Modelling: a core technique in anaesthesia and critical care research
As papers published in this1 and previous editions of the journal2–5 illustrate, modelling is assuming a more prominent role in mainstream anaesthesia and critical care research, becoming an accepted methodology and an ever-more useful part of the research process. This editorial explores the phenomenon of modelling and its uses and limitations. The integration of modelling in a structured way into research projects can enhance almost any project's potential. Modelling of anaesthesia, critical care and pain processes has matured into a robust discipline, and should be welcomed as another technique for addressing the problems of research in the 21st century.What is modelling?
Modelling is the representation of a complex system, designed to facilitate understanding the system and to facilitate predictions of how it functions. Modelling is now a stand-alone methodology of similar value to laboratory investigations and clinical trials. Clinicians have become very good at clinical research, despite its shortcomings (e.g. subject heterogeneity, confounders, ethics, bureaucracy), although we have, generally, not been so good at modelling, or at least, not so good at making use of it.
From any original research question flow many paths of enquiry, containing a variety of methodological approaches and questions regarding the issue. What has been in short supply in these investigative paths is formalized modelling, which can offer a path of low resistance and high yield. Ethical considerations, subject compliance, pragmatic issues and subject heterogeneity are not relevant in modelling investigations and costs are usually substantially less than in clinical or laboratory research. However, this does not make modelling the poor cousin of clinical research. Sometimes, modelling may offer an easier route; such in silico research may offer insights not possible using in vivo research.
Development of a model is founded in the results of the research processes that have gone before. Previous research provides the structure needed to create models, and in particular, the numerical data rather than qualitative theorems. Thus, modelling is not a ‘new broom’ to sweep away previous techniques and findings, but merely a way of addressing the problems of today's research environment. It is not a ‘new’ philosophy for anaesthetic research—pioneers in anaesthesia research have made extensive use of modelling in pharmacological and physiological investigations—Mapleson's modelling of volatile and White and Kenny's work on i.v. anaesthetic agent pharmacokinetics6–8 have had huge impacts in the field of anaesthesia. Similarly, West's and Hahn's lung models9 10 have enabled substantial strides in our understanding of pulmonary pathophysiology. Modelling has been in use for many years, but has been used by only a minority of researchers and wider application of this highly productive methodology is possible. In today's research environment, modelling is a powerful tool in the researcher's armamentarium, an important ‘second string to the bow’ and a means by which we may facilitate collaboration with experts from other disciplines.
Modelling is the reproduction of real-life shape, form or function.11 12 It can be physical, mental, computational, statistical (probabilistic), mathematical, animal or laboratory. It is an attempt to simulate or recreate the important and interesting aspects of a question or system, while excluding the unpredictable and irrelevant aspects of organisms and populations that add heterogeneity, random variation and pragmatic obstructions. Such modelling may allow us to elucidate complex issues or to formalize or automate decision-making processes with a view to treating patients.13–16
Methods of modelling can be complex, but a great deal may be accomplished using simple approaches.17 18 Simple, numerical, computational models may be constructed with basic programming skills18 and fuzzy control systems19 and even neural networks are not beyond the reach of many researchers. The key, though, is in collaboration, and anaesthetists are well placed here. Research opportunities include: bench-top models of the upper airway,20 computational models of the lung,10 21 decision-support systems16 22 and costing models.23 In each of these cited examples, the investigators created a surrogate for examination, and the surrogate offered more accurate, reproducible, credible and feasible research outcomes than the alternative patient-based approach.
What does modelling have to offer the researcher?
This approach allows the testing of hypotheses, sometimes to the point of providing the definitive answer.24 Sometimes, such hypothesis-testing might prompt further model-based research or may lead onto clinical or laboratory-based investigation. Modelling may result in refinement of a research question, allowing more efficient use of an investigator's time and resources.
