Saturday, January 17, 2015

Clinical Trialists Should Use Economies of Scale to Maximize Profits of Large RCTs

The lever is a powerful tool
I am writing (very slowly) a review article about ionized calcium in the ICU - should it be measured, and should it be treated?  There are several recent large observational studies that look at the association between calcium and outcomes of  critical illness, but being observational, they do not offer guidance as to whether chasing calcium levels with calcium gluconate or chloride will improve outcomes or whether hypo- or hyper-calcemia is simply a marker of severity of illness (the latter is of course my bet.)

Thinking about calcium levels and causation and repletion, one cannot help but think about all sorts of other levels we check in the ICU - potassium, magnesium, phosphate - and may other things we routinely do but about which we have no real inkling of an idea as to whether we're doing any patients any good.  (Arterial lines are another example.)  Are we just wasting our time with many of the things we do?  This question becomes more urgent as evidence mounts that much of what we do (in the ICU and elsewhere) is useless, wasteful, or downright harmful.  But who or what agency is going to fund a trial of potassium or calcium replacement in the ICU?  It certainly seems unglamorous.   Don't we have other disease-specific priorities that are paramount in importance to such a trial?

I then realized that a good businessman, wanting to maximize the "profit" from a large, randomized controlled trial (and the dollars "invested" in it), would take advantage of economies of scale.  For those who are not business savvy (I do not imply that I am), business costs can be roughly divided into fixed costs and variable costs.  If you have a factory making widgets you have certain  costs such as the rent, advertising, widget making machines.  These costs are "fixed" meaning that they are invariable whether you make 100 widgets or 10,000 widgets.  Variable costs are the costs of materials, electricity, and human resources which must be scaled up as you make more widgets.  In general, the cost of making each widget goes down as the fixed costs are spread out over more widget units.  Additionally, if you can leverage your infrastructure to make wadgets, a product similar to a widget, you likewise increase profits by lowering costs per unit.

A large multicenter RCT, like a business, also has fixed and variable costs and is, in theory at least, scalable.  The infrastructure that must be developed in terms of the protocol, the data management center, the data analysis, trial coordination center are all largely fixed costs.  Each "unit" or each patient that is enrolled is a variable cost, and fixed costs are spread out and the cost per unit decreases as more patients are enrolled.  Large trials are costly because of variable costs.  But here's the rub - almost all the fixed costs and a large portion of the variable costs offset the fixed and variable costs of a piggybacking sub-trial, making it almost free.

For example, suppose we have a large RCT of, say, enteral nutrition in the ICU but we also want to know whether potassium or calcium or whatever replacement is beneficial in basically the same population.  We piggyback a calcium replacement trial on the nutrition trial.  In so doing, the fixed costs are about the same as they were for the main trial (set up of coordinating center and so on) and the variable costs of each enrollment (screening, consent, monitoring, laboratory testing, data collection and entry, safety monitoring) are only marginally higher because of the relatively small costs of the additional laboratory assays for the calcium and the cost of the replacement solution.  Since the same outcomes can be utilized for both trials, extra data collection (a variable cost) is also minimized.  Similarly, patients could be enrolled in the first trial with its inclusion and exclusion criteria and then conditionally enrolled in the sub-trial if they meet additional inclusion and exclusion criteria for that trial, obviating separate screening and consent processes.

Purists will groan that they don't want to potentially compromise the integrity of their enteral nutrition study with the additional complexity of the electrolyte "rider" (that was a pun), that there may be interactions (these would be unexpected and could be tested for with interaction terms).  But these concerns underscore the point that I'm trying to make:  currently RCTs are about maximizing only one result, rather than about maximizing the "profits" (in terms of meaningful scientific discovery) of the entire research enterprise and dollars invested.

Funding agencies could and perhaps should take the lead by encouraging investigators to conduct multifaceted clinical trials to maximize the efficiency and profitability of the research enterprise and wring as much value as possible out of each research dollar invested.

1 comment:

  1. The RCTs must have accrued patients with 1 of 4 indications—nonspecific back or neck pain, shoulder pain, chronic headache, or osteoarthritis—with the additional criterion that the current episode of pain must be of at least 4 weeks duration for musculoskeletal disorders.