Vitamin C for cancer: Trying to rise from the grave once again

There are certain things that can happen that are the equivalent of the Bat Signal to me; that is, if you can swallow the idea of me being in any way like Bruce Wayne. Call them the Cancer Signal, if you will. When I see the Cancer Signal, I know that I have to head down to the Cancer Cave—wait a minute, that doesn’t sound right. Scratch that. In any caes, there are certain studies or stories that basically demand my attention and say, “Blog me, Orac! Blog me right now!!!” Either that, or they’re studies that capture my readers’ attention, leading them to e-mail me or Tweet at me plaintive requests to blog them. Sometimes I do it. Sometimes I don’t. Not every study that interests some of my readers interests me sufficiently for me to put forth the effort to blog about it, and, remember, this blog is all about me. I love that there are a few thousand out there who love what I lay down, but at the end of the day this is all about what interests me. I know that sounds arrogant, but if I started writing about things that didn’t interest me that much this blog would rapidly fade into oblivion. My writing would become less interesting, and I’d lose interest, maybe even to the point of giving it up.

Be that as it may, fortunately, the study of the week last week was one that actually did interest me. In any case, this week’s Bat Signal consisted of a series of news reports with titles like:


These stories, to varying degrees, miss the point, from utterly missing it to missing most of it. Unfortunately, I confess that I wasn’t able to help at least one of them. A reporter happened to leave me a message Tuesday morning, which is my operating room day, and I didn’t have time to read the paper and to get back to her before her deadline. Unfortunately, my real job sometimes gets in the way of my being able to help out a journalist. That paper, by the way, appeared in Science Translational Medicine from Jeanne Drisko and Qi Chen from, yes, Kansas University Medical Center, home to one of the more—shall we say?—dedicated centers of quackademic medicine. This indicates to me that STM’s standards are slipping. But then, STM did publish a rather credulous paper by our old friend Ted Kaptchuk on placebos less than a month ago; so maybe I expect too much. Clearly STM appears to be looking for more papers on “complementary and alternative medicine” (CAM) or “integrative medicine.” Worse, just last week, one of the associate editors of STM, Yevgeniya Nusinovich, hosted a lovefest of a web chat in which Dr. Josephine Briggs, director of the National Center for Complementary and Alternative Medicine (NCCAM), and Dr. Jeanne Drisko, chock full of pro-CAM tropes, distortions, and cherry picking. Among the cherry picked stories, besides the Trial to Assess Chelation Therapy (which Dr. Drisko was a co-author on), was this study on vitamin C and cancer, mentioned near the end of the web chat and the study that I promised to say more about.

A typical story describes the recently published research thusly:

People with ovarian cancer who receive high-dose vitamin C injections are less likely to report toxic side effects from chemotherapy than people who had chemotherapy alone, according to the results of a small clinical trial.

The study, published today in Science Translational Medicine1, was too small to assess whether the combination of chemotherapy and vitamin C combats cancer better than chemotherapy alone. But accompanying work in mice suggests that the two treatments could be complementary.

The results are the latest salvo in long-running controversy over the use of vitamin C against cancer. Early studies championed by Nobel-prizewinning chemist Linus Pauling in the 1970s suggested that vitamin C could help to fight tumours2. But larger clinical trials failed to substantiate those claims3, 4.

With the spin, from another typical story, being:

One potential hurdle is that pharmaceutical companies are unlikely to fund trials of intravenous vitamin C because there is no ability to patent natural products.

“Because vitamin C has no patent potential, its development will not be supported by pharmaceutical companies,” said lead researcher Qi Chen.

“We believe that the time has arrived for research agencies to vigorously support thoughtful and meticulous clinical trials with intravenous vitamin C.”