Dwindling research resources, fewer research staff, time constraints, funding pressures and the increasing complexity of the research environment (e.g. MHRA and COREC) have necessarily led researchers to consider their research hypotheses in isolation, without addressing the broader context in which the question is generated, its precursors, and the consequences that lie beyond the question asked. Seemingly simple questions frequently demand complex and laborious approaches for their resolution and we are compelled to take paths of high resistance and low yield. However, modelling may result in refinement of a research question, streamlining clinical or laboratory research and allowing efficient use of an investigator's time and resources.25 Short modelling projects that take a well circumscribed issue as a small bite from a big problem offer excellent opportunities for motivated but time-limited talent. In addition, the process of creating a model obliges the researcher to step back a little from the initial line of inquiry. The broader context of the question may then become clearer, and alternative, or better-focused questions may become apparent. Questions always have a context but many researchers ignore the context of a question, choosing to focus on an investigation of limited interest and narrow applicability. In viewing a question in context, researchers have the opportunity to identify appropriate investigative methodologies and collaborators at an early stage. These may encompass a multi-track approach. For example, clinical studies may utilize modelling to provide optimized patient stratification and inclusion/exclusion criteria; modelling studies may take a line of enquiry beyond the feasible reach of a clinical approach, and clinical studies may develop modelling research, moving forward and testing the novel theorems generated through modelling.
By considering the context of the questions we ask, we allow multiple strands within the research project to co-exist. These strands may run out, or may give rise to new strands, methodology leading into methodology, an answer spawning another question. These strands contain different questions and differing investigative methodologies: some clinical studies, some laboratory bench-top work, some in silico modelling and some a mixture. These strands do not run in isolation and in parallel, but may converge or bifurcate. Some will offer low resistance and high yield, while some may offer high resistance, but lead to a path of fruitful investigation. Investigators must be open to all the methodologies and tools available to them as medical researchers, choosing them appropriately as the situation demands. Above all, every question will benefit from the consideration of alternative investigative techniques, including modelling. This broader view, expanding the investigator's opportunities, can also increase the investigator's productivity, as the broader, contextual view impresses journal editors and funding charities alike.
Often, the process of model creation and refinement leads us to realize a deficiency in our knowledge. Issues of which we were certain are exposed as being understood qualitatively, and without quantitative basis. In such circumstances, modelling may prompt a search for data to fill the gap, developing new directions of research enquiry.
Collaborative model generation is an excellent way to combine the expertise of two disciplines who may otherwise find meaningful dialogue difficult through the paucity of shared knowledge. A model may contain and convey your expertise better than you can explain, allowing your collaborator to work alongside you far more effectively, bolstered with a deeper understanding of your own transferred expertise.
Is there validity in this approach to answering research questions? Experience in other fields suggests there is. Scale models of working prototypes are less sophisticated than a full-sized structure, but the substantial similarities make the derived data valid and pertinent. This approach was used by the aviation industries through the latter half of the 20th century for manned aircraft, and is still pursued vigorously in the new fields of unmanned and remotely piloted aerial vehicles.26 The results are less data rich than that gained from a fleet of full-size aircraft in a new environment, but they do answer questions of pertinence and specific relevance. Such models are controllable, affordable and flexible research tools and the data generated are highly-prized. In the clinical environment, modelling can go places that other forms of research cannot (e.g. lethal hypoxaemia).8 27 It can offer the chance to get rid of the perpetual confounding that is patient heterogeneity, although population variation can also be modelled, if required.
In fact, modelling runs, unrecognized, through all research methodologies. Surrogates, which are models by another name, are everywhere. We record the easily-measured (e.g. arterial pressure) as a substitute for what we really want to know (e.g. organ perfusion). We apply interpretative models to our clinical research to make its conduct feasible and its results publishable. For example, when we compare the arterial pressure responses in two groups of patients receiving different induction agents, we apply models of the way arterial pressure relates to the outcome of interest (e.g. cardiac output, injury, death), models of measurement (e.g. how equipment-derived numbers relate to the internal physiology of the patient) and statistical models (e.g. how each group of patients represents the population and how differences between groups may be extrapolated to differences between drugs). Frequently, these models are not formalized; usually, they are not even recognized as models and, consequently, are poorly thought-out and their intrinsic value ignored.
