Skip to main content

Regulatory Affairs

Tales From FDA Site Inspections

It’s a summer Friday, so I’ll just send along this link to a story at Wired. It’s similar to this recent post, in that it details some of the ridiculous ways that small overseas manufacturers have tried to get around FDA site inspections. As a correspondent noted, though, the title of the Wired article isn’t quite right: it should be how these people tried to dupe the FDA, because all of these were caught in the act. You will experience the thrills of seeing some guy hustling down a hallway with a black trash bag, which he pitches into a stairwell when he realizes he’s being followed. Or arriving a day before the announced inspection, only to find everyone busily working away in the QC lab back-dating stacks of analyses, surrounded by Post-It notes attached to all the things that still had to be fixed before the FDA showed up. And more! See everyone on Monday. . .

17 comments on “Tales From FDA Site Inspections”

  1. Anonymous 1 says:

    Not sure if you have mentioned this, but the book “Bottle of Lies” by Katherine Eban paints a disturbing picture about the extensive fraud at generic drug manufacturers revealed by FDA inspections.

    1. NJBiologist says:

      It looks like Bottle of Lies is the source for the Wired book.

      I really want to read it, but the excerpts I’ve seen are disturbing. Like Bad Blood disturbing, except for products that often make it to the market.

      1. Anonymous 1 says:

        The book will definitely make your blood boil and be concerned about taking generics.

  2. Anonymous says:

    I meant to add this comment on the previous blog on this topic (“Got It All Ready For You, Mr. FDA Inspector, 25 July, 2019) but I’ll add it here, instead. It’s a bench researcher’s perspective cf., that of the assay or QC lab. Many years ago, many analytical instruments could be controlled interactively and manually. E.g,. with LC or GC methods, if you wanted to change the flow rate or gradient during a run or change detector sensitivity throughout a run or just re-zero a drifting or noisy baseline, you could do it. For “prep” runs on analytical columns, you could actually overlap injections and increase throughput. Then came the FDA and QC problems …

    In order to prevent tampering with evidence, instruments became more automated and now it’s almost impossible to “tamper” with hardware or data. That was driven by the need to prove that you didn’t fudge your data to the FDA. It was now impossible to manually override anything. You could no longer make overlapping injections. You could no longer even abort a run for any reason. For decades now, everything is recorded and logged by the software. And the software requires a ton of info for each sample with no backsies.

    This was all driven from the QC side of the market, which is bigger and more powerful than the research side. So now the researchers are stuck with the QC driven instruments. Some of the software is so complicated that you need a full time tech using it 40 hrs per week to know how to best optimize conditions that a researcher might only need to do 4 times per year (and have to relearn the software and the updates each time).

    But I think I prefer that PITA to letting QC data get manipulated for FDA submissions. It’s either that or go back to prep TLC.

    1. electrochemist says:

      I understand your perspective. My experience is that the degree of (in-)flexibility of the software largely depends on how it has been implemented at your company. Many pharma companies have set up more than one instance of their chromatography data systems (like Empower), with a locked-down version for use in QC and a more flexible installation for use in Research and some parts of Development.
      Having had assignments more than once in my career to try to “harden” the security of GMP lab systems, I can tell you from experience that audit trails can be disabled on many types of instrument controller software, and data can be deleted by users given admin accounts.

  3. Jon Davies says:

    Is there anything more the FDA can do to protect US consumers?

  4. DV Henkel-Wallace says:

    > “an investigator scrutinized the results of high-performance liquid chromatography tests”

    I suppose in this case the “P” HPLC means “performance art”.

    I have a lot of family in Maharashtra and have spent a lot of time there and sadly can believe some of these horrible stories.

    A few years ago I worked with a head of QC who loved all things Chinese (he was Chinese himself, educated in the US and had gone back to work there for a while): he loved Chinese alternative medicine and basically advocated for all things Chinese. Except he was deeply opposed to our sourcing any APIs from China. Hmm…

    (Of course it’s not like crazy things don’t happen in the US, Europe or other wealthy countries.)

    1. Athaic says:

      “it’s not like crazy things don’t happen in the US, Europe or other wealthy countries”

      If that could convince the Chinese authorities to let EU or US inspectors check the QC of TCM factories, I have no issue allowing a busload of Chinese inspectors go visit the Laboratories Boiron, in Lyon, France. They are rather (in-)famous for their homeopathic duck liver pills.

  5. Curious Cat says:

    My question relates to API manufacturing and would appreciate it if someone could direct me to the answer:

    I understand that HPLC is used for analysing the purity of APIs since it has a high degree of precision, but why do people use HPLC analysis for assessing intermediates? Can’t NMR spectroscopy be used, not for the final API but at least for the analysis of intermediates? NMR is not as precise as HPLC, but processing time is significantly faster. Are there any USFDA-accepted guidelines on how to validate an NMR instrument?

    1. Some idiot says:

      A very good question… the proper answer is a lot longer, but a shorter answer is a bit like this. BTW, I love NMR as a technique, and I think it could be used a lot more…! But…


      Quantification by NMR is actually very simple, easy (assuming the experimental setup is right) and effective. Therefore you can quite easily set up an automated approach in a regulated environment (there is decent software for this). The two factors you can test (or control) here are identity (is the compound what you think it is) and content (is the compound as pure as you think it is, on a w/w basis). In addition, you can set the system up to quantify known impurities.