Yes, indeed. The same old tropes are there, from the claim that vitamin C has usefulness in treating cancer to the old ascorbate warriors’ lament that there’s no patent potential in vitamin C, which means that pharmaceutical companies don’t want to invest money into doing science and clinical trials on it because there’s no profit potential. Of course, I’ve written fairly extensively about vitamin C and cancer before, using it as an example of how even a two-time Nobel Prize winner like Linus Pauling could fall prey to bad science when he wandered outside of his area of expertise. Every so often these stories come up suggesting that Linus Pauling has somehow been vindicated and how vitamin C is the greatest thing for cancer patients since surgeons first discovered that some cancers could be cured by cutting them out. Inevitably, I have to throw cold water on such claims. No, Linus Pauling has not been vindicated, and, no, vitamin C for cancer is not all that great. It might even be harmful.

Also, no, contrary to what critics say, I’m not close-minded about vitamin C and cancer. Unlike so many “alternative” cancer treatments, it’s actually a chemical and, at the doses used by alternative cancer practitioners, a drug. There’s even a (very) weakly plausible mechanism by which it might work. However, in vitro, the concentrations required to provide even a whiff of a hint of antitumor activity are ridiculously high, and the same is true in animal models. Let’s just put it this way. Imagine a pharmaceutical company had developed a compound with properties identical to that of vitamin C and could thus own the complete patent on it as a drug. Given the ridiculously high concentrations and doses required in preclinical models to demonstrate a hint of antitumor activity against , that pharmaceutical company would probably retire that compound before even the animal model stage because, as I like to put it, getting any useful anticancer activity out of it would be such a long run for a short slide. A good drug for cancer is, at the very minimum, active at low or reasonable concentrations against the cancer cells being targeted, and vitamin C fails miserably on that count. Worse, there are at least indications that in some cases vitamin C might interfere with chemotherapy.

So does this study change my opinion? Not really. At best, it suggests there might be some utility for ascorbate (vitamin C) against ovarian cancer, but that ascorbate therapy for cancer still remains at best a long run for a very short slide right into the gloved ball of reality for a third out. (OK, I’ll stop with the baseball analogies.)

The dubious reasoning begins right in the first paragraph, with the authors’ justification for “re-examining” ascorbate as a cancer therapy. Basically, they point out that the pharmacokinetics of oral ascorbate dosing is different from intravenous dosing, to the point where it is possible to obtain serum ascorbate concentrations of 10 mM (millimolar). To give those of you who aren’t chemists a rough comparison of just how high that concentration is, most cancer drugs have active concentrations in the nanomolar (nM) to micromolar (uM) range, in other words, a thousand-fold to a million-fold lower than 10 mM. For example, the IC50 (the concentration that leaves only 50% of cells alive) for paclitaxel is in the 2.5 to 7.5 nM nM range, depending upon the cell line, and 50 nM is considered a good, effective therapeutic concentration. You get the idea. You need a lot of ascorbate:

By contrast, when ascorbate is injected intravenously, tight control is bypassed and pharmacologic concentrations of ascorbate are established until excess ascorbate is excreted by kidney. Plasma concentrations greater than 10 mM are safely sustained in humans for ~4 hours (10–13). When patients have normal renal function and glucose-6-phosphate dehydrogenase (G6PD) activity, toxicity is minimal even with intravenous doses as high as 1.5 g/kg, equivalent to 105 g for a 70-kg person (2, 12). These data indicate that intravenous administration of pharmacologic ascorbate doses is safe and similar to drug administration. Therefore, the effect of ascorbate in cancer treatment is worth reexamining.

These are huge doses, consistent with previous experiments in mice with a xenograft from an ovarian cancer cell line (Ovcar5) in which 4 g/kg of ascorbate was administered twice daily for a total of 8 g/kg/day. The result was an inhibition of xenograft growth of around one-third after 30 days. Results for a pancreatic cancer cell line and a glioblastoma cell line were only marginally better.

