"Quality control" for programmers in the pharmaceutical industry

July 31, 2016 12:10 PM

I recently applied for a job in the pharmaceutical industry as a statistical programmer. I was told that if I successfully move to the next stage of the selection procedure, I will need to set a "quality control" test, in addition to a programming test. I would like to know "quality control" means in this context, so that I can try to prepare myself.

I am interested in whether "quality control" might have a specific meaning in this industry for this kind of job. Is there a body of regulatory knowledge, or knowledge related to clinical testing that one could reasonably expect a statistical programmer to have? Or, could a "quality control test" mean some sort of generic test of error checking abilities for programmers? If it seems likely to be the former, could somebody suggest any relevant reading I could do on the subject to prepare myself? I have experience writing statistical scripts but I have never worked in the pharmaceutical industry. Neither do I have a legal or hard science background. In case it is relevant, the company is large and has a US presence.

posted by mister_kaupungister to Work & Money (8 answers total)

I can tell you from working in a hospital that many medication errors are caused by a mix of bad input not being error checked, and the many many alerts that care providers see on the screen that they ignore. (alert fatique). For example if a medication is prescribed based on the kilograms of the patient, (10mg for every 10kg) but the patient is very obese (160kg) , the doctor may say 100mg because it is the max are dosage before problems with other medications or overdoses appear. The computer could "correct the doctor" and kill the patient, based on the kg instructions alone, pop up a question about if 100mg will be effective which could add to error fatique or some other thing.

Machines do much of the actual depensing and counting of medication these days, and there are huge issues about computer trust and people die when the humans and the compuer does not notice.

Please note I am a social worker and would never actually be applying for your job in anyway, so get other information from others as well.
posted by AlexiaSky at 1:21 PM on July 31, 2016

I'm not familiar with that industry but "software quality" is a software engineering thing if the quality control reference is to the output from the programming work, as opposed to quality control of a pharmaceutical manufacturing process or something like that.
posted by XMLicious at 1:25 PM on July 31, 2016

don't know anything about pharm industry, but i would suggest you know what the following mean, and what libraries and/or tools are used to implement them: unit, integration, regression, black- and white-box testing; source control (eg git + github); code review; continuous integration; test coverage. those are the things i'd expect a generic quality test/interview to focus on.
posted by andrewcooke at 2:06 PM on July 31, 2016 [2 favorites]

It is common for statistical programming outputs (datasets/tables/listings/figures) to be programmed twice as a QC measure: once by the development/production programmer, and again by a validation programmer. Both programmers are given the same set of specifications and results are compared to ensure agreement. If disagreement arises, either someone made a mistake or the specifications were insufficiently detailed and interpreted differently. Critical review of code for best practices or errors is often part of the validation process. I can imagine a scenario where you are asked to review code/outputs for errors and agreement with specifications. However, I have not encountered such an exam myself.

ICH guidelines may be of interest to you. For example, E9 on statistical practice in clinical trials. The ICH document on good clinical practice (E6) describes the responsibilities of a clinical trial sponsor in conducting a trial and reporting results. These are less immediately applicable to your situation, but you can also find a list of FDA regulations related to clinical practice here.

Credentials: I worked for pharma clients as a biostatistician at CRO and am now in public health.
posted by esoterrica at 2:27 PM on July 31, 2016 [4 favorites]

You may also benefit from familiarizing yourself with data standards: CDASH and ADaM. I've linked you to the organization that manages the standards, but PharmaSUG or SAS Global papers may give a friendlier introduction.
posted by esoterrica at 2:28 PM on July 31, 2016

First, I think it's fair to ask. It could mean quality control in general, or some specific protocol.

In addition to the possibilities mentioned above, if you are looking at the manufacturing side, they could mean Six Sigma (getting to be old hat) or IS0 9000 or something equivalent.
posted by SemiSalt at 4:53 PM on July 31, 2016

I agree that the wording in your post could be taken a few ways. But an interpretation, and from my perspective an important one is the chain of of events starting with the FDA grappling with integration of electronic records into NDAs (New Drug Applications, historically an unfathomably large paper document), most familiar in the context of 21 CFR Part 11 to the current mandate of the eCTD (electronic Common Technical Document). For the FDA these are guided through the cGxP's, and in your case specifically cGCP, or Good Clinical Practices.

Coming out of this is essentially a formalised quality management and testing methodology known in the Pharma industry as Computer Systems Validation (CSV). It overlays most SDLC methodologies with the outcome of being Part 11 compliant.

I worked as a management consultant and program manager mainly around management of NDAs / CTDs. Not explicitly with the statistical systems themselves. I did find an interesting paper on SDTM. You'll notice toward the end of the document them referencing validation as a part of tracking person hours.

Of course, there are six overarching Quality Systems in Pharma. I don't think they would expect you to be across all of that.

When I was involved in staffing developers, I would want them to understand the principles of CSV and how it impacts their systems and chosen SDLC. I would want them to understand certain documents such as the Validation Master Plan (VMP) and its subdocuments as it relates to CSV and software development as well as how to work within highly-structured documentation expectations and with a, sometimes pedantic, expectation on traceability and testing.

Hopefully I haven't confused things too much. I've been retired for ~5 years and am a little rusty on my explanations. I've used a few wikipedia links for convenience. They are not the definitive source, but are reasonable starting points for some of the vocabulary.

Hope this helps and good luck.
posted by michswiss at 6:45 PM on July 31, 2016 [2 favorites]

Response by poster: Some useful answers. Thanks all.
posted by mister_kaupungister at 2:17 PM on August 2, 2016