What new skills are required of researchers?
Anaesthetists and intensivists work at the junction of many disciplines: surgery, medicine, biology, pharmacology, mathematics and physics, and are well placed to embrace modelling. We have access to knowledge and expertise of enormous breadth and have experience of a huge array of induced and pathological states. We are comfortable with biological science, physical science, numbers, technology and medicine. Anaesthetists and intensivists, above all, have clinical contact, a real understanding of real-world relevancy and empathy for the issues of importance. In addition, we have skills in managing teams of individuals, collaborating and coordinating their efforts towards a single goal. We have the skills, the opportunities and worthwhile goals. All that is required of the researcher who wants to use modelling is to get an idea of what may be achieved and find a suitable question to answer. Contact with an expert will be enormously helpful during the researcher's early forays into modelling. The authors of this editorial have found that collaborating and pooling resources with others in order to construct models is widely welcomed and very rewarding.
Any new techniques have their limitations, and modelling is no different. The credibility of model-based investigations relies upon readers' understanding of the strengths and limitations of modelling research and their acceptance of modelling as a viable paradigm for research. Just as readers currently can assess the strength of a clinical research manuscript, they must also develop the ability to appreciate correct model validation, application and limitation. All manuscripts describing research that is based in modelling must make their model valid in the eyes of the consumer. This usually involves demonstrating the acceptable similarity of the model and real patients and/or processes. Crucially, models must be validated in the task or application to which they are put. Generally, model space is the term used to describe the areas within which a model is validated and credible. Application of a model outside the model space may give rise to spurious or frankly nonsensical results, while application within the model space is credible and potentially useful.
Some systems are too complex to model adequately. A single model of the central nervous system, designed to elucidate the mechanism of seizures, is unrealistic, unless the question is divided into manageable parts and each addressed individually. Systems often are broken into components, and modular models may be developed and validated separately. As the need arises, such a modular system may be joined together. Even if each module has been shown to be valid, discrete validation is required of the newly assembled multi-modular model.
In the near future, modelling will allow us to develop closed-loop anaesthesia, ventilators that optimize patients' arterial gases and wean patients from mechanical support, evidence-based cardiac arrest management, cost-savings through operating department and intensive care unit efficiency, lung-protective ventilation and novel disease monitoring strategies and diagnosis. However, rather than a specialty in its own right, we must view modelling as an important tool in research just as statistics, ethics, laboratory techniques and software skills are currently considered useful tools. Modelling runs through all of our endeavours, and we stand to benefit hugely by becoming acquainted with this powerful device.
1 University of Nottingham and Queen's Medical Centre Nottingham, UK
2 University of Sheffield and the Northern General Hospital, Sheffield, UK
*E-mail: J.Hardman{at}nottingham.ac.uk
References
1 Yem JS, Turner MJ, Baker AB, Young IH, Crawford ABH. A tidally breathing model of ventilation, perfusion and volume in normal and diseased lungs. Br J Anaesth 2006; 97:718–31
2 Hope DA, Jenkins BJ, Willis N, Maddock H, Mapleson WW. Non-invasive estimation of venous admixture: validation of a new formula. Br J Anaesth 1995; 74:538–43
3 Moppett IK, Gornall CB, Hardman JG. The dependence of measured alveolar deadspace on anatomical deadspace volume. Br J Anaesth 2005; 95:400–5
