nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Tolerance/Prediction Intervals and Norms ISO16269-6, USP 1210

Steph_Georges

Hi everyone,

A prospect (Eurofins Amatsi Analytics, subsidiary of Eurofins Scientific, specializing in analysis, testing and inspection services) asked me the following question: "Does JMP calculate tolerance intervals and prediction intervals according to Iso 16269-6 and USP 1210? Or does it calculate without reference?"

The documentation is not explicit on this subject:

https://www.jmp.com/support/help/en/18.0/#page/jmp/statistical-details-for-prediction-intervals.shtml#

https://www.jmp.com/support/help/en/18.0/#page/jmp/statistical-details-for-tolerance-intervals.shtml#

What can I answer? Thanks for your support

Accepted answers

Chandramouli_R

Tolerance intervals are used in the pharmaceutical industry for ensuring product quality and consistency. They determine the range within which a specified proportion of a population's characteristic lies, with a certain level of confidence. These intervals are applied in quality control, process validation, stability studies, QbD and regulatory compliance.

This is a cross-cutting topic related to QbD (ICH Q8, Q9, Q10), specifications (Q6A, Q6B), stability (Q1A, Q1E), FDA Process Validation guidelines, EMA guidelines, pharmacopoeias, and WHO GMP guidelines. They help ensure batch-to-batch consistency, set specification limits, validate manufacturing processes, and assess product shelf life. Additionally, they play a vital role in risk management and follow the below stated USP and ISO standards (depending on the jurisdiction).

ISO 16269-6 provides guidelines for calculating statistical tolerance intervals, which are ranges within which a specified proportion of a population falls, with a certain confidence level. Key concepts include the tolerance interval, confidence level, and proportion of the population. The standard assumes a normal distribution and highlights that larger sample sizes lead to more precise intervals. Calculation methods involve formulas using the mean, standard deviation, and sample size to determine the interval bounds. The K-factor, dependent on sample size and confidence level, is used in these calculations.

USP 1210 outlines statistical methods for validating pharmaceutical procedures to ensure product quality and consistency. Key concepts include procedure validation, tolerance intervals to confirm consistent quality, and prediction intervals to estimate future observations within a certain confidence level. It emphasizes the importance of normal distribution assumptions, adequate sample sizes, and defines confidence and reliability levels required for validation. Additionally, it uses control charts and capability indices to monitor and measure process performance.

While ISO 16269-6 focuses on calculating tolerance intervals, USP 1210 (being a compendial reference) provides a comprehensive framework using various statistical tools to validate processes and maintain product quality.

Both standards highlight the importance of adequate sample sizes and normal distribution assumptions for reliable results.

Aspect	ISO 16269-6	USP 1210	JMP
Normal Distribution Assumption	Assumes normal distribution for data.	Assumes normal distribution for validating procedures.	Ref: Statistical Details for Prediction Intervals (jmp.com) Assumes normal distribution for calculations.
Confidence Levels	Uses t-distribution for determining confidence levels.	Uses confidence levels to ensure reliable prediction intervals.	Uses t-distribution for confidence levels, aligning with both standards.
Sample Size	Critical for accurate interval estimation.	Emphasizes adequate sample sizes for reliable interval estimation.	Considers sample size for precise calculations.
Proportion and Population	Ensures intervals contain a specified proportion of the population.	Focuses on prediction intervals for future observations, inherently addressing similar principles.	Calculates intervals ensuring future observations align with statistical principles.
Procedure Validation	-	Emphasizes validating procedures meet quality criteria.	-
Tolerance and Prediction Intervals	Focuses on tolerance intervals containing specified population proportions.	Uses prediction intervals to validate future observations within a defined range.	Provides methods aligning with both standards for tolerance and prediction intervals.
Confidence and Reliability	Emphasizes statistical confidence for interval estimation.	Defines required confidence and reliability levels for process validation.	Uses confidence levels to provide reliable intervals, consistent with both standards.
Control Charts and Capability Indices	-	Uses tools like control charts and capability indices to monitor and measure process performance.	Supported in JMP.

Note: Pharma can go to notorious extents to disagree until documented otherwise (If it is not documented, it is not true). This is a lucrative application and use case area in pharma which would ease sales. Perhaps we should refresh the documentation and references to include these technical details, which will convince pharma users.

All comments

Chandramouli_R

Both standards highlight the importance of adequate sample sizes and normal distribution assumptions for reliable results.

Aspect	ISO 16269-6	USP 1210	JMP
Normal Distribution Assumption	Assumes normal distribution for data.	Assumes normal distribution for validating procedures.	Ref: Statistical Details for Prediction Intervals (jmp.com) Assumes normal distribution for calculations.
Confidence Levels	Uses t-distribution for determining confidence levels.	Uses confidence levels to ensure reliable prediction intervals.	Uses t-distribution for confidence levels, aligning with both standards.
Sample Size	Critical for accurate interval estimation.	Emphasizes adequate sample sizes for reliable interval estimation.	Considers sample size for precise calculations.
Proportion and Population	Ensures intervals contain a specified proportion of the population.	Focuses on prediction intervals for future observations, inherently addressing similar principles.	Calculates intervals ensuring future observations align with statistical principles.
Procedure Validation	-	Emphasizes validating procedures meet quality criteria.	-
Tolerance and Prediction Intervals	Focuses on tolerance intervals containing specified population proportions.	Uses prediction intervals to validate future observations within a defined range.	Provides methods aligning with both standards for tolerance and prediction intervals.
Confidence and Reliability	Emphasizes statistical confidence for interval estimation.	Defines required confidence and reliability levels for process validation.	Uses confidence levels to provide reliable intervals, consistent with both standards.
Control Charts and Capability Indices	-	Uses tools like control charts and capability indices to monitor and measure process performance.	Supported in JMP.

Steph_Georges

Thank you very much for your comprehensive reply!

peng_liu

I would like to discuss further about the "tolerance intervals" part of this question.
"Does JMP calculate tolerance intervals and prediction intervals according to Iso 16269-6 and USP 1210? Or does it calculate without reference?"

To directly address the question, there are two things need to clarify:

Where is the referred "tolerance interval" in JMP? Find it, and find its definition/calculation in documentation.
Is the calculation of THAT tolerance interval in JMP the same as what is described in the external reference?

The reason that those questions are important is that the definition and calculation of "tolerance interval" may not be unique, but meanwhile asymptotically equivalent. There is no right or wrong, but just unfortunately different industries follow different conventions depending on the choices of experts in their field. If there is a discrepancy, then it is worth investigating the difference. Blindly following external industry documentation could be a wrong path.

Third, it is possible that you may not find "tolerance interval" in JMP where the answer is needed. In that case, the answer can usually be found by finding confidence bounds of quantile estimations. The details can be expanded case by case, depending on which part of the JMP is relevant.