There’s a chemical contamination story in the generic drug industry that just isn’t going away. Late last summer, some lots of valsartan were recalled due to detection of N-nitrosodimethylamine (NDMA), and the problem has just continued since then. We’ll get into the chemistry of this problem in a minute, but first off, looking at this situation in detail tells you a lot about the generic drug business. A key point is that there are often just a few sources of actual API (active pharmaceutical ingredient). Other companies buy this material and formulate it into capsules, tablets, etc., and then still more companies repackage these. The patented drug world, by contrast, is a lot easier to follow, since while a drug is under patent there’s only one company that’s authorized to market it. They may well have several suppliers for different starting materials or for different steps of their synthesis, but it’s at least one defined route managed by one company.
In this case, the valsartan recall was originally traced back to a problem with the material from Zhejiang Huahai Pharmaceuticals. That’s the Chinese manufacturer who made the API itself. There were originally two generic companies whose tablets included the ZHP material, Teva and Prinston. But the Teva material was sold by Major Pharmaceuticals and also by Actavis, while the Prinston was sold by Solco Healthcare – there are those three layers of API production, pill manufacturing, and repackaging. It turned out that there were nine more repackagers of the Teva and Prinston tablets, so valsartan packages under those other names were added to the recall list. But not all of them. To add the confusion, some of the packages from those companies were still OK, because they’d bought the tablets from more than one manufacturer. The recall had to be done by batch number. Meanwhile, Torrent Pharmaceuticals also emerged as a company who had bought the original valsartan API from Zhejiang Huahai, so their products went on the list as well.
But then the FDA found, while testing material from all over, that valsartan manufactured by Hetero Labs of India (and sold as Camber Pharmaceuticals tablets) also had NMDA contamination, so the problem wasn’t just that one manufacturing source in China. This takes us up to late October, and that’s then things really started getting messy. Another drug in the same angiotensin II antagonist class as valsartan (irbesartan) was found to be contaminated, but this time with the N-nitrosodiethylamine (NDEA) instead of the dimethyl compound. This was made by ScieGen, and again was repackaged under still more names. The API manufacturer for the ScieGen material was Aurobindo (of India), and they recalled material. Yet another compound in this class, losartan, turned out to have contaminated material also produced by ZHP, but since ZHP themselves had by that time already been placed on the FDA’s import restriction list, no more of that API was coming in.
Then Mylan pulled several lots of valsartan tablets, using API that they’d made themselves. Teva followed by recalling more lots of their product, since they’d also bought some of the Mylan API for manufacturing. Over the last ten days or so, Torrent and Aurobindo have expanded their own recall of material because of the second NDEA contaminant, and that’s where we are now. So what’s going on?
Well, any synthetic organic chemist will look at the situation and decide that (1) there’s a process chemistry issue, (2) it seems to apply to common chemical features of the “sartan” angiotensin II antagonists, and (3) it could well be related to some change in the synthetic conditions, because this all didn’t seem to be a problem before. And these are indeed the case, but the details don’t make anyone look particularly good. The common feature, in this case, is the synthesis of the tetrazole ring common to all the sartans. And the change apparently was a solvent switch to dimethylformamide (or presumably diethylformamide, in the cases where NDEA is the contaminant?) The dialkylformamides are often contaminated by small amounts of the corresponding dialkylamines, and they are well known to break down to give those (slowly) under heating.
But how do you get N-nitroso compounds from the amines, and why was the solvent switched? Well, the classic industrial syntheses of these molecules involved reacting an aryl nitrile with tri-n-butyltin azide (often formed in situ from the trialkyltin chloride). ZHP themselves appear to have introduced a cheaper, higher-yielding route using just sodium azide and zinc chloride in an aprotic solvent like DMF (here’s what seems to be the patent on that route). The excess azide is consumed at the end of the process using sodium nitrite – but nitrite under acidic conditions will give you some nitrous acid, and nitrous acid will react with secondary amines to give you N-nitrosoamines. That would seem to be the root of the problem.
Well, one of the roots. The second problem is that no one apparently picked up on the N-nitroso contaminants for years. ZHP probably brought this new process online around 2012, and it would appear that other manufacturers came up with their own variations on the route after that. The FDA and EMA have by now both detailed several analytical routes to assaying for the impurities, and the history of the recalls over the last few months is basically the history of more time spent testing more lots of material for what has probably been in there for some time.
That isn’t good, because N-nitrosoamines are definitely genotoxic, at least at high levels. We’re stepping off into a big toxicology argument at this point, because the genotoxicity of these things at lower levels is a matter for (heated) debate. It all depends on how the high-dose animal studies can be extrapolated down, how both high- and low-dose animal studies can be extrapolated to humans in general, and how to interpret human observational data (on, for example, the consumption of cured meats, which contain low levels of nitrosamines) in the presence of multiple other factors. Not least among these is the problem that some of the N-nitroso compounds are in foodstuffs themselves, while others are produced by gut bacteria. Everyone can agree, though, that large amounts of N-nitroso compounds are bad news, for some value of “large”. And everyone can agree that exposing yourself to such compounds for no reason at all is senseless.
The FDA’s standard is to be below an amount that would be expected (by their dose/response modeling) to cause 1 extra cancer case in 100,000 people who were taking the tablets at a standard dose for 70 straight years. That’s pretty stringent, considering the background rates of cancer in people who actually stay alive for 70 years in a row, especially when you factor in that no one goes on valsartan when they’re ten years old. Unfortunately, the ZHP material (according to the FDA) would be expected to cause one extra case of cancer with only 8,000 patients taking the highest dose of the drug for only four years, and that’s definitely unacceptable. It appears that the other manufacturers’ batches had contaminants at lower levels, from what I can see, although the amounts do not appear to have been specified.
So we’re going to have to think about the way that synthetic routes in the generic API business are monitored, it would seem. People seem to have missed that changing the chemistry for the sartans could lead to this problem, so what else are we missing? It’ll be tricky – generic drug synthesis is generally a low-margin business where people compete at least partially on price, so the companies are always looking for new routes that will give them an advantage. And as the present example illustrates, there can be many different manufacturing sources scattered over several different countries. But no one wants to see this sort of thing happening again, either. . .