participants in stages and evaluates the test after each stage. Multi-stage designs can reduce the total sample size if the test performs better or worse than expected.

Multi-stage designs are more complex to plan and analyze. The sample size for each stage must be calculated in advance, and the stopping rules must be specified. The researcher should consult a statistician before using a multi-stage design.

Fixed versus Adaptive Designs

A fixed design has a predetermined sample size that is not changed during the study. An adaptive design allows the sample size to be adjusted based on the results of an interim analysis. Adaptive designs may be more efficient than fixed designs but require more complex planning and analysis.

Adaptive designs are less common in diagnostic accuracy studies than in clinical trials. The researcher should consult a statistician before using an adaptive design. The adaptive design should be described in the study protocol before the study begins.

Observations and Measurements

Recording the Number of Participants with the Condition

The number of participants with the condition is the denominator for the sensitivity estimate. This number is determined by the reference standard result. The reference standard should be applied to all participants, regardless of the index test result. The reference standard result should be recorded for each participant.

The number of participants with the condition should be compared with the expected number from the sample size calculation. If the observed number is lower than expected, the sensitivity estimate will be less precise than planned. If the observed number is higher than expected, the sensitivity estimate will be more precise than planned.

Recording the Number of Participants without the Condition

The number of participants without the condition is the denominator for the specificity estimate. This number is also determined by the reference standard result. The number of participants without the condition should be recorded for each participant.

The number of participants without the condition should be compared with the expected number from the sample size calculation. If the observed number is lower than expected, the specificity estimate will be less precise than planned. If the observed number is higher than expected, the specificity estimate will be more precise than planned.

Recording the Index Test Results

The index test result should be recorded for each participant. The result should be recorded as positive, negative, or indeterminate. The index test result should be recorded independently of the reference standard result to avoid bias.

The index test results should be compared with the reference standard results to calculate sensitivity and specificity. The comparison should be performed after all data have been collected. The researcher should not calculate sensitivity and specificity during the study because this may influence the data collection.

Recording the Reference Standard Results

The reference standard result should be recorded for each participant. The reference standard should be performed by a person who is blinded to the index test result. The reference standard result should be recorded as positive or negative.

The reference standard result should be compared with the index test result to calculate sensitivity and specificity. The reference standard should be the best available method for determining the true disease status. If the reference standard is imperfect, the sensitivity and specificity estimates will be biased.

Records and Documentation

Study Protocol

The study protocol should include the research question, the target condition, the index test, the reference standard, the intended population, and the sample size calculation. The protocol should also include the recruitment plan, the data collection plan, and the analysis plan. The protocol should be written before the study begins and should be approved by the research team and any relevant institutional review board.

The protocol should be followed throughout the study. Any deviations from the protocol should be documented and reported. The protocol should be updated if the sample size calculation is revised.

Data Collection Forms

The data collection forms should capture the index test result, the reference standard result, and the demographic and clinical characteristics of each participant. The forms should be designed to minimize errors and to ensure that all required data are collected. The forms should be tested before the study begins.

The data collection forms should be completed for each participant. The forms should be reviewed for completeness and accuracy. The data should be entered into a database and verified for accuracy.

Statistical Analysis Plan

The statistical analysis plan should describe how the sensitivity and specificity will be calculated and how the confidence intervals will be constructed. The plan should also describe how the prevalence will be estimated and how the missing data will be handled. The plan should be written before the study begins and should be followed during the analysis.

The statistical analysis plan should be reviewed by a statistician. The plan should be updated if the study design changes. The analysis should be performed according to the plan, and any deviations should be documented.

Data Sharing and Management

The data management and sharing policy of the National Institutes of Health requires that researchers plan for the management and sharing of data generated by NIH-funded research. The data management and sharing plan should describe the types of data that will be generated, the standards that will be used, and the plan for sharing the data. The plan should be included in the grant application and should be updated as needed.

The data management and sharing plan should be consistent with the study protocol and the statistical analysis plan. The plan should describe how the data will be stored, preserved, and shared. The plan should also describe how the confidentiality of the participants will be protected.