The authors did several cell culture studies in which ovarian cancer cell lines were treated with ascorbate and various chemotherapeutic agents. The authors reported an IC50 of between 0.3 and 3.0 mM, which is still incredibly high for an anticancer drug. The authors blithely write that this is “easily achievable” with IV ascorbate. Maybe so, but given the quantities involved, if you’re going to use a drug that requires such high plasma concentrations to have activity, that activity had better be awesome. None of the activity shown in this paper can be characterized as being particularly impressive. Worse, the authors, despite testing several ovarian cancer cell lines, only tested one non-tumorigenic immortalized ovarian line, HIO-80, and, finding that the IC50 to kill HIO-80 cells was much higher than all but one of the other cell lines (SHIN3), proclaimed a high degree of specificity for cancer. Moreover, HIO-80 cells are hardly “normal.” They likely contain BRCA1 mutations. Finally, the authors only used one assay for proliferation, the MTT assay. This particular assay is very popular (I use it in my lab not infrequently) because it is faster and easier than counting viable cells and also allows for large experiments using 96-well plates. However, the MTT assay depends on the metabolism of cells to produce a dye that is detected. The amount of light absorbance due to the dye is assumed to be proportional to the number of viable cells. Usually, this assumption is reasonable accurate, but lots of things can interfere with this and render that assumption incorrect. For instance, one wonders if very high concentrations of ascorbate can interfere. I’d want to see a control demonstrating that the MTT results correspond to cell number.

In other words, if I were a reviewer for this paper, not so fast, I’d have said. I want to see the results for at least a couple of more non-tumorigenic cell lines and a control validating the MTT in the presence of so much ascorbate (even if just a reference) before I’ll let you conclude that the effects of ascorbate are highly specific for cancer over normal ovarian cells. At the very least, I wouldn’t have considered it unreasonable to ask for a couple more non-tumorigenic ovarian epithelial cell lines to be tested.

In any case, the authors also did some mechanistic studies, the results of which were consistent with the activity of ascorbate in cancer requiring the production of peroxide (H2O2), as H2O2 scavengers blocked the effect. They also did a series of experiments that indicated synergy between ascorbate and carboplatin, a common chemotherapy drug used in ovarian cancer. One area that, as a reviewer, I’d have gotten on the authors’ case was the series of xenograft experiments using ovarian cancer cell lines implanted under the skin of immunodeficient mice, specifically this part:

Two-tailed Student’s t test was performed for comparison of treated groups to control group in the cell and animal experiments, as well as for toxicity comparison between chemotherapy group and chemotherapy + ascorbate group.

No, no, no, no, no! This is some pretty basic stuff here. There are eight different experimental groups, and the authors didn’t control for multiple comparisons, as far as I am able to tell. Pair-wise two-tailed t-tests are not the correct statistical test for determining statistical significance in such a case; likely some form of ANOVA would be, given that the dataset consists of tumor weights and volumes of ascites, the latter being a common estimate of ovarian tumor burden in mouse models. Some form of ANOVA, likely factorial ANOVA, would have been the proper test, given that there are combinations of three drugs being used. Whatever the correct test is (and I’ll leave that to the statisticians out there), I know that Student’s t-test isn’t it, and that using Student’s t-test will often produce “false positives” that appear statistically significant but aren’t.

All of this, however, is the warmup to the part of the study that got it noticed, namely the clinical trial. Without the clinical trial, this would have been yet another in vitro and animal study of high dose vitamin C that provokes a collective yawn throughout the scientific community. The clinical trial itself, was a randomized prospective phase I/IIa clinical trial, which means that the trial was designed to combine an evaluation of toxicity with a pilot study to evaluate efficacy and safety. Its primary objective was to “determine the safety of high-dose intravenous ascorbate when combined with first-line chemotherapy paclitaxel and carboplatin in the treatment of advanced-stage ovarian cancer,” along with evaluation for toxicity. Consequently, the two groups were (1) standard carboplatin plus paclitaxel (Cp + Pax) and (2) carboplatin plus paclitaxel plus ascorbate (Cp + Pax + AA) according to this design:

Ascorbate dose for the Cp + Pax + AA arm was established via dose escalation initiated at 15 g per infusion titrated up to a therapeutic range of 75 or 100 g per infusion, with a target peak plasma concentration of 350 to 400 mg/dl (20 to 23 mM) (12, 13). The ascorbate infusion was given at a rate of 0.5 g/min, and 400 mg of magnesium chloride (Wellness Pharma) was supplemented into each infusion. Once the therapeutic dose was established, the Cp + Pax + AA group received ascorbate two times per week in conjunction with chemotherapy for 6 months, and injectable ascorbate was continued for another 6 months after chemotherapy completion.

