You're addressing two different problems: the cut off of a given test and the clinical cut off using that test. To determine the intrinsic detection threshold of your test, you don't need control subjects. But to assess clinical value and cut off you need an healthy population and both its mean and SD, assuming a gaussian distribution. Your cut off can  then be chosen  at two SD for 5 % positive samples or higher....   J.

If this a preliminary assay, I would suggest to use ROC analysis and then choose the cutoff value based on the highest Youden's score.

ROCR and pROC packages for R are very useful packages to do such analysis.

Edit: I agree with @Jacques. Limit of blank and limit of detection are different from clinical cutoff.

You're addressing two different problems: the cut off of a given test and the clinical cut off using that test. To determine the intrinsic detection threshold of your test, you don't need control subjects. But to assess clinical value and cut off you need an healthy population and both its mean and SD, assuming a gaussian distribution. Your cut off can  then be chosen  at two SD for 5 % positive samples or higher....   J.

First select at least 100 healthy individuals or who lack the disease under study. Subsequently to the average readings +/- 2 standard deviations.

Hi,

If using kits, ELISA assays limit values are mentioned in the prospectus of the kits provided by the manufacturer. Additionally, you should use the standard concentrations (known) to plot a calibration curve / standard.

Statistically, you can analyze percentiles 5-95 to determine the positive and negative values. Levels of testing of controls could be used as cutoff values. For example, at 5 percentile, the mean level of a given cytokine is 3 pg / mL. Thus, 3 pg / mL will be considered as cutoff value. The upper levels will be considered as positive values.

This problem is 2-fold.  First, you must define the analytic sensitivity (LoB, LoD and LoQ) of the method.  Once this is done you proceed to establish the clinical cutpoint.  This can be determined by a number of methods including analysis of  a normal population and setting the 95% as the cutpoint, or by performing ROC calculations using defined populations of normals and abnormals to establish clinical sensitivity and specificity.

I agree with Guadalupe Garcia-Elloriaga. About standart deviations - may be 1SD, 2SD, 3SD. Its your decision for your ELISA-kit.