      The problem is that if there are new impurities at low levels (like <1 % or <0.1 %) you may risk not seeing them by such a method (particularly an automated system). That is where HPLC is usually pretty good. It is (generally) intrinsically much better at picking up low levels of impurities than NMR (particularly <0.1 %).

      Again, I am painting with a broad brush here. You _can_ (even routinely) detect impurities at the low-ish ppm level by NMR (have done so many times). Plus, NMR will give you a way better chance of giving you a good idea as to what the impurity is. But (in general) HPLC is the go-to for analysis. But it is horses for courses. And I think NMR is a horse that could be used more.

      1. Anonymous says:

        Some comments on characterization, testing, purity, analytical methods, etc., not specific to APIs but more general. There are many issues to consider when doing analysis including specificity and sensitivity. GOOD analysts do their best to find everything relevant. I have known several groups that have deliberately performed analyses to hide, mask, or not detect impurities. I consider that to be UNETHICAL and FRAUDULENT.

        A few examples illustrate the concepts.

        If I ran a reaction in perfluorobenzene and I think there might by some C6F6 in my product, H-NMR is of no use in finding or measuring it. F-NMR would detect C6F6 but not my non-fluorinated compounds. C-NMR could pick up both. However, C-NMR sensitivity varies from C to C in different compounds. Primary Cs give stronger signals; tert-Cs give weaker signals. Without a lot of signal response curves, it won’t be easy to quantify.

        I knew a group that routinely characterized their samples using “low field” NMR (less sensitive than high field; less dispersion than high field which means that signals are more likely to overlap and not be spread out (dispersed), e.g., diastereomers overlap at low field and look like one pure compound but are spread out at high field). The PI would not allow researchers to get high field NMR without his permission … which would have exposed how bad the reactions were actually working.

        Suppose you’re doing HPLC with a UV-VIS detector tuned to some long wavelength. All of your hydrocarbon and non-absorbing impurities are suddenly invisible because the detector has been set to have ZERO sensitivity to those entities. You’ve got one strong peak for your UV-absorber but everything else is buried in the baseline. That does not mean “a single peak” is a 100% pure product.

        If looking for metal impurities (e.g., catalyst residue), ICP (inductively coupled plasma) or AA (atomic absorption) should pick it up with very high sensitivity and selectivity … unless the sample prep is not done correctly. (Yes, I’ve seen that, too.)

        GC FID detectors (yeah, old school) have higher sensitivity than TCD detectors but FID has low sensitivity to halocarbons. A “cheater” who wants to under-detect halocarbon contamination will use FID. LC refractive Index (RI) detectors have low sensitivity but are almost universal unless the RI of your impurity happens to closely match your baseline solvent. And so on.

        If looking for radioactivity (e.g., metabolic studies or just checking for lab contamination), you better choose the right detectors for what you think you are detecting. A Geiger Counter won’t detect tritium. LSC is mostly for beta emitters, not alphas or gammas. And so on.

        MS is usually pretty universal, pretty sensitive (sub-femtomole detection) but you do have to be aware of possible factors that could hide or alter a detectable compound or impurity. (I can’t think of a specific example, so I’ll make one up: your impurity is thermally unstable, you run it through a GCMS at a temperature that decomposes it on-column so it is never detected by the MS.)

        Good researchers and QC labs apply their skill and knowledge to make sure that they are properly using the correct methods to honestly and accurately characterize their samples and are looking for the right things the right way. Be wary of those who deliberately perform fancy science-y sounding alphabet soup tests but the WRONG tests in order to avoid detecting impurities and providing accurate characterizations.

        1. NJBiologist says:

          “MS is usually pretty universal, pretty sensitive (sub-femtomole detection) but you do have to be aware of possible factors that could hide or alter a detectable compound or impurity.”

          How about ion suppression? Every bioA person I’ve talked to has developed a really nasty tic on hearing the words “lithium heparin.” I don’t fully understand the issue (biologist here–I get to ask experts to work this out for me), but it sounds like LiHep impairs ionization.

    2. A Nonny Mouse says:

      You will find that the most advanced laboratories are using UPLC, which can have run times in the order of 5 mins and it will, most likely, be attached to a mass spec system to help identify your impurities.

    3. PharmaJohn says:

      The issue with NMR result is that there is a lot more to interpret than HPLC result. The manufacturer’s QC perspective is: did it pass spec? So until all the possible impurities and specs are given to them, it would be difficult to judge who’s cutting corners.

    4. Process analytical guy says:

      It’s actually a matter of money. QC labs in manufacturing sites are typically not as well equipped as your research labs. Process Control Tests and intermediate analyses need to be developed such that the manufacturing QC lab can perform them. NMR is way too expensive and MS detectors are rare in these settings.

  6. Joe psycho says:

    Please help me get some prednisone
    I am in a living hell and 40mg prednisone completely returns me to normal and finally enjoy something.

Comments are closed.