Quality Controls and Verification

Blinding of the Index Test and Reference Standard

The index test result should be interpreted without knowledge of the reference standard result. The reference standard result should be interpreted without knowledge of the index test result. Blinding prevents bias in the interpretation of the test results.

Blinding should be implemented in the study design. The index test should be performed and interpreted by a person who does not know the reference standard result. The reference standard should be performed and interpreted by a person who does not know the index test result. The blinding should be documented in the study protocol.

Verification of the Reference Standard

The reference standard should be verified to be the best available method for determining the true condition status. The reference standard should be applied to all participants, regardless of the index test result. If the reference standard is applied only to participants with a positive index test result, the sensitivity and specificity estimates will be biased.

The reference standard should be applied consistently to all participants. The reference standard should be performed by a competent person. The reference standard result should be recorded for each participant.

Quality Control of the Index Test

The index test should be performed according to a standardized protocol. The protocol should describe the equipment, the reagents, the procedure, and the interpretation of the results. The protocol should be followed consistently for all participants.

The quality of the index test should be monitored during the study. The quality control samples should be run at regular intervals. The results of the quality control samples should be recorded and reviewed.

Handling of Indeterminate Results

Indeterminate results should be handled according to a pre-specified plan. The plan should describe how the indeterminate results will be classified and how they will be included in the analysis. The plan should be described in the study protocol.

The indeterminate results should be reported separately from the positive and negative results. The number of indeterminate results should be reported in the study results. The analysis should be performed with and without the indeterminate results to assess the impact on the sensitivity and specificity estimates.

Common Failure Patterns

Underestimating the Prevalence

A common failure in sample size calculation is underestimating the prevalence of the target condition. If the actual prevalence is lower than the expected prevalence, the study will enroll fewer participants with the condition than planned. The sensitivity estimate will be less precise than planned.

The researcher should use a conservative prevalence estimate. The prevalence should be based on the study population and the clinical setting. If the prevalence is uncertain, the researcher should calculate the sample size for a range of prevalence values and use the largest sample size.

Overestimating the Sensitivity or Specificity

Another common failure is overestimating the sensitivity or specificity of the index test. If the actual sensitivity or specificity is lower than the expected value, the confidence interval will be wider than planned. The study may not have enough precision to answer the research question.

The researcher should use a conservative estimate of sensitivity and specificity. The estimate should be based on the literature or a pilot study. If the estimate is uncertain, the researcher should calculate the sample size for a range of values and use the largest sample size.

Ignoring the Prevalence Adjustment

A common failure is calculating the sample size for sensitivity and specificity but not adjusting for prevalence. The sample size for sensitivity is the number of participants with the condition, not the total number of participants. The total number of participants must be larger than the sample size for sensitivity if the prevalence is less than 100%.

The researcher should always adjust the sample size for prevalence. The adjustment should be documented in the study protocol. The total sample size should be reviewed to ensure that it is feasible.

Using the Wrong Formula

A common failure is using the wrong formula for the sample size calculation. The formula for a diagnostic accuracy study is different from the formula for a clinical trial. The formula for a diagnostic accuracy study uses the expected sensitivity and specificity, not the expected difference between groups.

The researcher should use the formula that is appropriate for the study design. The formula should be reviewed by a statistician. The formula should be documented in the study protocol.

Not Accounting for Missing Data

A common failure is not accounting for missing data and indeterminate results. If the study has missing data, the number of participants with valid results will be lower than the number enrolled. The sensitivity and specificity estimates will be less precise than planned.

The researcher should account for missing data in the sample size calculation. The sample size should be increased by the expected rate of missing data. The expected rate should be based on the prior experience or a pilot study.

Limitations and Interpretation

The Sample Size Is a Minimum Estimate

The sample size calculation provides a minimum estimate of the number of participants needed to achieve the target precision or power. The actual sample size may need to be larger to account for missing data, indeterminate results, and subgroup analyses. The researcher should consider the sample size as a starting point and adjust it as needed.

