Saturday, 28 January 2017

The perils of in vitro drug testing

I've discussed this issue before. It comes up regularly because people often want to base treatment decisions on studies in which some drug or supplement seemed to have potent effects on cancer cells in a dish.  The problem is that the scientists carrying out these studies often seem to pay no attention to the achievable levels of the drug or supplement in human plasma or other tissues in the body with reasonable dosing schedules.  It's fairly common to see studies showing marvelous effects, without mention that the drug is being tested at 100 or 1000 times the achievable level in the body.

As a case study, here is one example:

Genistein Suppression of Matrix Metalloproteinase 2 (MMP-2)and Vascular Endothelial Growth Factor (VEGF) Expressionin Mesenchymal Stem Cell Like Cells Isolated from High andLow Grade Gliomas

"Results: Expression of MMP-2 demonstrated 580-fold reduction in expression in low grade glioma cells post treatment with genistein compared to untreated cells (P value= 0.05). In cells derived from high grade lesions, expression of MMP-2 was 2-fold lower than in controls (P value> 0.05). Genistein caused a 4.7-fold reduction in VEGF transcript in high grade glioma cells (P value> 0.05) but no effects were evident in low grade glioma cells. Conclusion. Based on the data of the present study, low grade glioma cells appear much more sensitive to genistein and this isoflavone might offer an appropriate therapeutic intervention in these patients. Further investigation of this possibility is clearly warranted."

Oh, and by the way, the study states that genistein was tested at a variety of concentrations: "variety concentrations of genestein (Sigma, UK) (0, 0.01, 0.004, 0.002 and 0.001M)"

This is actually not much of a variety of concentrations because all these concentrations are absurdly high.  M stands for molar (moles/liter).  0.001 M is therefore 1 millimolar.  The only drugs I'm aware of that even come close to 1 millimolar in the plasma are lithium and phenylacetate and maybe DCA with high doses.  Most drugs are achievable in the nanomolar or low micromolar range at best. For genistein we'll be generous and include glucuronidated genistein (the form predominantly found in the plasma which may or may not have pharmacological activity compared to free genistein).  Even still, concentrations of around 2 micromolar is about what is achievable in the blood with a single 300 mg dose.  Let's be generous again and assume that 10 micromolar total genistein would be achievable with a high dose.  The 1 millimolar tested in this study is 100 times higher the 10 micromolar achievable at best, including glucuronidated and free genistein.  In many cases the only way to match concentrations used in vitro would be high dose intravenous injection, but the safety of that would be unknown without phase 1 clinical trials.  It's possible that 1 millimolar free genistein in the bloodstream would be very toxic, and that normally non-toxic substances would become toxic at these high concentrations.

I've also seen studies testing melatonin up to 1 millimolar, whereas the maximum concentration achievable with an 80 mg dose (4 times more than most GBM patients are using) is only 500 nanomolar.  1 millimolar used in vitro is then at least 2000 times higher then what we could expect in the plasma.

These studies are very misleading, especially when read without a scientific background that can make sense of the concentrations being used (what is 1 mM?).  I do my best to bring awareness to this issue, that in vitro work can usually not be taken at face value (thinking the conclusions reached and the concentration-dependent mechanisms described necessarily have any clinical relevance).


  1. Ok, so you have to find achievable levels of a drug in blood first. I guess you look that up on pubmed? What are the usual "search words"? Bioavailability, pharmacokinetics,...of a X drug?

    Also, you mentioned free and glucuronidated genistein in your example. Only the "free" form matters (for other supplements, drugs as well) ?

    And as usually...thanks Stephen :)

    1. I've personally compiled a rather large spreadsheet on this sort of pharmacokinetic data for several hundred drugs and supplements. I should probably just publish this spreadsheet and save people repeating my work.

      I would usually (either on Google or pubmed) search for "clinical pharmacokinetics drug_x" or "Cmax drug_x" where Cmax in pharmacokinetic jargon stands for the maximum plasma concentration of the drug at a given dose.

      It is more complicated than this as in vitro drug studies express concentration in molar form (millimolar, micromolar, nanomolar) while pharmacokinetic studies usually express concentration in grams (ng/ml) and you have to know how to do the conversion, which is a different conversion for each drug as every drug has a different molecular weight.

      You also have to consider the plasma protein binding of each drug. My working assumption is that only free drug (unbound to plasma proteins) will be pharmacodynamically active. Also for many of the nutraceutical polyphenols like genistein or curcumin, glucuronidation and other kinds of conjugation result in the molecule mostly being found in the blood in conjugated rather than free form. Most studies don't consider the relative activity of say, glucuronidated curcumin versus free curcumin, so this is not much information on this question.

      Nearly all in vitro cancer cell studies testing non-oncology (repurposed) drugs or supplements are using higher-than-clinically-achievable drug concentrations. This is why I trust in vivo (animal) studies more than in vitro studies. At least with animal studies there is limit to the dosing you can give (this limit is undue toxicity or death of the animal from using too high a dose). Of course, in vivo rodent studies are problematic too as humans and rodents are separated by about 80 million years of evolution have a rather different metabolism, pharmacokinetics, blood-brain barrier activity, immune system etc. It's often the case that rodents get much higher plasma concentrations of the drug in these studies than a human would get with usual human dosing.

      In short, preclinical work gives some indication of whether a drug might be worth pursuing in the clinic, but only human studies can indicate how the drug might perform in a human (although even here interpatient variability in tumor genetics and differences in drug metabolism/pharmacogenetics complicates matters).

    2. Come to think of it I think there's a column in the "pharma and non-pharma" list in the library for maximum plasma concentration. It's one of the last columns in the sheet (column X). I still need to fill this column out a lot more than I have.

    3. Another big consideration is that the amount of drug reaching the extracellular fluid in the brain will often be much less than the unbound drug in the blood plasma. This is because of the blood-brain barrier. Even in an active tumor, the blood-brain barrier is only partially disrupted.