March 17, 2011
Follow up on pediatric seizures and genomics
I received the following question via email after my recent post:
Depakene is about the only thing that has controlled my child's seizures. But the 23-times greater risk is scary. Do the POLG gene variants cause higher drug blood levels or what? Should those be monitored more often? We'd hate to stop the drug and 'fix' something that's not broken. My daughter's been on Depakene for 4 years with no problems. If she has that SNP genotype, is her risk 23 times higher, or might it be some other amount?
Those are excellent, very practical points and questions you raise!
There is not much known yet about what, if any, statistical association there may be between the POLG polymorphisms mentioned in my previous blog post and the peak valproate drug levels or rates of drug metabolism and excretion (pharmacokinetics). At least, nothing has so far been published about this by the neurology researchers who have been studying the POLG variations in children and adults with epilepsy.
It is possible that there are some as-yet unknown variations in the genes for one or more of the liver enzymes that metabolize valproate--variations that are frequently concomitant with the gene variations in POLG.
In response to your question, the mechanism explaining why and how the liver injury happens is not yet known. It is not yet known whether valproate causes the same sort of POLG-related mitochondrial abnormalities as some NRTI antiretroviral medications do (see Duong article). So far, it is only known that the POLG gene variants empirically predict a 23-times increased probability of serious liver injury happening. The fact that the mechanisms or reasons for it are not yet well-understood does not make this empirical fact any less real.
You might be interested to read the book chapter published last year by Drs. Bruce Cohen, Patrick Chinnery, and Bill Copeland (links below) which recommends that blood tests for liver function be repeated every 3 months, especially if it is not practical to avoid prescribing the valproate.
Regarding individualized quantitative risk prediction, it's important to note that the 23-fold increased risk (odds ratio) and the 95% 'confidence interval' (95%CI) for the odds-ratio (8.5 to 66.7) reported in the article by Stewart and coworkers is a population average and the range within which the true odds ratio is expected to fall with 95% certainty based on the finite number of people they studied.
For predicting the range within which a single person's odds would fall with 95% certainty, statisticians use what is called a 'prediction interval' (95%PI; see book by Hahn & Meeker, link below, if you're interested). The 95%PI is wider than the population-average 95%CI 'confidence interval'. The equations for calculating the 95%PI from the data in the Stewart paper give a range from 5.3 to 97.1. So the risk might be expected with 95% certainty to range from "only" 5.3-times the baseline risk if you didn't have one of those POLG gene variants, to as much as 97.1-times.
Obviously, even the lower 95%PI figure of 5.3 is a big multiple, worthy of attention.
A lot of genomics is, oh, your risk is increased by 1.4 times compared to normal or other amount so modest that a person hardly feels like a choice really matters. By contrast, the POLG valproate 23-times risk multiple for serious injury is an actionable example, too big for any sensible person to dismiss. And the POLG variants reported by Stewart affect about 3.1% of people, so it's not like these are rare mutations that are rarely seen in routine everyday practice.
• rs3087374 (3708G>T; Q1236H) on Chr. 15 (both of these are in your 23andme raw file download)
• rs2307441 (3428A>G; E1143G)
In summary, the usefulness of genomics or other personalized medicine information or quantitative risk prediction is not that it becomes a hard-and-fast criterion of what is right or what is wrong to prescribe. There will always be unknown factors or additional considerations that preclude any 'protocol' being suitable or 'best' for everybody. There will always be a diversity of goals and values--what's important to you; how much risk you're prepared to accept in order to have chance of achieving the benefits that are important to you; and so forth.
Instead, the usefulness of genomics and other testing, like in this epilepsy medications example, is that it can inform your decisions about treatment options and about ongoing monitoring activities once those treatment choices have been made--in exactly the way that your email question indicates. It's not mandatory to fix what's not 'broken'. Just be mindful of the extra risk and be extra vigilant by checking the liver function blood tests frequently, so that if any problem does arise it's caught early and corrective action and different meds choices can be taken at that time. It empowers you to prevent bad things from happening and to get the most value from the care decisions at a risk-level that's acceptable to you.
Thanks for your email!
Douglas McNair, MD PhD, Senior VP, is one of three Cerner Engineering Fellows and is responsible for innovations in decision support and very-large-scale datamining. McNair joined Cerner in 1986, first as VP of Cerner’s Knowledge Systems engineering department; then as VP of Regulatory Affairs; then as General Manager for Cerner’s Detroit and Kansas City branches. Subsequently, he was Chief Research Officer, responsible for Cerner’s clinical research operations. In 1987, McNair was co-inventor and co-developer of Discern Expert®, a decision-support engine that today is used in more than 2,000 health care facilities around the world. Between 1977 and 1986, McNair was a faculty member of Baylor College of Medicine in the Departments of Medicine and Pathology. He is a diplomate of the American Board of Pathology and the American Board of Internal Medicine.
Anderson G. Children versus adults: pharmacokinetic and adverse-effect differences. Epilepsia. 2002;43(Suppl 3):53-9.
Bondreva I, et al. Predictability of individualized dosage regimens of carbamazepine and valproate mono- and combination therapy. J Clin Pharm Ther. 2010 Nov 10 [ePub ahead of print]
Botha J, et al. A model for estimating individualized valproate clearance values in children. J Clin Pharmacol. 1995;35:1020-4.
Cohen B, et al. 'POLG-related disorders', in Gene Reviews, ed. R. Pagon, et al. Univ Washington, 2010. (free download)
Bruce Cohen page at UMDF
Bill Copeland page at NIEHS, RTP, North Carolina
Duong J, et al. Alteration of cytochrome oxidase subunit I labeling is associated with severe mitochondriopathy in NRTI-related hepatotoxicity in HIV patients. Mod Pathol. 2006;19:1277-88.
Hahn G, Meeker W. Statistical Intervals. Wiley, 1991.
Lee M, et al. The relationship between glucuronide conjugate levels and hepatotoxicity after oral administration of valproic acid. Arch Pharm Res. 2009;32:1029-35.
Levy R. Idiosyncratic Reactions to Valproate: Clinical Risk Patterns and Mechanisms of Toxicity. Raven, 1992.
McNair D. DTC Genomics tests, Using your results: Pediatric seizure medication safety/prevention example. Cerner blog, 10-FEB-2011.
Stewart J, et al. Polymerase γ gene POLG determines the risk of sodium valproate-induced liver toxicity. Hepatology. 2010 Nov;52:1791-6.