Bioequivalence and Bioavailability Forum

Hi Constantinos!

❝ I need to write an SOP regarding codification of samples, blinding and unblinding during the conduct of a BE study (clinical and analytical phase). Could anyone propose or highlight any specific guideline or other resource i should have in mind?

If you are with a full-service CRO it’s most important that the randomization cannot leak from the clinical to the analytical department. Labels of the sample vials should contain at least the following information: study code, subject ID, period, scheduled sampling time. Deliver the final analytical report to the biostatistician. If you send data in electronic form, make sure that the integrity of data can be assessed (e.g., a MD5 check sum). Avoid proprietary formats (like XLS); use CSV instead.^* The statistician should officially request the randomization from the PI. I use to estimate the elimination still blind for treatment and join the randomization with NCA’s results afterwards.

❝ Does anyone know whether MS Office Xcel has a random number generation function?

ElMaestro already gave RAND() above. More comfortable functions are available in M$-Excel’s Analysis ToolPak add-in (part of M$-Office – but not installed by default).
A simple randomization would use CHAR(RANDBETWEEN(65,66)) in cell A1. In cell B1 write =IF(A1='A','B','A'). Drag these two cells down to the number of subjects getting a two column list. You have to check whether the list is balanced (equal numbers of As and Bs). In cell C1 write =IF(COUNT(A:A)/COUNTIF(A:A,'A')=2,'balanced','imbalanced!'). Hit F9 until you get a balanced list.
For a better method see this example. Again, check whether the list is balanced (I suggest to use 1 000 000 in the code).
Example of a 3-period 6-sequence Williams’ design:
Subj. Seq.               Rand.
 1     1    697961    7    1
 2     2    876928    4    4
 3     3    518166   12    6
 4     4    115046   22    4
 5     5     72219   23    5
 6     6    472942   13    1
 7     1    966161    1    1
 8     2    876936    3    3
 9     3    592926   10    4
10     4    329605   19    1
11     5    253962    2    3
12     6    370378   16    4
13     1    688949    8    2
14     2    377098   15    3
15     3    687419    9    3
16     4    345694   17    5
17     5    269087   20    2
18     6    906461    2    2
19     1    558561   11    5
20     2    835174    6    6
21     3    447320   14    2
22     4    332128   18    6
23     5     35092   24    6
24     6    843748    5    5
Validating Excel-sheets is a nightmare. Play around with the random-functions first, and then write a macro – or better – a VBA-script.

---

• Type =PI() in cells A1:A3. Leave A1 in the default format (3.14159265), format A2 to 3 decimals, and A3 to 14. If you send this data as XLS, the statistician will get all data in full precision. But this is likely not what you have listed in your analytical report. If you format cells in Excel accordingly and save them as CSV, only the displayed numbers will by used.