28-Jul-2021 Analyse-it v5.90: Support for the updated CLSI EP6-Ed2 protocol and inverse predictions

Recently we’ve been busy updating Analyse-it to stay aligned with the latest updates to the CLSI protocols, and added a new inverse prediction feature.

If you have active maintenance you can download and install the update now, see updating the software or visit the download page . If maintenance on your license has expired you can renew it to get this update and forthcoming updates, see renew maintenance.

New CLSI EP6-Ed2

The CLSI recently released guideline EP06-Ed2 on the Evaluation of Linearity of Quantitative Measurement Procedures , which replaces the EP06-A published in 2003.

EP06-A relied on fitting a linear (straight line), 2^nd (parabolic) and 3^rd (sigmoidal) order polynomials to the data. A method was then determined to be linear or possibly non-linear based on statistical criteria. The degree of nonlinearity was then calculated as the difference between the linear fit and the best fitting non-linear model (parabolic or sigmoidal curves). Nonlinearity could then be compared against allowable nonlinearity criteria.

The new CLSI EP6-Ed2 protocol no longer requires fitting polynomial models to determine linearity. Instead, the deviation from linearity is calculated as the difference between the mean of each level and a linear fit through the data. That can then be compared against the allowable nonlinearity criteria. Other changes to the protocol include experimental design and there is now more focus on the structure of the variance across the measuring interval.

By default, the settings in Analyse-it are still configured for EP6-A as that is still in widespread use. However, to perform a new EP6-Ed2 analysis follow these steps:

1. Select a cell in the dataset.
2. On the Analyse-it ribbon tab, in the Statistical Analyses group, click Linearity.
3. In the Y drop-down list, select the measured variable.
4. In the By drop-down list, select the level variable, and then:
• If the values are identifiers, select the Identifier, and then in the Assigned values grid, under the Value column enter the values alongside each level.
• If the values are dilutions made by diluting a high pool or mixing high and low pools, select the Relationship, and then select Mixture, Dilution, or Addition depending on how the levels were prepared. In the Assigned values grid, under the Value column for the first and/or last level, type the value (intermediate values are automatically calculated using relative values).
• If the values are known/expected/assigned values, select Known values.

Note: Computation of linearity only requires the relationship between the levels, so you do not need to enter the assigned values if they are unknown.

5. Optional: To compare the nonlinearity bias against performance requirements:
• If the allowable nonlinearity bias is a constant or proportional value across the measuring interval, select Across measuring interval, and then: in the Absolute edit box type the bias in measurement units, and/or in the Relative edit box type the bias as a percentage (suffix with % symbol).
  
  Note: The allowable bias is the greater of the absolute bias and the relative bias for each level. So, with an absolute bias of 5mg/dL and a relative bias of 10%, the allowable bias will be set at 5mg/dL for all values 0 mg/dL up to 50mg/dL and then at 10% of assigned value for values above 50mg/dL.
• If the allowable nonlinearity bias varies for each level, select Each level and then in the Allowable nonlinearity grid, under the Absolute / Relative column alongside each level, type the bias in measurement units or the bias as a percentage (suffix with % symbol).
6. On the Fit Model panel, in the Fit drop-down list, select Linear.
7. In the X drop-down list, select Expected values and in the Y down list, select Mean.
8. In the Weights drop down list, select:
   • None - Fit an ordinary regression. Use when measurement procedure exhibits constant SD over the measuring interval.
   • Var(Y level) - Fit a weighted regression with weight based on the variance at each level. Recommended when number of replicates per level is 4 or more.
   • Var(Y level pooled) - Fit a weighted regression with weight based on the pooled variance over a subinterval of levels. Recommended when number of replicates per level is 2 or 3.
   • VarFn(Y) - Fit a weighted regression with weight based on the variance function at the mean of each level. Recommended when the precision can be modeled by a variance function.
9. If the levels are made by a dilution of a high level, select Force through zero check box. Otherwise, if the levels are produced by mixing a high and low level, clear the Force through origin check box.
10. Click Calculate.

New Inverse prediction feature

This release also includes a new inverse prediction feature for simple linear regression models. Inverse prediction has several uses including estimating the shelf-life of a product. In the context of CLSI protocols it is useful for CLSI EP25-A – Evaluation of Stability of In Vitro Diagnostic Reagents .

To make an inverse prediction use Fit Model to fit a simple regression model, then:

1. Activate the analysis report worksheet.
2. On the Analyse-it ribbon tab, click Predict, and then click X given Y.
   The analysis task pane Inverse Predict X given Y panel opens.
3. In the Predictions list box, under the Y column, type the values for the response variable.
4. Optional: In the Confidence interval edit box, type the confidence level, and then in the drop-down list select the confidence bounds. Then in the Method drop-down list, select Simultaneous for intervals that ensure you achieve the confidence level simultaneously for all predictions or Individual for intervals that only ensure confidence for each individual prediction.
5. Optional: On the Fit panel, in the Predicted values drop-down list, select the style to plot the predicted values on the scatter plot.
6. Click Recalculate.