We conducted clinical study of S-007 in two phases. Phase-I involve single dose treatment of S-007 ( dose- 500 mg tablet). In phase-II we administered PL-45 drug along with S-007 tablet on same patient. Now we have plasma level of S-007 for checking any drug interaction.
Currently i am facing problem in data analysis because patients were already on the treatment of S-007, so for 0 hr each patient has different drug level. How to check drug interaction for S-007?
Thanks for reply. I am handling first time population data so i am unaware of interaction model differnital equation. I want to check whether the treatment of PL-45 has affecting the plamsa level of S-007 or not.
I’m rather puzzled by your question, it seems that you’re saying that there was no, or insufficient washout between your treatment periods, so you have postive predose for phase II for your S-007 compound, yet I aslo see it in Phase I of your data? Was this expected by your protocol ? Was PHase II at some approximation of steady state? quite often higher concs are seen in phase I that phase II
What was your SAP (Statistical Analysis Plan) for this? a simple ‘BE’ type assessment?
Phase-I and Phase-II also have positive predose level.
Patients were on the therapy of once a day single dose treatment of S-007 for 15 days as per our protocol. Phase-I is nothing but the sampling of these patients for 0-24hr on the 6th day so these patients also have postive predose level.
On the 13th day we adminstered another drug P-45 along with single dose of S-007 i.e. phase-II.
Now we want to know, is there any statistical difference in the pharmacokinetic of S-007 due to concurrent administration of P-45.
Sachin, Are you asking for a statistical consultant’s opinion? (I’m not one - but you may find you need to pay). My point is that it seems dubious scientifically and ethically to run a study before deciding how you will define/measure the outcome, there’s a quote;
“You can’t fix by analysis what you bungled by design.” from a book 'By Design’ by RJ Light, JD Singer & JB Willett.
Looking briefly at your data I would say exposure is so variable, phase to phase that it would be hard to identify a statistical difference between these treatment periods, I am not sure that is the question you are asking though.
Dependent Units Subject Phase-I Phase-II Test_Ref Ratio_PctRef
Ln(Cmax) A 1380 533 -847 38.623188
Ln(Cmax) B 1140 1000 -140 87.719298
Ln(Cmax) C 576 1400 824 243.05556
Ln(Cmax) D 764 863 99 112.95812
Ln(Cmax) E 1010 1520 510 150.49505
Ln(Cmax) F 2020 1830 -190 90.594059
Ln(Cmax) G 707 1260 553 178.21782
Ln(Cmax) H 1320 1360 40 103.0303
Ln(Cmax) I 424 345 -79 81.367925
Ln(Cmax) K 1270 629 -641 49.527559
Ln(AUClast) A 9218.0431 4341.3202 -4876.7229 47.095898
Ln(AUClast) B 6701.8075 6217.5432 -484.26426 92.774125
Ln(AUClast) C 3017.1795 5596.4349 2579.2554 185.48565
Ln(AUClast) D 4449.5352 4252.4958 -197.0394 95.571686
Ln(AUClast) E 5192.2851 6311.5582 1119.2731 121.55646
Ln(AUClast) F 10908.761 6389.6497 -4519.1118 58.573558
Ln(AUClast) G 4221.9838 4629.5077 407.52392 109.65243
Ln(AUClast) H 6359.6183 7147.8373 788.21898 112.39412
Ln(AUClast) I 3342.7754 2062.3691 -1280.4063 61.696311
Ln(AUClast) K 7466.7948 3771.0433 -3695.7515 50.504178
Good idea! Your design is completely wrong for a DDI study. There are established rules for its design for more than two decades. The current versions:
Steady state is mandatory. Seems that you were in steady state for drug 1. But: Then you have to saturate with drug 2 as well (while still administering drug 1).
Your design is paired (1 followed by 1+2 in all subjects) and therefore, useless. AUCs in Phase II are ~85% of Phase I. Does that mean you can conclude that there is an interaction (drug 2 reduces concentrations of drug 1)? Not at all. Imagine: You have no (zero!) “true” interaction, but a period effect of minus 15% (AC broke down, phase of the moon, whatsoever). What will be the treatment-estimate? Correct, 85% – although it should be 100%! Hence, a cross-over study (which will correct for period-effects) is mandatory. BTW, stated in all guidances. @Simon; I disagree. The CV of AUC is ~32%. Of course 10 subjects are too few. But given the variability, a properly designed DDI study is not that difficult. Cmax with ~41% is more tricky. Reference-scaling would help.
Last, but most important:
You are giving drugs to patients. Before you do so the next time, ask yourself:
– Would I take it myself?
– Would I give it to my wife/husband/etc.?
– Would I give it to my children? Only if the answer is yes to all questions, go ahead with the next point.
Take a course in medical ethics. Seek statistical consultancy. Never start any study before you understand a good part of the statistics behind. If you have difficulties in understanding what the statistician tells you, go for a second opinion.
Sadly enough, 90% of studies on my desk failed by design and go straight to the waste bin.
To quote RA Fisher:
To call the statistician after the experiment is done may be no more than asking him to perform a postmortem examination: he may be able to say what the experiment died of.
The CV of AUC is ~32%. Of course 10 subjects are too few. But given the variability, a properly designed DDI study is not that difficult. Cmax with ~41% is more tricky. Reference-scaling would help.