Developing Treatment Guidelines for Potentially Lethal Infections
Rapidly fatal infections need urgent treatment with optimum doses of appropriate antimicrobials. Such doses should ideally produce maximum effects as quickly as possible, and provide the greatest differential between lives saved and lives lost because of toxicity. If the antimicrobial drug is not eliminated rapidly, a loading dose should be given to provide therapeutic concentrations as soon as possible.
This dosing strategy for rapidly fatal infections contrasts with dose recommendations for uncomplicated infections—such recommendations are aimed at lower microorganism burdens, where rapidity of action is less important and adverse effects are of greater significance. In other words the risk-benefit trade-off, commonly termed the therapeutic ratio, is different in severe and uncomplicated infections.
This difference has important implications for treatment guidelines. According to the World Health Organization (WHO): “Guidelines are formal advisory statements which should be robust enough to meet the unique circumstances and constraints of the specific situation to which they are being applied” . Treatment guidelines are best when they rest on a sound and consistent evidence base ,. Randomised clinical trials (RCTs) are considered to provide the best evidence. But what if the evidence from controlled trials is insufficient, or there simply isn't any? Guidance and specific recommendations are still necessary. Inadequate initial treatment of life-threatening infections has serious consequences. Therefore, common sense argues for recommending higher doses for such infections, at the expense of increased toxicity, to avoid any possibility of under-dosing those patients with unusual pharmacokinetics and more resistant organisms. If intravenous administration is not possible, absorption from the gut or intramuscular injection site may be compromised in the most seriously ill, arguing again for higher doses.
In this context of critical uncertainty, and against a background of concerns over liability and consequent risk aversion, physicians often seem more worried about the risks of adverse effects than of under-dosing, even though antimicrobial adverse effects are rarely fatal. Seldom is an infectious disease death ascribed to administration of inadequate doses.
Treatment Guidelines for H5N1 Influenza
H5N1 influenza is regarded by many as the greatest threat to human health and national security . Fortunately human infections are still rare, but this rarity also means that there are no published RCTs of treatment. The oral viral neuraminidase inhibitor oseltamivir (Tamiflu) is considered the drug of choice ,. There is no parenteral formulation. H5N1 influenza replicates more rapidly than seasonal influenza viruses ,, reaching much greater viral burdens than do other human influenza viruses . Resistance arises readily. Mortality consistently exceeds 50%, which puts H5N1 influenza amongst the most lethal of human infections.
Experimental studies with H5N1 viruses in animal models suggest that high doses and long courses of neuraminidase inhibitors provide optimal treatment ,. Despite this evidence, a “rigorous and transparent” process, led by WHO, to develop treatment guidelines for H5N1 influenza has recommended an adult dose of 75 mg twice daily for five days. This is the “standard” dose regimen for uncomplicated seasonal influenza ,. If absorbed well, this 75 mg dose might provide maximal neuraminidase inhibition at the sites of infection in all patients seriously ill with H5N1 influenza. In other words it might be enough, but the truth is that we just do not know. The concentration-effect relationship in patients has not been characterised. Oseltamivir doses of up to 1,000 mg have been given to volunteers. High doses of oseltamivir are reasonably well tolerated in humans, and there is experimental evidence to suggest they could be more effective ,,,,.
There seems little to gain and everything to lose by using a low dose of this potentially life-saving drug in a highly lethal infection. The “evidence-based approach” (Grading of Recommendations Assessment, Development and Evaluation or GRADE; see http://www.gradeworkinggroup.org/), now considered “state of the art” for guideline development ,,, has been constrained by lack of RCT evidence on higher doses of oseltamivir. In recent years a hierarchy of the quality of evidence has been increasingly promoted, particularly for the formulation of guidelines. “Hierarchies place randomised controlled trials (RCTs) at their summit, with various forms of observational studies nestling in the foothills,” says Rawlins , but information from observational studies and other foothill inhabitants (experimental investigations, analogy with similar conditions and processes, pathological and pharmacological understanding and reasoning, and a derived assessment of risks and benefits) is also valuable. In the case of pandemic influenza, the GRADE process has resulted in a dose recommendation for H5N1 influenza that could be too low.
Using All the Evidence To Assess the Risks and Benefits
In contrast to the GRADE approach, a “mechanism-based” approach, incorporating current understanding of this lethal disease and of antimicrobial pharmacology, and assessing the risks and benefits, would lead to initial use of the highest oseltamivir doses considered to have a low risk of major toxicity in H5N1 influenza. This fundamental difference in analytical and deductive approaches is analogous to the frequentist versus Bayesian debate in statistics. Rawlins has recently articulated the important limitations of relying too much upon evidence from RCTs and has argued cogently for greater use of a Bayesian approach in decision making on recommendations for therapeutic interventions . Different approaches to the same problem may yield different results initially, although as evidence accrues, results of the two approaches tend to converge.
The current approach to guideline development may be too restrictive. Where there is little or no direct evidence from RCTs, the current “evidence-based approach” to treatment guidelines certainly needs reconsideration. In the absence of direct evidence on dosing in a rapidly lethal infection, basic precepts of antimicrobial pharmacology (“Bayesian priors”) and common sense argue that the highest possible dose of an antimicrobial should be used, at least initially, until evidence becomes available to inform the recommendation.