A Blood Test to Predict the Future

Over the years, the Journal has received manuscripts that can best be described as falling into the category of “a blood test to predict the future.” These are manuscripts most often submitted by younger colleagues who are trying to build their investigational credentials. Surgery and particularly cardiothoracic surgery have a long tradition of proposing the use of a blood test to “predict patient outcome and the occurrence of postoperative complications” to quote from a recent manuscript. After all, surgery is often a predictable event with a well-defined start and end, and particularly in the realm of cardiothoracic surgery a well-defined conduct of operation.

It appears, therefore, worthwhile to spend a moment to look at the requirements for what such a test may look like. It may seem intuitive that the sum of the probabilities of all post-op complications should equal 100%. It follows that a single complication can be derived from a complex array of symptoms and derangements. However, applying this principle of a single cause (often referred to as Occam's razor)[1] to a complex set of circumstances discounts the possibility of more than one thing going on. Yet, it is so attractive to surgeons because the one event they know is when surgery took place. Surgeons like studies along a classic and well-established pathway where several biomarkers are chosen and assayed sometimes before, during, or shortly after surgery. Then an outcome is observed and subsequently it is investigated if there is a difference in the chosen biomarkers for that particular outcome. It may be helpful to review the term “biomarker.” The term was first accepted by the US Food and Drug Administration in 1989.[2] To quote from Circulation where a statement from the Center of Disease Control and Prevention and the American Heart Association was published in 2003: Biomarkers are measurable and quantifiable biological parameters that can have an important impact on clinical situations. Ideal biomarkers are those that are associated with disease, clinical end points in observational studies, and clinical trials, and in some cases, they may even be used as surrogate end points. Biomarkers must also be both independent of established risk factors and recognized to be a factor in the disease for which they are a marker. The normal physiological expression of the potential biomarker must also be known to interpret results, as well as to generalize results to various population groups. Finally, potential biomarkers must also have the ability to improve overall prediction beyond that of traditional risk factors, while assays to detect them must have an acceptable cost and be subject to standardization to control for the variability of measurements.”[3] The definition was further refined over the years.[4]

There are some other factors that also deserve consideration. A test should only be done if its result will affect management. When the disease pretest probability is above a certain threshold, treatment is warranted and testing is not indicated. Below the treatment threshold, testing is indicated only when a positive test result would raise the post-test probability above the treatment threshold. When the diagnosis has some uncertainty or depends on a test with some uncertainty as it is almost always the case, the decision to treat must balance the benefit to treat a sick patient versus erroneously treating a well patient and doing some harm, even if this just means spending unnecessary money.

So, what does that mean if “disease” or complication is dependent on one test? The most common medical tests provide results along a continuous quantitative scale, but clinicians use them by classifying at test positive or negative based on some established criterion or cutoff. Cutoffs are based on some statistical and conceptual analysis that attempts to balance the rate of false positive results against false negative results. Identifying cutoffs also requires a gold standard. The other underlying concept is that patients with and without disease/complications have two different mean test values. This implicitly implies that there is an overlap of results. Thus, the establishment of receiver operating characteristic (ROC) curves where true positives are plotted against false positives which allows for different cutoffs is helpful. The greater the area under the curve, the better the test is able to discriminate.

In modern cardiac surgery, most procedures carry a mortality of < 2% and a morbidity of around 5%, that is, events are rare. Most patients stay ∼5 days in the hospital. Therefore, construction of reliable cutoff points takes thousands of patients. This is often complicated by the fact that the investigator wants to obtain the test after surgery and predict “disease”/complications for the next 5 days. The test should be independent of ever improving risk models such as the Euroscore or the Society of Thoracic Surgeons score which are based on thousands of patients. While these models are steadily refined, there has not been one routine new biomarker test introduced into routine clinical cardiothoracic practice in the past 20 years. If one wanted to attempt such an endeavor, it is proposed that the test fulfill the following criteria:

1. Be highly discriminative (ROC > 95%)

2. Be highly accurate

3. Make a difference to the patient

4. Fast turnaround (point of care/real-time if it is to make a difference within 5 days)

5. Ubiquitous available

6. Cheap

Particularly the inflammatory reaction present after all surgery but more so after surgery using cardiopulmonary bypass appears attractive to the search for biomarkers. This appears to stem from the idea that the inflammatory reaction is initiated by surgery and then develops in a cascading orderly fashion so that if the induction could be suppressed, all the subsequent expression of molecules/phenomena could be suppressed avoiding complications. Suffice it to say that the concept of “homeostasis” introduced by Claude Bernard 150 years ago holds that inflammation is part of the “milieu interieur” necessary to assure survival. Thus, suppressing all or parts of it has serious often unforeseen consequences. It also is highly individual and dependent on the “milieu exterieur.”[5] Thus far, despite studying these phenomena for 25 years or more, no biomarker out of this line of research has made it into routine cardiothoracic surgical practice. This should serve as a word of caution for colleagues who are embarking on developing a predictive biomarker test.

• References

• 1 https:// www.britannica.com/topic/Occams-razor
• 2 Biomarkers Definitions Working Group. Biomarkers and surrogate endpoints: preferred definitions and conceptual framework. Clin Pharmacol Ther 2001; 69 (03) 89-95
• 3 Naghavi M, Libby P, Falk E. , et al. From vulnerable plaque to vulnerable patient: a call for new definitions and risk assessment strategies: part I. Circulation 2003; 108 (14) 1664-1672
• 4 Chow SL, Maisel AS, Anand I. , et al; American Heart Association Clinical Pharmacology Committee of the Council on Clinical Cardiology; Council on Basic Cardiovascular Sciences; Council on Cardiovascular Disease in the Young; Council on Cardiovascular and Stroke Nursing; Council on Cardiopulmonary, Critical Care, Perioperative and Resuscitation; Council on Epidemiology and Prevention; Council on Functional Genomics and Translational Biology; and Council on Quality of Care and Outcomes Research. Role of biomarkers for the prevention, assessment, and management of heart failure: a scientific statement from the American Heart Association. Circulation 2017; 135 (22) e1054-e1091
• 5 Holmes FL. Claude Bernard, the milieu intérieur, and regulatory physiology. Hist Philos Life Sci 1986; 8 (01) 3-25