“If the United States is ever going to control our health
care costs, we have to demand better evidence of effectiveness, and stop
handing out taxpayer dollars with no questions asked.”
The above quote is from an op-ed article published yesterday in
the New York Times by two influential health policy makers Drs’ Ezekiel J. Emanuel
and Steven D. Pearson. They were referring to two expensive proton beam
machine being built at the Mayo clinic in Minnesota and Arizona. However, a
similar question could be posed for some of the therapies used in dialysis
patients. Shouldn’t Medicare demand evidence on effectiveness before paying for
epo or sevelemar or paracalcitol?
Take ESA treatment of anemia in patients with kidney
disease. The FDA has indicated in the label for these drugs that therapy should
be interrupted or the dose reduced if the hemoglobin level rises above 11 g/dL. The FDA
has not provided a lower Hb limit. Still, Medicare continues to reimburse providers
for the cost of ESAs for a target Hb concentration of 10 to 12 g/dL. There is
no randomized controlled trial (RCT) evidence that supports the idea of targeting a Hb above 11 g/dL in dialysis patients. A Hb>11 g/dL is not associated with better
outcomes than aiming for a Hb of <11 g/dL. However, Medicare expenditure for ESAs has increased almost exponentially (Fig.1). Indeed, the latest numbers show that over 40% of dialysis patients have Hb concentrations above 12 g/dL.
![]() |
| Fig. 1 Costs (Per Person Per Year) of dialysis services between 1991 and 2004 Source: USRDS |
In the Normal Hematocrit study, published in NEJM, dialysis patients at high
risk for cardiovascular disease were randomized to a higher Hb (13-15 g/dL)
versus a lower Hb concentration (9-11 g/dL). Patients in the lower Hb arm had
fewer vascular access thrombosis events, lower mortaility and fewer MI's
compared to those randomized to the higher Hb arm. A Hb target range of 9-11
g/dL is supported by evidence from a randomized trial. Forget about disincentives, should Medicare stop reimbursing providers if the Hb concentration is >12 g/dL range?
![]() |
| Fig. 2. Total Medicare spending for injectable drugs between 1993 and 2003 Source: USRDS |
The situation is even more uncertain with metabolic bone disease.
While observational studies suggest an association between hyperphosphatemia
(and the calcium x phosphorus product) and mortality in dialysis patients, RCTs have so far failed to demonstrate that control of
hyperphosphatemia is associated with improved survival. Worse still, these trials have been relatively modest in size and none have been placebo controlled. As Block pointed out recently in Current Opinions, a placebo controlled trial is long overdue.
Nevertheless, despite the lack of
randomized control data, Medicare continues to reimburse for expensive non-calcium
containing phosphorus binders like sevelamar and lanthanum. A similar situation exists for vitamin D analogues compared to generic calcitriol. Medicare costs are rising (Fig.2) and yet there is a paucity of RCT evidence.
In the NY Times article, Emanuel and Pearson make an interesting recommendation: that
Medicare adopt “dynamic pricing”. They argue that Medicare should only pay more for therapies that are proven. For treatments that are unproven, they recommend that Medicare should pay only what it pays for the
cheaper alternatives. In other words, Medicare should only pay for sevelemar or lanthanum at the rate it pays for generic calcium carbonate phosphate binders. Or for paracalcitol at the rate paid for generic calcitriol.
If studies were done showing that these therapies are better than other
treatments, the payment should go up. If no studies were done, or the new
evidence demonstrated no advantages, then coverage would continue, but at the
lower reimbursement. They write: “Of
course hospitals could continue charging patients more … and patients who
wanted the treatment could pay the difference themselves.”
So, should Medicare adopt dynamic pricing? I am not advocating Medicare withhold payment for medications that are used by dialysis patients. But, as resources become constrained at the federal level, policy makers will look to options like dynamic pricing as being attractive. Unless of course industry gets the message and starts investing in clinical trials to demonstrate the effectiveness of their drugs in dialysis patients.
