Posted by Larry Hoover on March 28, 2007, at 10:58:36
In reply to Re: neighbourhood, posted by Klavot on March 27, 2007, at 9:03:34
> http://jama.ama-assn.org/cgi/content/abstract/297/8/842
>
> Mortality in Randomized Trials of Antioxidant Supplements for Primary and Secondary Prevention
> Systematic Review and Meta-analysis> Conclusions Treatment with beta carotene, vitamin A, and vitamin E may increase mortality. The potential roles of vitamin C and selenium on mortality need further study.
I found the full-text version.
http://www.jhsph.edu/welchcenter/_pdf/3_6_07.pdf
Where to start? I cannot believe stuff like this gets published. This product is so biased, it does not even support its published conclusions within its own text. Its basic premises are fatally flawed, methodologically. It's junk science.
A priori assumptions for meta-analysis require that the study groups and study methodology are similar. That's the whole point, to create something of the sort of a larger homogenous group than was available through separate and smaller subject pools, subject to similar control of variables, to increase the power of the discriminant measure(s). Here, primary (healthy people) and secondary (those with diagnosed disease/pathology) prevention trials are lumped together without any rationale for doing so. Disease prevention and disease recurrence are different entities, certainly. Also, a prevention trial is disease-specific. They didn't look at disease outcomes, they looked at all-cause mortality. A car accident would count. Using other people's data for purposes not within the bounds of the intent of the original data collection constitutes the derivation of a new hypothesis. It is not a test of the hypothesis, however, as no control of variables is possible. It is purely speculative.
Another disturbing methodological flaw is the a priori exclusion of all studies in which no subject died. As the analysis was (supposedly) to find correlations between antioxidant supplementation and mortality, excluding 747 trials with no mortality, and analyzing only 68 trials with mortality seems illogical, at best. But apart from a single paragraph in the preamble, I didn't see it mentioned in the discussion. Maybe I missed it.
Another a priori discriminant (with subsequent effects on the statistics), a completely arbitrary separation of trials into two groups which they call low-risk and high-risk, is without any realistic relevance. Does it really matter if subjects took vitamin E or whatever in a double-blind trial environment (as one of the four criteria for risk was applied)? No, it doesn't. So long as some people took it and some people didn't, whether they died or not can't be attributed to placebo effect. All this arbitrary distinction does is provide you with yet another form of data exclusion, which permits derivation of new statistics. If you divide a heterogenous group without significant variation enough times, you will find statistical significance, eventually. This is called data mining, pejoratively, amongst scientists.
Consider yet another a priori decision, with subsequent profound effect on mortality stats....the lumping together of chronic vs. acute exposure. For example, one study in which bed-ridden geriatric patients were given a single injection of 200,000 IU vitamin A is lumped together with data from long-term oral supplementation of healthy people with 2,000 IU/day. Remember, meta-analysis requires similarity of subject pools and methodology, yet the obtained statistical relative risk for vitamin A, by dose, was 1.000006. That was found to be significant, but with the underlying assumptions (or ignorance, as you might alternatively conclude), does that have any relevance to others? Does it warrant inclusion in the abstract?
Now, within the obtained statistics, there is a hierarchy of groups. There is the statistic derived from all data. That would be the main outcome of the study. There are then statistics for the individual antioxidants, but with all data included. Then analyses within those groups for factors influencing the outcomes. And only at the end, was the high-risk/low-risk data exclusion criterion applied. Way down the relevance/validity tree. Here are the study outcomes, in order of validity (notwithstanding the methodological flaws):
"The pooled effect of all supplements vs placebo or no intervention in all randomized trials was not significant (RR, 1.02; 95% CI, 0.98-1.06)."
No harm done. Where is that in the abstract?
Next in line, the individual antioxidants, with dose alone amongst lone variables that led to significant statistics....
"Univariate meta-regression analyses revealed significant influences of dose of beta carotene (RR, 1.004; 95% CI, 1.001-1.007; P = .012), dose of vitamin A (RR, 1.000006; 95% CI, 1.000002-1.000009; P = .003), dose of selenium (RR, 0.998; 95% CI, 0.997-0.999; P = .002), and bias-risk (RR, 1.16; 95% CI, 1.05-1.29; P = .004) on mortality. None of the other covariates (dose of vitamin C; dose of vitamin E; single or combined antioxidant regimen; duration of supplementation; and primary or secondary prevention) were significantly associated with mortality."
So, beta carotene (a single form of provitamin A)/Vitamin A have statistical mortality risk, of 0.4% and .0006%, respectively, *by dose*. Selenium significantly reduced mortality. Do you see that in the abstract? No, it "needs further study". And vitamin E has yet to be implicated. Please also note the other variables which did not obtain significance when regressed alone (i.e. by univariate analysis).
