Contaminated TB Treatments

(None of this is medical advice.)

In 2020 the FDA identified contamination in two drugs used to treat Tuberculosis, Rifampin and Rifapentine. The contaminants were 1-methyl-4-nitrosopiperazine (MNP) in rifampin or 1-cyclopentyl-4-nitrosopiperazine (CPNP) in rifapentine, and levels were ~10x the FDA’s acceptable threshold (1-4ppm for MNP, 8-14ppm for CPNP). At the same time, these are two of the standard treatments for Tuberculosis and the alternatives come with downsides of their own, so the FDA permitted these drugs to continue being prescribed and sold in the US.

My question was: What risk does this contamination pose?

My tentative answer is: tiny. Specifically, my best estimate is that taking these treatments increases lifetime risk of cancer by roughly one in a million. The basic reason is just that the drugs are really minimally contaminated and the course of treatment is limited in duration.

In what follows I’ll focus on the treatment regimen for latent TB. I haven’t checked the numbers for active TB, though I would be surprised if they were so different as to change the conclusions.

Evidence from Rats

As far as I can tell we don’t actually have experimental/observational data on the cancer risks of MNP and CPNP, just of related compounds in the same family (nitrosamines).

The European Medical Authority (EMA) has a report on nitrosamines and acceptable concentrations, and they reference studies in rats finding that a number of nitrosamines produce tumors with something like a year of exposure at the level of 5-10mg/kg/day (TD50, meaning 50% lifetime risk), and they base their upper bounds on those numbers. I spot-checked one of the numbers to make sure and was able to confirm it in the underlying database.

There are some caveats in interpreting these numbers for humans:

1. Rats get cancer at way higher rates than humans.

2. 1. There’s actually a trend here: shorter-lived creatures have fewer protections against cancer because something else always kills them first, while longer-lived creatures have many more protections.
   2. The famous example is cisplatin, widely used as a cancer treatment in humans and derived from trees, which live longer than us and so have more need of cancer prevention.
   3. On the whole this suggests that the TD50 for humans is higher than for rats.

3. Rats have less time for replication errors to accumulate.

4. 1. This is speculation on my part, but one could imagine that toxins like nitrosamines produce some DNA damage and that that’s the mechanism for causing tumors in rats.
   2. If that’s the case, the model one might have is that this is more dangerous for humans, because we live longer and so there are more chances for subsequent DNA damage to produce cancer.

5. The doses rats are exposed to in these studies are massive. Like more than 100,000 times higher than the biggest dose one would get from TB treatments.

6. 1. We know that lots of toxins/stressors/drugs have a response threshold.

   2. The most famous example I can think of is radiation exposure: below a certain threshold DNA repair mechanisms are able to keep up and there’s a negligible effect on adverse outcomes (at least in model organisms).

   3. 1. There’s even some evidence of improved outcomes at very low radiation doses relative to zero radiation (speculated to be due to engaging repair mechanisms more actively than they would be by default, thereby averting more problems than the radiation causes).

   4. That said, this study found in rats that the threshold dose at which cancer risk starts increasing (linearly with dose) is quite low, around 0.1ppm. So my best guess is that there isn’t a significant threshold effect at play and that the risk is indeed linear in dosage.

7. Humans are bigger than rats (~300x by weight), so the same dosage per body weight produces proportionately more chances of cancer.

8. 1. This isn’t considered or mentioned in any of the analyses of risk I’m seeing, so I might be missing something, but this seems like an important factor.
   2. At the same time, factor (1) probably matters a lot here in pushing the number back down, so I don’t want to naively apply this correction factor and get a wild overestimate.

Disregarding these caveats (i.e. naively assuming human response per bodyweight dose scales like rat response) and using dosing from the CDC, I’d estimate the cancer risks as:

1. Ioniazid + Rifapentine (30mg/kg/week for 12 weeks, CPNP at ~10ppm): Total exposure is 0.004mg/kg, versus 2500mg/kg of exposure for TD50 for rats (5mg/kg/day for 60 weeks). Suggests risk of ~1/million of cancer.
2. Rifampin (10mg/kg daily for 120 days, 3ppm MNP): Total exposure is 0.004mg/kg, same risk as above.

As a sanity check, the EMA’s threshold limit on nitrosamine exposure is set at 18ng/day based on a lifetime risk tolerance of 1/5 million. Exposure on the first treatment would be 2000ng/day (100x higher) but for much less than a lifetime (~400x less), suggesting the EMA might call the increased risk of cancer 60/billion.

As another sanity check, the FDA’s threshold limit is 96ng/day, set by a cancer risk of 1/100,000 developed over 70 years. Scaling that to the first treatment option gives 700/billion.

Both of these are lower than the numbers I estimated above, suggesting my estimates erred on the cautious side.

Evidence from Humans

This study on contaminated Valsartan seems promising. Valsartan was contaminated with nitrosamines at similar concentrations to Rifampin and Rifapentine.

The study looked at two cohorts in France: 1 million people who took contaminated Valsartan and 700k people who took uncontaminated Valsartan. There was no overall increase in cancer risk detected.

There were two cancer types that showed increased incidence (liver & melanoma) but they did enough different cuts of the data that I worry those are just false positives (the effects come and go depending on what they control for and disappear in the cohorts they have the most followup data for, and shows the wrong dependence on dose, etc.).

Even if the liver and melanoma increases are real, they had increased incidence of ~100/million/exposed yr (not clear if the increased incidence runs past end of exposure, I couldn’t tell that from the study). The typical person in the contaminated cohort was on a daily dose of ~20ug/day of nitrosamines for several years, for a cumulative exposure of (~0.2mg/kg). That’s a dose 100x what latent TB treatments involve, so adjusting down by 100x gives 1/million as the extra cancer risk, roughly in line with the rat numbers (which the study notes, through a slightly different calculation).