4 Farmery AD and Roe PG. A model to describe the rate of oxyhaemoglobin desaturation during apnoea. Br J Anaesth 1996; 76:284–91
5 Joyce CJ and Hickling KG. Permissive hypercapnia and gas exchange in lungs with high Qs/Qt: a mathematical model. Br J Anaesth 1996; 77:678–83
6 Mapleson WW. Circulation-time models of the uptake of inhaled anaesthetics and data for quantifying them. Br J Anaesth 1973; 45:319–33
7 Mapleson WW. Mathematical aspects of the uptake, distribution and elimination on inhaled gases and vapours. Br J Anaesth 1963; 36:129–39
8 White M and Kenny GN. Intravenous propofol anaesthesia using a computerised infusion system. Anaesthesia 1990; 45:204–9[Web of Science][Medline]
9 West JB. Regional differences in gas exchange in the lung of erect man. J Appl Physiol 1962; 17:893–8
10 Hahn CEW and Farmery AD. Gas exchange modelling: no more gills, please. Br J Anaesth 2003; 91:2–15
11 Greenway CV. Mechanisms and quantitative assessment of drug effects on cardiac output with a new model of the circulation. Pharmacol Rev 1981; 33:213–51[Web of Science][Medline]
12 Masuzawa T, Fukui Y, Ty Smith N. Cardiovascular simulation using a multiple modelling method on a digital computer–simulation of interaction between the cardiovascular system and angiotensin II. J Clin Monit 1992; 8:50–8[CrossRef][Web of Science][Medline]
13 Bates JHT and Young MP. Applying fuzzy logic to medical decision making in the intensive care unit. Am J Respir Crit Care Med 2003; 167:948–52
14 Hoekesl SAAP, Blom JA, Jansen JRC, Maessen JG, Schreuder JJ. Automated infusion of vasoactive and inotropic drugs to control arterial and pulmonary pressures during cardiac surgery. Crit Care Med 1999; 27:2792–8[CrossRef][Web of Science][Medline]
15 Lau F. A clinical decision support system prototype for cardiovascular intensive care. Int J Clin Monit Com 1994; 11:115–69
16 Kwok HF, Linkens DA, Mahfouf M, Mills GH. Rule-base derivation for intensive care ventilator control using ANFIS. Artif Intell Med 2003; 29:185–201[CrossRef][Web of Science][Medline]
17 Hanson CW III and Marshall BE. Artificial intelligence applications in the intensive care unit. Crit Care Med 2001; 29:427–35[CrossRef][Web of Science][Medline]
18 Hardman JG, Bedforth NM, Ahmed AB, Mahajan RP, Aitkenhead AR. A physiology simulator: validation of its respiratory components and its ability to predict the patient's response to changes in mechanical ventilation. Br J Anaesth 1998; 81:327–32
19 Mason DG, Ross JJ, Edwards ND, Linkens DA, Reilly CS. Self-learning fuzzy control with temporal knowledge for atracurium-induced neuromuscular block during surgery. Med Biol Eng Comput 1997; 35:498–503[Web of Science][Medline]
20 Buczkowski PW, Fombon FN, Russell WC, Thompson JP. Effects of helium on high frequency jet ventilation in model of airway stenosis. Br J Anaesth 2005; 95:701–5
21 Hardman JG and Bedforth NM. Estimating venous admixture using a physiological simulator. Br J Anaesth 1999; 82:346–9
22 Ross JJ, Mahfouf M, Denai MA, King OK. Modelling decision–making strategies in supporting the shocked patient. Br J Anaesth 2006; 97:432
23 Negrini D, Mills GH, Jacobs P, Edbrooke NR, Edbrooke DL. Intensive care costing and international comparisons. Clin Int Care 2005; 16:27–31
24 Heldt T, Shim EB, Kamm RD, Mark RG. Computational modelling of cardiovascular response to orthostatic stress. J Appl Physiol 2002; 92:1239–54
25 Tjardes T and Neugebauer E. Sepsis research in the next millennium: concentrate in the software rather than the hardware. Shock 2002; 17:1–8[Web of Science][Medline]
26 Beckett N. BAE Systems UAVs—closing the capability gap. Air Int 2006; 71:57–60
27 Hardman JG, Will JS, Aitkenhead AR. Factors determining the onset and course of hypoxemia during apnea: an investigation using physiological modelling. Anesth Analg 2000; 90:619–24
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  A. Ercole The Effect of Injectate Conductivity on the Electric Field with the Nerve Stimulator Needle: A Computer Simulation Anesth. Analg., October 1, 2008; 107(4): 1427 - 1432. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||