How much specific knowledge of a trial reqd to extract test results

#1
This concerns the data from a controversial medical trial (involving over 600 participants) in which the thresholds for the outcome measures were altered after the data had been collected, though not analysed.

When some of the patient-level data were released after an FOIA request and reanalysed, the claims made by the trial investigators were not substantiated.

I have made an FOIA request for patient-level data for six measures (questionnaires, scale scores, etc). I have asked simply for the results of the tests and not for any kind of processing of those results or application of any thresholds. It is said that SPSS and Stata software is required.

The trial was large and publicly funded. The Chief Statistician was of considerable experience.

The University is claiming the following. To fulfil this request it is absolutely necessary to involve someone who was involved in the trial and as no such person remains at the University, it should not be required to provide the data. The data are 'very specific sets', so that someone familiar with both the software and the terminology or notation used in this specific database would be needed to run queries to attempt to locate/retrieve the correct data. Only someone who worked on the trial would understand the structure and terminology of the data and so be able to provide the information.

It is also claiming that it would take 3 days' work to extract the results or 6 days if a statistician unconnected to the trial was guided by someone who was.

I am surprised by these claims.

I have no knowledge of statistics. If I obtained the database schema, data dictionary (or descriptions of the data) and file structures for the data, would a qualified statistician be able to check their claim? Would they be able to say for sure what needed to be done? If not, exactly what else would I need to ask for? How easy would it be for the University to extract the schema, dictionary etc if the data have been archived?
 

hlsmith

Not a robit
#2
Interesting situation. So the University is pretty much saying they don't have the prior code or a good enough data dictionary to procure the subset of data you are requesting without considerable effort and time. What is so complicated about the dataset? Did they collect lots and lots of time series data? Are you requesting a subset of subjects or just a subset of variables for all subjects? Provide some more information please. Also, would you be will to link they clinicaltrials.gov url for the study or the generated manuscript, so we have an idea of what you are referencing.
 
#3
Interesting situation. So the University is pretty much saying they don't have the prior code or a good enough data dictionary to procure the subset of data you are requesting without considerable effort and time. What is so complicated about the dataset? Did they collect lots and lots of time series data? Are you requesting a subset of subjects or just a subset of variables for all subjects? Provide some more information please. Also, would you be will to link they clinicaltrials.gov url for the study or the generated manuscript, so we have an idea of what you are referencing.
Thanks for your reply.

The protocol for the trial: https://bmcneurol.biomedcentral.com/articles/10.1186/1471-2377-7-6

The original paper: https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(11)60096-2/abstract

The reanalysis: https://bmcpsychology.biomedcentral.com/articles/10.1186/s40359-018-0218-3

A critique of the trial by Stats USA: http://senseaboutscienceusa.org/pace-research-sparked-patient-rebellion-challenged-medicine/

A journal special edition on the trial: http://journals.sagepub.com/toc/hpqa/22/9

I have requested:

Please provide for the following patient-level data for baseline, 24-week and 52-week assessments, where available.

EuroQOL (EQ-5D) scores.

Hospital Anxiety and Depression Scale scores in both anxiety and depression sub-scales.

Work and Social Adjustment scale scores.

Borg Scale scores

Physical Health Questionnaire 15 items (PHQ15) scores.

Client Service Receipt Inventory scores. (Please also include the 6-months-prior assessment.)

Hope that's enough information. If there's more that is needed, let me know.

Thanks for your help.