Antioxidants, including vitamin A and its analogues (retinoids) are hypothesized to prevent cancer, but results of randomized trials have not been very encouraging. A synthetic compound, polyprenoic acid, binds to the cellular retinoic acid-binding protein and has been shown to inhibit hepatocarcinoma in rats and mice. In this trial from Japan, polyprenoic acid was administered to patients with surgically or "chemically" resected hepatoma, who are at particularly high risk for recurrent or second primary hepatoma.
Methods
111 patients screened, 89 randomized. Of these, 78% were treated
surgically, 22% by injection with ethanol. 75% had hepatoma due to hepatitis
C, 13% hepatitis B, 11% other causes, including alcohol abuse.
The authors speculate that, since retinoids are considered chemopreventive rather than chemotherapeutic, their effectiveness against new, second hepatomas rather than against disease recurrence is logical. They estimate that a real survival benefit will become apparent in about 48 months, if current trends continue. They note that the mechanism of action of polyprenoic acid is not known, but that hepatocarcinogenesis is linked to impaired retinoid metabolism.
Editorial
David Shafritz, from the Albert Einstein College of Medicine, points out that distinguishing new primary tumors from tumor recurrence based on clinical criteria alone, as was done in this study, is not an ideal method. Some recurrences were probably counted as new primary hepatomas. He notes that this distinction can be made reliably, using viral DNA analysis, for example, when the hepatoma is caused by chronic viral hepatitis.
The authors first present the results comparing the recurrence rates in both groups, one against the other, and then present them using Kaplan-Meier survival curves.
The problem with directly comparing the data from the placebo and treatment groups is that patients with very different follow-up times are included and compared. This is particularly problematic when data are examined shortly after termination of a relatively long study, since there will be a wide disparity in lengths of follow-up under these circumstances: some patients will have just completed the trial whereas others will be several years out. This can be problematic if the effect of a therapy is just to delay the onset of disease by a short time, rather than to prevent it. In such a case, including patients with only short follow-up will inflate the calculated therapeutic effect.
Using the Kaplan-Meier survival curve method eliminates this problem by explicitly looking at development of disease over time. Patients with short follow-up are appropriately "placed" on the curve, and not compared with patients with longer follow-up. In the current study, I'm not sure the direct comparison should have been presented at all, but I would be interested to hear what others with more epidemiologic/biostatistics knowledge have to say.
The effect of polyprenoic acid on development of a second primary hepatoma was more pronounced when the authors compared the two groups directly, than when the (appropriate) Kaplan-Meyer analysis was presented. This makes me wonder whether some of the effect on hepatoma development wasn't due to a delay in appearance rather than actual prevention. Presumably, longer term follow-up will clarify this matter.
It was an intriguing study, primarily as it dealt with an unusual aspect
of altering the natural history of cancer. I would have been interested
in knowing if this compound has activity in other cancers. I was also troubled
by the way recurrent cancers were distinguished from second primaries.
Letters
to the editor about this article from the New England Journal of Medicine
website.