Bioequivalence and Bioavailability Forum

Hi Marcelo,

welcome to the 12^th member of the BEBA-Forum from Brazil!

❝ 1 - zero is zero.

Zero does not exist in practice. Not in bioanalytics. Very rarely in PK (by convention: the pre-dose concentration in the first period of a crossover study and by assumption: the one in higher periods).

1. After some shouting matches (some people call that “lively discussions”) already in the first Crystal City Conference (Arlington 1990) a consensus was reached that concentrations below the Lower Limit Of Quantification (LLOQ) should be reported with a non-numeric code (i.e., “BLQ”) – not zero.
   If someone reports 0 anywhere in the profile that is either a typing error or proof of lack of under­standing how calibration works. The LLOQ is the lowest concentration where we have still acceptable inaccuracy and precision (for chromatographic methods 20%/±20% and for LBAs 30%/±30%). Generally the LLOQ is ~3–5 times the Limit of Detection (LOD). In other words, you see something but should not use it. It‘s not zero – look at the chromatograms. The LOD itself is (depending on the definition) 3–6 times the response of the blank.
   Imagine: In theory you have a perfectly linear calibration curve through the origin. In 50% of analytical runs you will find a negative intercept and in 50% a positive one. Is it possible to back-calculate a response resulting in exactly zero? Yes, but only if the response equals the intercept. If the chromatographic data system is properly set up, the integration threshold generally is around the LOD. You will never be able to get such a peakarea.
   
2. Lack of understanding PK. You have a good bioanalytical method with an LLOQ of 1% of the C_max and a washout of ten half lives. At the second period of a crossover you have a residual concentration of ~0.1%. Will the analytical method be sufficient to quantify it? No way. Likely it will be even below the LOD (0.2–0.3%). But since the concentration exists it should be reported as “BLQ”, not zero.

❝ 2 - missing must be interpolated.

Agree. But: Throw the linear trapezoidal rule into the waste bin. Gives a positive bias in the distri­bu­tion / elimination phases. It is a miracle to me why I see it in so many studies. The reign of the pocket calculator is history.

❝ 3 - No exclusion for outliers.

Disagree (see 1.a.) and read my posts above. Regulatory views are not necessarily good science.

❝ Must be considered in ours SOP's.

Agree. No cherry-picking. Must be done before unblinding.