The sample size calculation is based on assumptions about the expected sensitivity, specificity, and prevalence. If the assumptions are incorrect, the sample size will be incorrect. The researcher should review the assumptions and adjust the sample size if necessary.

The Sample Size Does Not Guarantee a Precise Estimate

The sample size calculation provides the sample size needed to achieve a target precision on average. The actual precision of the estimate will depend on the observed sensitivity and specificity. If the observed sensitivity or specificity is lower than expected, the confidence interval will be wider than planned.

The researcher should report the observed precision of the estimates. The confidence intervals should be reported for the sensitivity and specificity. The confidence intervals should be interpreted in the context of the sample size and the expected precision.

The Sample Size Does Not Address All Sources of Bias

The sample size calculation addresses the precision of the estimates but does not address the validity of the estimates. The sensitivity and specificity estimates can be biased if the study design is flawed. The bias can be introduced by the selection of the participants, the reference standard, or the interpretation of the test results.

The researcher should design the study to minimize bias. The study should use a consecutive or random sample of participants. The reference standard should be applied to all participants. The index test and reference standard should be interpreted independently.

The Sample Size Does Not Address the Clinical Utility of the Test

The sample size calculation addresses the precision of the sensitivity and specificity estimates but not the clinical utility of the test. A test with high sensitivity and specificity may not be clinically useful if the prevalence is low or if the test is not feasible in the clinical setting.

The researcher should consider the clinical utility of the test in the interpretation of the results. The sensitivity and specificity should be interpreted in the context of the prevalence and the clinical consequences of the test results. The researcher should also consider the positive and negative predictive values of the test.

Safety and Regulatory Context

Institutional Review Board Approval

The study should be reviewed and approved by an institutional review board before the study begins. The institutional review board will review the study protocol, the informed consent form, and the data collection plan. The board will assess the risks and benefits of the study and the protection of the participants.

The institutional review board approval should be obtained before the study begins. The approval should be documented in the study records. The study should be conducted according to the approved protocol.

Informed Consent

The participants should provide informed consent before they are enrolled in the study. The informed consent form should describe the purpose of the study, the procedures, the risks and benefits, and the confidentiality of the data. The consent form should be written in language that the participants can understand.

The informed consent should be obtained from each participant before the study procedures are performed. The consent should be documented in the study records. The participants should be able to withdraw from the study at any time.

Data Management and Sharing

The data management and sharing policy of the NIH requires that researchers plan for the management and sharing of data generated during NIH-funded research. The data management and sharing plan should describe the types of data that will be generated, the standards that will be used, and the plan for sharing the data. The plan should be described in the grant application and should be updated as needed.

The data management and sharing plan should be consistent with the study protocol. The plan should describe how the data will be stored, preserved, and shared. The plan should also describe how the confidentiality of the participants will be protected.

Publication Ethics

The publication of the study results should follow the ethical standards of the Committee on Publication Ethics. The authors should be listed according to the authorship criteria. The data should be reported accurately and completely. The conflicts of interest should be disclosed.

The study should be reported according to the STARD 2013 guidelines for diagnostic accuracy studies. The guidelines should be used to ensure that the study is reported transparently and completely. The reporting should include the sample size calculation, the recruitment, the reference standard, and the results.

Professional Escalation Criteria

When to Consult a Statistician

The sample size calculation should be reviewed by a statistician if the researcher is not confident in the calculation. The statistician can review the assumptions, the formula, and the resulting sample size. The statistician can also advise on the analysis plan.

The researcher should consult a statistician if the study design is complex, such as a multi-stage design or an adaptive design. The statistician can help to plan the design and the analysis. The statistician should be consulted before the study begins.

When to Revise the Sample Size

The sample size should be revised if the assumptions change during the study. The assumptions may change if the prevalence is different from the expected value or if the sensitivity or specificity is different from the expected value. The sample size should be revised if the study is not feasible with the original sample size.

The sample size should be revised before the study begins if the assumptions change. The sample size should not be revised after the study begins unless the study design is changed. The revised sample size should be documented in the study protocol.

When to Stop the Study

The study should be stopped if the sample size is not feasible or if the study is not safe. The study should also be stopped if the test is performing much worse than expected. The stopping rules should be specified in the study protocol.

