A great leap from clinical trials to clinical practice
There is potential for better tying clinical practice to clinical trials, but when will we get there?
In this era of SARS-CoV-2 we have an opportunity to educate the motivated. Pharmacology is front of mind for many of us, and there are some inconvenient truths about pharmaceuticals most don’t consider.
We now know that individuals respond to SARS-CoV-2 very differently with respect to depth of risk, time to recovery, organ systems affected, and chronic damage. We also have seen, in less than a year, very different conclusions about the effectiveness of experimental therapies from oxygen to anti-inflammatories to antivirals. We can’t be definitive about anything yet beyond that some approaches may benefit some of the patients some of the time at doses that remain unclear. On average a therapy may be helpful or not. But averages are not patients.
What is not well-appreciated is that this is normal. Every mammal from a companion animal to a grandparent subjected to pharmacology for the first time is participating in an informal N=1 clinical trial. This applies to the choice of an over-the-counter drug or a prescription to be filled.
Anyone following public reports on the trials of older drugs, newer drugs, or experimental drugs for SARS-CoV-2 will have heard declarations of miracles, tentative benefits, nothing positive, or adverse reactions. The reports have come from public figures, news outlets, medical journals, and public health agencies around the world. Quite often, these conclusions may apply—but never simultaneously. “Does it work?” is a very complex question. Viable answers are “It depends” or “Maybe.”
Let’s look at remdesivir as an example. Multiple trials have been carried out with mixed results (for example: Beigel, J.H. et al. N Engl J Med 2020;383:1813-26 and Goldman, J.D. et al. N Engl J Med 2020; 383:1827-37). The first of these studies utilized a 200 mg loading dose intravenously followed by 100 mg daily for the following nine days. This trial was placebo-controlled and double-blinded. The second study was randomized open-label with the same IV dose protocol, but for one group the treatment was terminated at day five and for the other at day 10. These studies (and others) were designed and executed early in the pandemic.
Reported conclusions suggested a benefit in reduced days in an intensive care unit on average for those with severe illness, suggesting the drug showed some benefit. Six months later the conclusions for remdesivir are very mixed. The largest study (the Solidarity trial), organized by the World Health Organization (WHO), reported that there was no faster recovery time or reduced mortality. The FDA granted an Emergency Use Authorization in early May and then approved the drug in late October. Wise or premature? More recently, a WHO panel recommended against the use of remdesivir (Veklury) while the manufacturer Gilead remains positive.
Just as things seemed to be focusing, we learn of other Veklury trials that appear to have been quite successful, including a combination with the Lilly arthritis drug baricitinib (Kalil, A.C. et al. N Engl J Med 2020; DOI: 10.1056/NEJMoa2031994). Other drug combinations are in still active trials.
Likewise, very positive clinical results were announced by OncoImmune for its Saccovid (now MK-7110). Merck was impressed and bought the company. This is all getting hard to follow, but again, the uncertainty is normal, and we remain in the early innings.
Automobile and smartphone models come off the assembly line with Six Sigma consistency. Humans do not. Therefore, we test drugs in subpopulations, including gender, race, age, size, diet, comorbidities. We usually start simple with healthy subjects (cancer drugs have long been an exception). If things go wrong there, it’s usually game over. Do no harm. When an experimental drug is dosed in a volunteer patient at the early stages there is much that can be measured in response to the several doses explored. With indications of positive results (pharmacology) and minimal negative results (toxicology) the trials are extended to larger populations.
With that extension, what is measured quickly becomes very sparse. Opportunity for understanding is thus lost. Should those COVID patients have all received 100 mg? That was a guess, perhaps a hurried one given the emergency. We really have no way of knowing how much drug those patients were exposed to over time.
The term “area under the curve” (AUC) describes this. Because drugs will clear from different patients at different rates and those rates will depend on metabolism, organ function (kidney), and other drugs in circulation, we can’t fully explore the benefits/risks with respect to exposure. Will it matter? We can’t know if we don’t measure. Will adverse events relate to exposure? Will time in the ICU relate to exposure? We’ve had no convenient means to make exposure measurements in clinical practice.
We use the term EC for effective plasma concentration against a virus. EC50 is half that. These are low-quality numbers—rules of thumb—varying with the virus. For bacterial infections, MIC is a related term, the minimum inhibitory concentration. Killing microbes tends to be logarithmic in response vs. concentration. EC50 and MIC50 are convenient ways of comparing the potency of different drugs. Success is also dependent on the time the infector is exposed to the drug. If we are not monitoring exposure over time in an individual, we are guessing.
“It is more important to know what kind of a patient the disease has, than to know what kind of a disease the patient has.” The literal origin of these popular words is not clear, although the meaning goes back to physician Claudius Galen (130-210 CE). COVID-19 exemplifies this old saw very well, being horrible in some of the patients some of the time but not all patients all the time. Cancer also fits very well. Every day in oncology clinics we are infusing drugs into very sick people with no measure of exposure at all. There is potential here for saving lives. To say “It’s impossible” is wrong. The better way is to say: “Let’s go”—and see what is needed to help some of those patients some of the time.
There is potential for better tying clinical practice to clinical trials. It may well help to make measurements of drug concentration vs. time. This is becoming feasible for hospitalized patients. In the remdesivir trials referenced above, blood samples were taken very sparsely for the purpose of determining drug concentration. Those samples likely ended up frozen and were not looked at for weeks. We thus can’t know if the efficacy of the drug might have been improved by adjusting the dose along the way. We can’t go back. Funding such a study in a new set of patients would be very costly. Will we get around to it? As this column has emphasized before, we don’t yet get around to it. 100 mg for every patient? That’s nuts.
I use remdesivir only as an example from current events. There are hundreds of generic drugs from decades past where improved safety and efficacy are a few measurements away.
What tools do we need to optimize exposure for finicky drugs and bring the great leap to a soft landing?
- Automated, time-stamped, no-blood-waste serial phlebotomy.
- Methods for drug monitoring that enable a rapid turnaround time for small sets of samples at known times post dose.
- Mathematical modeling of pharmacokinetics based on clinical trials and clinical experience over time but augmented with data for the individual patient.
- Guidance to adjust the dose: In the form of clinical decision support considering all relevant phenotypic and genotypic data as it becomes available.
Of these, the weakest link is currently number 2. Costs are too high for even a single measurement, but in many cases the costs of subsequent measurements are minimal once the validated method is launched.
With mass spectrometry, the first measurement is likely to cost 95 percent of the total, with the next five at 5 percent (to start a debate). The economics and speed improve when a series from a single patient are processed together.
There is much work published on number 3 for subpopulations. Several commercial firms provide software for modeling using sparse sampling from individual patients. With number 3 in hand, number 4 will be refined.
Over time, other phenotype and genotype data will be integrated. The whole process can be done in 24 hours or less. The science and math are known; the logistics and financial incentives remain weak both for pharma and the in-vitro diagnostics industry. Once a drug is approved, optimizing performance with respect to dosing is not a priority, particularly for pediatrics. The innovation engine of pharma moves over to the next experimental drug orphaning the earlier progress.
Clinical pharmacology gets less respect and less funding than it should. For many drugs, the best companion diagnostic is likely to be the drug concentration the patient is exposed to.
Peter T. Kissinger is a professor emeritus at Purdue University, founder of BASi, chairman of Phlebotics and director of both Prosolia and Tymora.