In addition, the authors noted:

Two subjects voluntarily withdrew from the Cp + Pax arm before treatment commenced because they wanted intravenous vitamin C, and they were excluded from data analysis. Two subjects were removed from the Cp + Pax + AA arm because they were noncompliant with tobacco use, and one was removed from the Cp + Pax + AA arm after in vitro cytotoxic assays detected that her tumor cells were resistant to all chemotherapy. These three subjects received doses of chemotherapy and ascorbate, so their adverse events were counted, but they were excluded from the survival report (table S3). Double blinding was used at enrollment and randomization, but was not maintained during the treatment because no placebo control was used.

So what we have here is a small clinical trial with a 19% dropout rate that wasn’t even blinded. It reported zero difference in overall survival (both were, as one would expect for ovarian cancer at this stage, abysmal), and zero statistically significant difference in time to relapse/progression. In all fairness, there would have had to have been an enormous effect to produce a statistically significant effect on survival or progression in such a small study, but these are the two “hard” endpoints that would be least affected by the lack of blinding, although one notes that time to progression could be affected by lack of blinding when the definition depends on interpreting scans. It’s also hard not to note that the differences in toxicities are all in the mildest reported toxicities, grades 1 and 2 (out of a scale of 1 to 5, with a score of 1, which denotes mild toxicity that requires no intervention to 5, which is death). There were no statistically significant (or even close to statistically significant) differences in toxicities graded 3 or 4, which are the most troubling kind. Take a look at the graph:

fig4a

Then, when the authors broke it down, this is what they found:

fig4b

Notice the types of complaints with the biggest difference: gastrointestinal (which usually includes symptoms such as nausea, abdominal pain; dermatology, which usually includes itching and rashes of various types); pulmonary, which often includes symptoms of shortness of breath, cough, and the like, and renal/genitourinary, which is the only one that’s less objective. So, basically, what we have is a study that found no benefit in overall survival or time to progression (not unexpected for such a small study). More importantly, contrary to the way it was trumpeted to the press, the decrease in adverse events actually observed was limited to the least serious adverse events (grade 1 = minor, causing no limitation of activity, no intervention required; grade 2 = moderate, some limitation of activities, minimal intervention indicated) with the most potential to be subject to reporting bias, which in the context of a trial that is not blinded makes the difference reported probably meaningless. In other words, this was probably a negative study, a long run for a short slide, indeed. (Sorry, couldn’t resist. Again.)

Not that any of this stops Dr. Drisko from saying things like:

“We now have a better understanding of vitamin C’s anti-cancer action, plus a clear safety profile, and biological and clinical plausibility with a firm foundation to proceed,” said researcher Dr. Jeanne Drisko, director of the integrative medicine program at University of Kansas Medical Center. “Taken together, our data provide strong evidence to justify larger and robust clinical trials to definitively examine the benefit of adding vitamin C to conventional chemotherapy.”

“Firm foundation to proceed”? You keep using that term. I do not think it means what you think it means. All Dr. Drisko has shown is that there is no difference in survival between the group receiving high dose ascorbate and the group receiving standard of care and that, putting the very best possible spin on the data, maybe adding ascorbate to carboplatin and paclitaxel in the treatment ovarian cancer might decrease the most minor side effects of chemotherapy; that is, if there wasn’t reporting bias due to the lack of blinding after the randomization.

I’m not impressed. Neither should you be.