Wednesday, August 5, 2015

Process Wednesday: stress tests aren't just for the Treasury

Credit: Practical Process Research and Development
From our mentor-by-literature Neal Anderson's second edition of "Practical Process Research and Development", a good thought about stress tests of plant processes on page 416:
With stress tests reaction mixtures are purposely exposed to more extreme conditions to assess the potential impact. For instance, a key reagent may be added in portions to determine wither higher level of impurities will form due to micromixing. Another example of a stress test is shown in figure 15.1 (CJ's note: above). Merck researchers anticipated that by using n-BuLi as base, small amounts of excess base could lead to metalation of one aryl ring, leading to a bis-aldehyde. When two equivalents of n-BuLi was charged in a stress test, 10% of this side product was indeed formed. Further screening showed that EtMgBr could be used as a base, even in 20% excess, avoiding the need to titrate and control the addition of the n-BuLi solution. 
Every once in a while, you have this terrible "what if" dream where the horrible "what if" scenario is "what if the operator makes a math error and adds twice as much as they should of a key reagent?" Performing these sorts of stress tests (and yes, math errors definitely happen) are a good idea and an important part of developing a plant process.

This section also talks about identifying a "normal operating range" and a "proven acceptable range" for operating parameters - I think part of the issue is that the temptation is to make the proven acceptable range to be the norm, and then mistakes happen, and then you're into "here be dragons and impurities and reworks" land. 

12 comments:

  1. In our place, once we had fixed the synthetic route and the drug substance specifications such stress test were the norm. We would, in lab experiments, look for a range of conditions,, temp, time, concentration, rate of addition, etc that would still give the acceptable drug substance according to the specs.
    Sometimes the range was extremely narrow, sometimes wide. But anything inside that range was acceptable. We kept on going until we found the parameter that did not give the correct quality.
    This involved a tremendous amount of experimentation and we included pilot runs in the data. This info was part of the package that was handed over to production.
    In production they continued to expand the information.
    It is vital because in production errors do occur through nobodies fault, just human error. So we could always refer to the document to ensure the production run was producing defined material.

    ReplyDelete
    Replies
    1. As a former DOE wonk wannabe, just wanted to point out that frequently it's not just an issue of a single parameter but rather interactions between parameters that are the sources of undue stresses - sometimes these aren't even on the investigators' radar...

      Delete
    2. Sure thing. But when you start altering two parameters at once and get an unexpected result, which one caused it? So, while I agree, you can't really take this approach.

      Delete
    3. Not wanting to have an argument here, but yes, sometimes it's the interaction between two variables rather than the action of one parameter. Design of experiments allows you to quantify this - and yes, you can take an approach that allows you to model the interactions.

      Delete
  2. Here's what Shel Silverstein has to say (apologies for the children's poetry; I just cannot resist):

    Last night, while I lay thinking here,
    Some Whatifs crawled inside my ear
    And pranced and partied all night long
    And sang their same old Whatif song:
    Whatif I'm dumb in school?
    Whatif they've closed the swimming pool?
    Whatif I get beat up?
    Whatif there's poison in my cup?
    Whatif I start to cry?
    Whatif I get sick and die?
    Whatif I flunk that test?
    Whatif green hair grows on my chest?
    Whatif nobody likes me?
    Whatif a bolt of lightning strikes me?
    Whatif I don't grow taller?
    Whatif my head starts getting smaller?
    Whatif the fish won't bite?
    Whatif the wind tears up my kite?
    Whatif they start a war?
    Whatif my parents get divorced?
    Whatif the bus is late?
    Whatif my teeth don't grow in straight?
    Whatif I tear my pants?
    Whatif I never learn to dance?
    Everything seems swell, and then
    The nighttime Whatifs strike again!

    ReplyDelete
    Replies
    1. Ohrwurm

      http://andrewhammel.typepad.com/german_joys/2005/04/german_word_of__1.html

      Delete
    2. Also, "das Kopfkino" - lit. "head cinema" - those rolling pictures in your head that predict with such unfounded certainty the most unfortunate outcomes of a meaningful situation or event.

      I guess it's true, those Germans really do have words for everything...

      Delete
  3. Not sure if the scheme shown is directly from the book or redrawn here (and obviously not the main point of the post), but the formaldehyde electrophile shown would not give the aldehyde product. Is it DMF or something else?

    ReplyDelete
    Replies
    1. You know, that could have been my error. I'll check tonight.

      Delete
    2. I'd figure DMF, but I assume Grignards can act as hydride donors if the environment is hindered enough?

      Delete
    3. I checked the book briefly, and "HCHO" was what was written in the scheme; I'm going to dig further...

      Delete
    4. Looks like an error in the book. Scifinder says that in the original patent and 05 OPRD paper the electrophile is indeed DMF.

      Delete