I was thinking about submitting this for a pioneer award but upon learning of the near impossible odds and apparent requirement for gravitas, which I don’t have enough of, I decided against it. I put it on the blog to lay out the idea and solicit comment.
Establishing efficacy of new neuroprotective therapies in neurocritical care and stroke has proven to be an exercise in futility. Over 475 completed clinical trials are listed on the Internet Stroke Trials Registry with few apparent reproducible results of any demonstrable efficacy in the acute context. However, these many negative studies belie the supportive basic laboratory studies that justified the time and enormous expense for such translational clinical trials. I will provide a rationale to suggest that such results, in retrospect, are altogether predictable, suggest an explanatory model for such reproducible futility in a complex biological system, and propose a research program to develop a multifaceted approach which, taken altogether will produce breakthrough level data which will then be generalizable to for multi-institutional application or study. Donnan  in the 2007 Feinberg lecture suggests: “We have reached a stage at which research in this area should stop altogether or radical new approaches adopted." The proposal is an answer to his plea for a new approach.
I. Impact of perturbations in complex systems.
Single facet futility---multifacet therapeutic breakthrough
Imagine a factory that makes widgets. A number of processes are important for the quality of the final widget as it proceeds: conveyor speed(x1), presence of raw materials and power(x2), quality of bolts(x3), quality of steel(x4), and type of metal used for circuits(x5). A weighting factor can be applied to each variable wi leading to the following general equation describing the widget quality:
Each variable x can be precisely known with very small variation so any change in any of the variables will produce a reproducible and predictable change in the widget quality Q.
In a biological system characterized by severity of a pathophysiologically complex injury, S, a similar equation can be derived with important pathophysiologic factors, xi, and weighting factors, wi:
Notably different from the widget however is that there are a large number of disparate and potentially interacting factors known to contribute to S with also an unknown number of as yet unknown factors with correspondingly unknown weighting factors and variability. Moreover, each pathophysiologic factor xi has to be described over a biologically diverse population such that each factor has an associated central tendency and large normal or non-normal distribution about that mean. Additionally, in the context of clinical medicine there are also associated system factors, Hi, like nursing ratio, nursing experience, availability of drugs and technology, efficiency of rapid response teams, and so on, which are also important to the severity of injury such that the equation can be written as
S=∑WiXi + ∑WiHi
Given the above characterization of the multiple highly variable biological and system factors that enter into a given outcome, it should come as no surprise that clinical studies directed at improving only one of the above noted numerous complex factors tend to show no effect, especially if multi-institutional in design (increasing variation in H factors), unless it is truly a breakthrough phenomena (large W factor like early thrombolysis in ischemic stroke) or the therapy exerts a multifaceted effect (e.g., hypothermia). This then leads to the notion that the current widely accepted methods of advancing clinical knowledge for complex problems is generally a fruitless waste of public resources which produces innovation paralysis on the part of institutions, third party payers, clinicians, pharma, and investigators, and that an alternate method is needed which is based on a multifactorial approach. Rogalewski et al  have recently reviewed and endorsed this concept, however, they fail to suggest a rational means for building the multimodal approach other than trying everything at once…another prescription for trouble. A rational method is needed.
II. PDSA (Plan-Do-Study-Act) Cycles -- From QI To Generalizability
A process designed to produce local improvement in the quality of care (QI), the so-called PDSA method advocated by Berwick  could provide a means to develop an incrementally implemented multifactorial approach as a means to serially test and add single clinically unproven but safe, pathophysiologically sensible and scientifically supported facets in the therapy of a disease…eventually producing a multi faceted approach which would then be amenable to more widespread testing and/or application.
The PDSA method entails application of a nonrandomized process, using institutional and individual (paired or N of 1) historical control data, to introduce incremental improvements in processes of care. The first step, planning, entails identification of a process to improve with a plan for implementation. The do phase entails the actual systematic implementation of a new process of care. Studying then is the procedure for collecting and analyzing the results of the new intervention. And the act phase provides for coming to a conclusion that the results are worthy or not worthy of then producing a permanent change in a system’s health care protocols.
The PDSA cycle is designed for processes of therapeutic QI which are meant to improve process-of-care outcomes locally and not necessarily produce generalizable knowledge. Thus by the federal definition such processes technically are not research. Nonetheless the PDSA process is a widely accepted type of institutional process-of-care experimentation which very well could be a method to produce or contribute to the production of new generalizable knowledge.