The study should be stopped if the sample size is not feasible. The study should also be stopped if the test is performing much worse than expected. The decision to stop the study should be made by the research team and the institutional review board.

Frequently Asked Questions

What is the difference between the precision-based and power-based sample size approaches?

The precision-based approach calculates the sample size needed to estimate sensitivity and specificity with a target confidence interval width. The power-based approach calculates the sample size needed to test whether sensitivity or specificity exceeds a minimum threshold. The precision-based approach is more common for diagnostic accuracy studies because the primary goal is usually to estimate the test performance.

How does prevalence affect the sample size for a diagnostic accuracy study?

Prevalence determines the number of participants with and without the condition. The sample size for sensitivity is the number of participants with the condition, and the sample size for specificity is the number of participants without the condition. The total sample size is calculated by dividing each component by the prevalence or its complement. A lower prevalence requires a larger total sample size.

What is the role of the reference standard in sample size calculation?

The reference standard determines the true disease status of each participant. The sample size calculation uses the expected sensitivity and specificity, which are defined relative to the reference standard. The reference standard should be the best available method for determining the true condition status. The reference standard should be applied to all participants.

How should missing data be handled in the sample size calculation?

Missing data should be accounted for by increasing the total sample size. The expected rate of missing data should be based on prior experience or a pilot study. The sample size should be increased by the expected rate of missing data. The missing data should be handled in the analysis according to a pre-defined plan.

What is the STARD 2013 guideline and why is it important?

The STARD 2013 guideline is a reporting guideline for diagnostic accuracy studies. The guideline provides a checklist of items that should be reported in a diagnostic accuracy study. The guideline is important because it promotes transparent and complete reporting of the study design, methods, and results. The guideline is available through the EQUATOR Network.

What is the role of the institutional review board in a diagnostic accuracy study?

The institutional review board reviews the study protocol to protect the rights and welfare of the participants. The board reviews the informed consent form, the risks and benefits of the study, and the data collection plan. The board approval should be obtained before the study begins. The study should be conducted according to the approved protocol.

How should the results of a diagnostic accuracy study be reported?

The results should be reported according to the STARD 2013 guideline. The report should include the sample size calculation, the recruitment, the reference standard, the index test, and the sensitivity and specificity estimates. The report should also include the confidence intervals for the estimates. The report should be transparent and complete.

What is the role of the data management and sharing plan in a diagnostic accuracy study?

The data management and sharing plan describes how the data will be managed and shared. The plan should describe the types of data that will be generated, the standards that will be used, and the plan for sharing the data. The plan should be described in the grant application and should be updated as needed. The plan should protect the confidentiality of the participants.

Using the Evidence

Related Diagnostic Guides

• Sample Size Calculation for Diagnostic Accuracy Studies: A Practical Guide
• Nanotechnology in Rapid Viral Diagnostic Tests
• Nested PCR: Enhancing Sensitivity and Specificity
• Biochemical Tests in Microbiology: A Practical Overview
• Clinical Chemistry Interference Studies: Design and Interpretation

References and Further Reading

• Research Methods Resources. National Library of Medicine.
• EQUATOR Network. EQUATOR Network.
• Core Practices. Committee on Publication Ethics.
• NIH Grants and Funding. National Institutes of Health.
• ORCID for Researchers. ORCID.
• Data Management and Sharing Policy. National Institutes of Health.
• Laboratory Biosafety Manual. World Health Organization.
• NCBI Literature Resources. National Center for Biotechnology Information.
• Sample size estimation in diagnostic test studies of biomedical informatics.. Journal of biomedical informatics, 2014.
• Blinded sample size re-estimation in a comparative diagnostic accuracy study.. BMC medical research methodology, 2022.
• Determining Risk of Falls in Community Dwelling Older Adults: A Systematic Review and Meta-analysis Using Posttest Probability.. Journal of geriatric physical therapy (2001), 2017.

This article is educational and does not replace validated laboratory procedures, institutional biosafety review, manufacturer instructions, or professional interpretation.