So, should Medicare adopt dynamic pricing? I am not advocating Medicare withhold payment for medications that are used by dialysis patients. But, as resources become constrained at the federal level, policy makers will look to options like dynamic pricing as being attractive. Unless of course industry gets the message and starts investing in clinical trials to demonstrate the effectiveness of their drugs in dialysis patients.



Wait a minute. As a CKD patient myself, I've been taught for years that high serum phosphorus levels are very dangerous. Is that an exaggeration at best and a lie at worst? I've just started taking binders, and they make me feel awful. Are you saying that there is no proven reason for patients to take these things, that we have only "observational studies" to rely on? Seriously?
ReplyDeleteFor the CKD (non-dialysis) population, more recent cohort studies such as Kovesdy's VA study Konta's Japanese study found upwards of 39% mortality risk reduction on binders vs. without. (Yes, these are "observational" studies, but the samples were large and the reported effects were substantial.) This is for binders in general--the superiority of non-calcium binders is not clear.
ReplyDeleteUnfortunately, for the dialysis population, there is so much other noise contaminating the signal--that any survival benefit from binders is likely completely drowned out by other confounders. Perhaps this money might be better spent reducing the numerous other factors far more likely to cause death.
Additional dialysis might arguably be a more effective way to reduce phosphate levels, so perhaps we should spend the money there first. For preventing mortality, LVH prevention is far more important than MBD. Few patients die from fractures.
Sample size doesn't reduce the potential for bias in observational studies. The size of treatment effect doesn't reduce the potential for bias in observational studies. RCTs are needed.
ReplyDeleteRCT's can bring forth incorrect and erroneous treatment recommendations such as seen in the NCDS that implemented urea kinetics and ignored the effects of TIME on dialysis outcomes based on an insignificant 0.06 outcome. In this instance, the NCDS is clearly in error as we wade through the wake of 30 years of misplaced efforts to secure adequate dialysis that has left hundreds of thousands if not millions of patients dead and disabled.
ReplyDeleteWe are also overlooking the relationship between RCT's and observational studies to see if the results of an RCT are generalizable into the treatment population without all of the artificial selection restraints inherent in a modern RCT. The FHN nocturnal study is a failure due in part to this artificial restraint on participation and related recruitment difficulties. For an RCT to be valid, the population studied must be generalizable to the population of patients that it will be applied.
RCT's may offer the best chance of eliminating bias, but even here, all biases cannot be completely eliminated and RCT's can be plagued by other errors. The most important aspect of an RCT is to gain insight into causality. Association of outcomes is also an important aspect needed to practice optimal medicine.
Lastly, the BMJ had an incredible spoof on the lack of RCTs for parachute use that illustrates some of the problems with rejecting observational data. If anyone is willing to participate, the selection criteria for this RCT is completely unselected except by those willing to volunteer.
http://elucidation.free.fr/parachuteBMJ.pdf
Other sciences (e.g., astronomy, physics, general biology) do not have this neurotic obsession that all observational data is therefore inherently useless. Are they careful, accurate observations? Do the principles yield valid, accurate, REPEATABLE predictions? (THIS is the standard that most sciences use.)
ReplyDeleteMedicine ITSELF grew out of observational data and somehow managed to advance and improve for centuries prior to the invention of RCT.
Large RCT's can inform decision-making especially when comparing 99% equivalent drugs, or to pull such weak signals from the data, but that does not mean they are ALWAYS necessary.
Half as many people within a sample dying (for which we can't find any other explanation) PROBABLY "means" something, REGARDLESS of how the sample was created--so perhaps we OUGHT to be paying attention. "Probably" is not "definitely"--but it is still far more than the "completely meaningless" that everyone assumes.
We also deceive ourselves into thinking of randomization as a magic wand which automatically abolishes errors in sampling (selection bias), measurement, or interpretation of results. Worse yet, RCT designs can using composite endpoints which sometimes achieve significance at the expense of comprehensibility. Great. We've "proven" "something" - now, what does it mean again????
Who cares if the experiment was poorly designed or if we haven't a clue what the results mean. We randomized! Therefore it MUST be good science.
Don't use that parachute - it hasn't been adequately tested! We don't know what the risks are, (and parachutes have been known to fail.)