1

I am planning a fixed-sequence 2-treatment (placebo followed by drug) crossover study. There is a Placebo run-in period of 15 days. Placebo is the control and has no effect on BP. The aim is to prove the efficacy of the medicine for treatment of blood pressure. I take one measurement while on placebo (A) (control) for one month. And then treatment is given for 1 year. Measurements are taken after 3 months and 9 months of treatment. After this, dose of medicine is tapered by 30% for one month, BP is recorded again at 10 months and the minimum dose required to maintain BP is given till 1 year and the final measurement is taken. Note is made about the experience of the recording nurses-junior or senior (>5 years’ experience). BP recordings are made in A/C and Non-A/C rooms for studying the effect of ambient temperature. There is no funding for the study. The trial is a double-blind study till the 3rd month's data collection and open-label after that because of practical difficulties in blinding the medicine. This fixed sequence 3-period 2-treatment (AB) crossover design study is done on about 150 patients. The hypothesis is that medicine normalizes BP, Non-A/C rooms increase BP and senior nurses’ recordings show lower readings.

DrWho
  • 799
  • 4
  • 12
  • 23
  • You created three tags here, could you please edit their descriptions to provide guidance in their usage? It's not clear to me from their tag names alone what they mean. – Firebug May 17 '20 at 16:32
  • I am a beginner in statistics. I am sorry, I could not understand your query. I could not find a way to edit the tag descriptions. Because no patient will agree to stop the drug when there is improvement, I had to make it fixed-sequence. Because there is long term additional effect appearing after 9 months, I thought of taking an additional recording of measurements at the end of 1 year treatment. I will be eager to provide any additional clarifications, is required. – DrWho May 18 '20 at 03:33
  • Nah, you're good. I came here through the moderating tools because I saw the new tags. You can keep them if you think there are no good tags that convey the sense of your question (keep in mind that it doesn't help people to see your question though, because they have no other questions in them and so no followers). It's just that, reading their names, I couldn't judge what they refer to and if there are synonym tags to merge perhaps. But it's fine, don't worry :) – Firebug May 18 '20 at 13:59
  • Thank you. I thought somebody would answer in a few minutes. I am surprised at the time being taken. – DrWho May 19 '20 at 02:35
  • Which hypothesis do you want to test with this experiment? What kind of noise/variations do you expect? What is the point of the placebo? What is your theory/hypothesis for presence/absence of the medicin? – Sextus Empiricus Jun 14 '20 at 12:24
  • How does this 'double blind' work when there is no randomisation of treatments? – Sextus Empiricus Jun 14 '20 at 12:34
  • It's not clear what was crossed over in this design. Just a cross-over from placebo to drug for _all_ patients between periods A and B? Or were some started on drug during period A and then crossed to placebo for period B? Or was there a placebo group maintained throughout the study? – EdM Jun 14 '20 at 18:35
  • The hypothesis is that compared to placebo, this drug (which is being used for other diseases so far) reduces blood pressure. The noise -There may be human / machine errors in recording BP. Initial response is seen in days but the maximum effect is seen after many months. So multiple readings are taken at various periods. Double blinding: Both patient & doctor are not aware of what is first and what is second for the first 3 months. Later it is open label because blinding for 9 more months is difficult in practice. – DrWho Jun 15 '20 at 02:17
  • EdM: Placebo was given during period A and drug during period B for all patients throughout the study because once a patient finds relief with the drug, he will not be willing to go to placebo and so dropouts cannot be prevented if randomization is used. Fixed sequence is used because it is unethical to withdraw a drug when patient responds well without any side effects. – DrWho Jun 15 '20 at 02:17
  • I do not believe that this design is able to test the hypothesis that, *compared to placebo*, the drug reduces blood pressure. The design of the experiment does not include comparisons where the drug is tested relative to the placebo (but maybe you have prior information on how placebo's work? you could test against that). You only use the placebo as a baseline, there is no measurement that can determine what the effect is of treatment with placebo. – Sextus Empiricus Jun 15 '20 at 08:18
  • Is this really the hypothesis that you want to test with this experiment? Or do you, less specifically, just want to test whether or not the patients got better (independent from whether it is due to the medicine or not). – Sextus Empiricus Jun 15 '20 at 08:21
  • Thank you for your interest. The placebo has no effect on blood pressure. Period A gives the Placebo baseline record. Period B is drug therapy. It is given for 1 year and 2 measurements are taken at 3 months and 1 year. Both these measurements will be compared with baseline placebo measurements. I am not comparing this drug with any another drug. – DrWho Jun 15 '20 at 11:34
  • @DrWho I am asking these questions in order to be precise and correct about your experiment and desired analysis. I am confused about the hypothesis that you want to test with this experiment. The question is relatively simple when you just want to show that the different measurements are different but you seem to be adding something into the mixture (a medicine) that is confusing and unclear. – Sextus Empiricus Jun 15 '20 at 15:55
  • Let us [continue this discussion in chat](https://chat.stackexchange.com/rooms/109418/discussion-between-drwho-and-sextus-empiricus). – DrWho Jun 16 '20 at 02:16
  • @DrWho you added extra information about parameters that you wish to control for, but now your question has entirely changed. In fact the question/questionmark is gone as well. – Sextus Empiricus Jun 16 '20 at 15:26
  • There is no actual question. I am voting to close it as needing focus. There are quite a few concerns with the design/approach in my opinion. But what is the specific question? – AdamO Jun 17 '20 at 18:53
  • @Sextus As more and more information becomes available, we have to make changes to the study design. Please let me know whether my selection of statistical tests is acceptable or not – DrWho Jun 18 '20 at 01:52
  • @AdamO Thank you for your comments. Please let me know what are your concerns with the design/approach. – DrWho Jun 18 '20 at 01:53
  • @DrWho I believe that you better get in close contact with a local statistician because a study design involves many little questions and good information about the background. – Sextus Empiricus Jun 18 '20 at 06:07
  • ASextus Thank you. I want to acknowledge your great constructive comments in improving my article when it is published. Can you please send your details to my email address nagabhushanarao@gmail.com – DrWho Jun 18 '20 at 06:13
  • I don't see any icon to award the bounty. Please advice. – DrWho Jun 19 '20 at 00:45
  • The 'human' factor, if one is in a study suspected to have a BP benefit, that alone is comforting and a stress relief which potentially could reduce BP. Instead, claim the purpose of the study is multi-fold relating to say vitamins, to perhaps reduce any psychological induced placebo effect on BP. I would recommend taking several BP readings during the day. – AJKOER Jun 20 '20 at 16:02
  • @AJKOER I totally agree. The human factor is quite a frequent confounder. – DrWho Jun 21 '20 at 01:53
  • I don't see any icon to award the bounty. Please advice. – DrWho Jun 22 '20 at 02:28

0 Answers0