III. Using a single NICU with PDSA cycles to develop a multifaceted program to arrest secondary brain injury.
A busy NeuroICU admits patients with traumatic brain injury,
intracerebral hemorrhage, subarachnoid hemorrhage, subdural hemorrhage, ischemic
stroke, and spinal cord injury. Each of
these diseases is associated with a
variety of similar secondary injury processes
which each individually are suitable targets for therapeutic intervention. One example of the widespread, albeit
incomplete, nature of these factors is illustrated in figure 1. Each of these
factors is associated with nucleotide polymorphisms such that a significant
variation between individuals can be expected in the response of any given
factor in the response to the initial insults, production of secondary injury,
and response to focused therapeutic interventions. Thus, as suggested by the equations modeled
in section I, these large variations in each factor makes it even more
difficult to identify a therapeutic effect. Moreover, treatment of one factor may be associated with exacerbation of
other factors. For example initial
efforts to identify calcium antagonists in the treatment of vasospasm were
fraught with problems with hypotension, which acted to exacerbate the ischemic process.
These notions lead to the proposed concept to incrementally introduce therapeutic measures, each individually directed against single facets in known processes of secondary injury after a neurologic insult, followed by three primary measures:
1. Impact on the target pathophysiologic pathway
2. Surrogate neurologic outcome measures
3. Functional neurologic measures.
Notably, chosen pathophysiologic goals should be widely
accepted to have neuroprotective potential based on either laboratory animal
studies or clinical studies. Modalities that prevent extremes of CBF,
anaerobic metabolism, glutamate toxicity hyper-metabolsim, and peroxidation are
examples of accepted pathophysiologic goals. Providing a therapy that produces such a goal will then be followed by
surrogate measures of neurologic effect along with subsequent measures of clinically
meaningful outcome. The decision flow
diagram is illustrated in figure 2.
Some examples of PDSA cycles with concomitant pathophysiologic goals that can be envisioned are described below:
· Adjust blood pressure and paCO2 to optimize CBF. Over the first five days after an insult using continuous measurement technology and portable XeCTCBF optimize CBF aided by measures of ICP, pbO2, EEG, and MD L/P.
· Administer antioxidant drugs. Measure urinary isoprostanes at baseline and after implementation of therapy to confirm an antioxidant effect.
· Administer magnesium infusions to control shivering. Evaluate effects on serum magnesium levels, shivering index, temperature, and MD Mg++, glutamate and glycerol levels
· Administer reserpine as an endogenous sympatholytic. Evaluate impact of this therapy on MD catecholamine levels.
Abbreviations: CBF-cerebral blood flow, MD-microdialysis, L/P lactate pyruvate ratio, pbO2- pO2 of brain tissue, ICP-intracranial pressure, EEG- continuous electroencephalography,
If the neurophysiologic or neurochemical pathway objective
is achieved at a statistically meaningful level and there is no evidence of
harm based on within patient and overall population neurological and other
medical measures, then the new therapy is added to the multifaceted protocol as
local standard of care. Then, using the PDSA process incremental therapeutic
facets are added sequentially. Each intervention
is evaluated for surrogate neurologic measures of impact on: (a) biochemical
biomarkers of brain injury; (b) CT-
and (c) MR-based infarct morphometry. Clinically meaningful functional outcomes are
evaluated one and six months after admission and both surrogate and functional
outcomes are statistically evaluated for a multifaceted therapy effect over the
course of the QI project.
CT- and (c) MR-based infarct morphometry. Clinically meaningful functional outcomes are evaluated one and six months after admission and both surrogate and functional outcomes are statistically evaluated for a multifaceted therapy effect over the course of the QI project.
By the end of the project multiple new neuroprotective therapies (vs secondary injury)will be employed for several clinical problems with concomitant QI measures for surrogate and actual neurologic outcomes at baseline versus the end of the project as part of an IRB overseen PDSA process. A robust difference in surrogate and actual outcomes should be apparent. These results will then form the basis for a rationally developed multifaceted approach to prevention of secondary injury for multiple neurocritical care diseases with robust effects apparent due to the multimodal approach. These resultscan then be offered in aggregate as multifaceted therapies ready for multi institutional trials or…for a true breakthrough effect… possibly for immediate generalized implementation. Indeed this might be a new alternate model for creation of new knowledge for complex medical problems.
1. Stroke Trials Registry. The Internet Stroke Center 2008 January 4, 2008 [cited January 7, 2008]; Available from: https://www.strokecenter.org/trials/index.aspx.
2. Donnan, G.A., The 2007 Feinberg lecture: a new road map for neuroprotection. Stroke, 2008. 39(1): p. 242.
3. Rogalewski, A., et al., Toward a multimodal neuroprotective treatment of stroke. Stroke, 2006. 37(4): p. 1129-36.
4. Berwick, D.M., Developing and testing changes in delivery of care. Ann Intern Med, 1998. 128(8): p. 651-6.
5. Siman, R., et al., Novel surrogate markers for acute brain damage: cerebrospinal fluid levels corrrelate with severity of ischemic neurodegeneration in the rat. J Cereb Blood Flow Metab, 2005. 25(11): p. 1433-44.