Next, you have the multivariate analyses, which attempt to tease out the variables which significantly influence the collective statistic, from those that don't seem to make a difference one way or the other. We're getting into data mining territory, here. This is where you typically obtain rationale for conducting future hypothesis-testing. We're already beyond the true scope of meta-analysis available from this diverse data-set, but lets see what they find, anyway.
"In multivariate meta-regression analysis including all covariates, dose of selenium was associated with significantly lower mortality (RR 0.998; 95% CI, 0.997-0.999, P=0.005) and low-bias risk trials with significantly higher mortality, (RR, 1.16; 1.05-1.29; P=.005). None of the other covariates was significantly associated with mortality."
Let's start at the last sentence. The mortality risk of beta-carotene and vitamin A do not survive data-mining, no matter how they slice it. No evidence against vitamins C or E, by any measure. Selenium still looks golden, except when the arbitrary data exclusion of "bias risk" is applied. Look at the expansion of the 95% confidence interval following this data exclusion. Heterogeneity has gone way up (larger CI), so the subject pool was drastically reduced, or they were dissimilar in some way.
When bias risk itself was examined, we find significant effects.
"In trials with low-bias risk mortality was significantly increased in the supplemented group (RR, 1.05; 95% CI, 1.02-1.08)", whereas in high-bias trials, "...mortality was significantly decreased in the supplemented group (RR, 0.91; 95% CI, 0.83-1.00)."
What does this mean? One possible variable (of four; they never say which one(s) were applied) is 'absence of blindedness'. Is it inconceivable that people who take a supplement knowing it is good for their health might have a state of mind that is different from someone who is being given a supplement that may be a placebo? I still contend that this data exclusion criterion is arbitrary, and in any case, it has not yet had any influence on the obtained statistics, save that one adverse finding for selenium trials.
The researchers continue to massage the data, however, and obtain some further statistics. Now they consider only whether a supp was used alone or in combination, but not dose.
When we examine the stats, we find that beta-carotene used alone is bad: RR 1.06 (CI = 1.01-1.11), but when combined with other antioxidants, it's okay: RR 1.01 (CI = 0.94-1.08). Only when you further exclude data, i.e. combinations which include selenium, *and* they apply the arbitrary risk criterion, is there any significant risk found in combinatory analyses: RR 1.07 (CI = 1.02-1.11). We already knew you shouldn't take beta-carotene alone. Why were such studies even conducted? Anyway....
When we look at vitamin A, there is no significant risk of using it alone or in combination. That is the main finding. Only when data are excluded, do you see any significant risk. Again, it is when the risk bias is applied, *and* selenium combinations are excluded. Why don't they say that?
I've still yet to see anything about vitamin E being bad. Oh, wait. Vitamin E, alone, in combination, in arbitrarily low or high dose, had no effect on mortality. However, when data are excluded, specifically when the risk criterion *and* selenium combinations are excluded, we see a significant finding of increased mortality. Are you starting to see a pattern here?
Vitamin C? They couldn't find anything. Even when they excluded any data they could. So why does it "need further study"?
Selenium? When used alone and/or in combination, it significantly reduced mortality: RR 0.91 (CI 0.84-0.99), and survived the risk criterion. And they conclude further study is required?
Here, once again, is the published conclusion in the abstract:
"Treatment with beta carotene, vitamin A, and vitamin E may increase mortality. The potential roles of vitamin C and selenium on mortality need further study."
In fact, their own evidence shows that selenium reduces mortality, and is an essential component of any antioxidant supplementation regime. Beta-carotene should never be taken alone, but in a combination of antioxidants, it is safe to use. That would be a valid abstract conclusion.
The interdependence of antioxidants is worthy of an entirely separate discussion. For example, vitamin C is required to regenerate oxidized vitamin E. And I've totally ignored the fact that d-alpha-tocopherol (synthetic) is *not* a surrogate for natural vitamin E. d-alpha-tocopherol displaces d-gamma-tocopherol (one of eight structural variations of vitamin E), but fails to accept electrons in the chain reaction in which d-gamma protects membranes from oxidation. It is thus toxic, if absorbed in this synthetic unnatural form, without simultaneous increases in gamma-tocopherol. We haven't even considered beta- and delta-tocopherols, or the tocotrienols (eight more, which some include with tocopherols under the term vitamin E).
Nothing was said here about the absence of cofactors, which explain *why* beta-carotene is toxic if taken alone. It is reductionist science in the extreme, reductio ad adsurdum, to isolate a nutrient and manipulate that sole variable.
Nutrients act like instruments in an orchestra. You can't give cellos to the violinists and expect them to play their role, nor silence all but the percussionists, and expect to hear the melody. What are they thinking? Moreover, what do these experiments collectively demonstrate? That is the real question. If all experiments are methodologically suspect, then GIGO is the only conclusion. We got garbage out, here, without even considering if it was garbage going in.
Lar
poster:Larry Hoover
thread:744072
URL: http://www.dr-bob.org/babble/alter/20070320/msgs/744928.html