sudy ☆ India, 20180711 12:31 Posting: # 19032 Views: 1,529 

Hi all, I have a query regarding acceptance criteria for RSABE as per FDA guideline. As per progesterone guidance and FDA statistical guideline, 95% uppper confidence bound should be less than or equal to zero. Now, 1) less than zero mean negative value, it’s very clear. 2) Equal to zero, i have a doubt that it is exactly zero. Because in practical situation during statistical calculation, we never get exact zero (very less chance). So if we have a value e.g. 0.01 or 0.001 or 0.06, If we round this value we get zero. Can we consider it as zero? What FDA stands for this? FDA given clarity for 90% confidence interval but not for 95% upper confidence bound. Is there any reference to report value upto 2 digits after decimal place without rounding for 95% upper bound? If anyone has any experience on this please share. 
Helmut ★★★ Vienna, Austria, 20180711 14:31 @ sudy Posting: # 19033 Views: 1,230 

Hi sudy, » As per progesterone guidance and FDA statistical guideline, 95% uppper confidence bound should be less than or equal to zero. Now, » 1) less than zero mean negative value, it’s very clear. » 2) Equal to zero, i have a doubt that it is exactly zero. Because in practical situation during statistical calculation, we never get exact zero (very less chance). Is is so difficult understanding what the “≤” symbols means? ≤ 0 reads “less than or equal to zero” (equivalently: “not greater than zero“, or “at most zero”). » So if we have a value e.g. 0.01 or 0.001 or 0.06, If we round this value we get zero. » Can we consider it as zero? Nope. » FDA given clarity for 90% confidence interval but not for 95% upper confidence bound. » Is there any reference to report value upto 2 digits after decimal place without rounding for 95% upper bound? No rounding according to the FDA’s guidance. As you rightly stated, the FDA’s rounding rules are for the CI in percent. That means at the 4^{th} (<100%) or 5^{th} (≥100%) significant digit. IMHO, the same “logic” would mean a rounding cutoff at 0.0001 or 0.00001. If you want to go that way (which I don’t recommend) initiate a controlled correspondence with the FDA in order to avoid an RtR (see this post). — Cheers, Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. 🚮 Science Quotes 