Sep 08

The MIC: A Misunderstood Animal?

I read recently a paper which studied ‘relevance’ of MIC data against clinical outcomes. The conclusion of the paper as that there is no link between MIC data and clinical outcomes in the case of vancomycin for left sided endocarditis. While there may be some merit in the claim that the MIC is relied on a little too heavily for presumptive clinical outcomes, this bought into sharp contrast something that is deeply concerning for someone who works predominantly with in vitro MIC testing. It is almost always vilified by people who don’t truly understand it.

Let me start by saying this is not a direct criticism of the paper I’m referring to. It seems to be very well written and the majority of the data supports their conclusions, but there are a few decisions with study set up which as an experienced MIC scientist confuse me. For instance the fact that they have decided that the cut off for low MIC/high MIC is 1.5 mg/L. The CLSI and EUCAST breakpoint is ≤2 mg/L and the EUCAST distribution tables show a very clear end of the wild type population between 2 and 4 mg/L. So first off I’m confused why they chose 1.5 mg/L and not 2 mg/L. Also they chose to establish MIC by eTest. eTests are great in a hospital setting because they are a quick, flexible, easy option to test a rough MIC which in turn will tell you if you are dealing with a resistant infection, but if you are doing a solid comparison of the MIC value I’d argue they are not accurate enough. Also they are very easy to run and read incorrectly, providing incorrect data particularly when you are trying to read on or around the clinical breakpoints. Again I am not suggesting this is the case with the paper but is a possibility.


All of that is beside the point. As I say, the paper itself is not the problem. The problem is that the MIC is seen as a ‘simple’ test. That in part is due to the fact that the science behind it is inherently also very simple. Yet that simple data forms a cornerstone of drug development from early stage in vitro testing, preclinical in vivo testing alongside PK/PD work to ascertain dosage and on into clinical trials. It is also used to determine development of resistance and set clinical breakpoints through surveillance.

You are literally just throwing in a culture of the test pathogen in with the test compound and if the bug is inhibited by that, it doesn’t grow and it spits out a value. Simple, right?

Having spent the past 10 years carrying out hundreds of thousands of MIC tests (mainly broth and agar dilution but the odd eTest too) I can tell you it is far from simple. Take for example the standard guideline methods for carrying out these ‘simple’ tests. Both the CLSI and EUCAST produce massive tomes and have put in countless amounts of validation and research into establishing as near to a reliable and solid method as you can get for the humble MIC. They state highly specific test conditions, quality controls, reagents and strict methods. Why do they do this? Because without them the method is not reproducible and breakpoints cannot be applied to clinical outcomes. The guidelines have to be followed as stated, otherwise all you have at the end of your testing is a value, which has no data whatsoever linking it to a clinical outcome.

There is more, in that the test is not a simple one to interpret. Okay, for the example in the paper above a guideline method dilution MIC for MSSA against vancomycin is often easy to read. For example, you need to be aware of the extended incubation period for vancomycin when compared to other test compounds (16-20 vs 24 hours). That’s because the resistance mechanisms take longer to kick in for this bug/drug combo. Sadly this is not something you’d know unless you’d done a lot of research, are following a guideline method to the letter or are experienced in the test method. Also if you are testing a worthwhile amount of clinical strains, you’ll always end up with a few ‘freaks’ which require a bit of experience to read correctly because they’ll do some strange things on the test plates. You might think that for a small number of isolates it’s not worth the concern, but remember that what you are looking for with the test are those isolates with MICs that differ from the wild type, which are usually the ‘freaks’. That makes interpreting those strains most important of all. Then there are all of the ways a test can fail. You can have failed growth for a whole host of reasons, contaminated cultures which are hard to spot with the untrained eye, trailing endpoints, bubbles in the media and/or incorrect isolate identification can happen even if the most experienced analyst has carried out the test. Still think it’s simple?

I’m not suggesting for one second that the MIC is perfect. In fact I and many others are working to develop alternative tests that may one day replace it. However right here and now, it does a better job that it is given credit for, but only if it is done right. All too frequently it isn’t and the differences between tests run by MIC experienced scientists and those who are new to the technique are significant. When the data produced is found to not match up with expected results or clinical outcomes, the method itself is blamed.

The dilution MIC is still the best method we have for determining bacterial resistance. Although other methods are popping up all over, they are commonly expensive, restricted in what they can show or unreliable. It also, despite its weaknesses, still forms a cornerstone of drug development testing. Meaning scientists time and time again come back to the MIC, which is part of the reason why it hasn’t been replaced in decades. Surely that has to stand for something?

I am begging for someone to show me a better way of accurately, affordably, reproducibly and reliably determining antimicrobial activity. But I haven’t seen it yet. That is as frustrating for me as for any other scientist, but please, don’t take it out on the MIC.



Ed Siegwart is a Senior Scientist for LGC’s Drug Development Services

Leave a Reply

%d bloggers like this: