Clinical Index of Stable Febrile Neutropenia (CISNE)

Interview provided by Dr. Carmona-Bayonas and Dr. Jiménez Fonseca.

Why did you develop the CISNE? Was there a particular clinical experience or patient encounter that inspired you to create this tool for clinicians?

CISNE means “swan” in Spanish, and in fact, the metaphor of Nassim Nicholas Taleb's black swan theory is applicable to this clinical scenario, since complications in patients with apparently stable febrile neutropenia have a high impact. These complications are difficult to predict through routine clinical inspection because of the frequent absence of symptoms and/or signs. When they have already occurred, someone may think in hindsight that they could have been avoided with an appropriate classification. It's therefore an inspiring philosophical argument.

The idea of the CISNE Score arose in the emergency department of a high-volume hospital. We hypothesized that not all cases with febrile neutropenia had the same degree of prognostic uncertainty. Dying patients, or those with very serious infections, did not really require a stratification tool to provide additional information about a clinical picture that was already expressive enough to guide admission.

However, as a result of immunosuppression, the human body is temporarily unable to generate inflammatory reactions. As a consequence, in the early stages of neutropenic infections, some patients may show a deceptive state of health. The aim of the CISNE Score is to provide a greater degree of certainty about whether or not the apparent stability is real, which may prevent early discharges from hospital in patients at risk and may assist the physician in decision-making, something that other models do not perform.

What pearls, pitfalls and/or tips do you have for users of the CISNE? Do you know of cases when it has been applied, interpreted, or used inappropriately?

The misuse of CISNE should result in decision-making errors, and we have detected some examples of this kind. First, patients should not be evaluated only with a numerical score—it is essential to take full account of the general principles of febrile syndromes in immunocompromised patients. The overall picture needs to be addressed, not just a number. It seems obvious, but when one reviews the literature, there is a strong tendency to believe that decisions are made by numbers alone. In this sense, CISNE is safe because it is based on low-risk criteria that are pre-established in the literature (e.g. ASCO’s clinical guide on febrile neutropenia, Flowers 2013). In contrast to other scores, CISNE reinforces the safety of these appraisals, providing additional information, but does not contradict them, since it has been developed specifically for patients considered apparently stable by other methods including vital signs and physical examination.

The second problem we have observed is that some researchers are tempted to also apply CISNE to the group of unstable patients, but the model was not designed for them. In fact, there is no point in using CISNE to evaluate the prognosis of immunosuppressed patients that we already know are at high risk before applying the model. That has nothing to do with CISNE's philosophy, and obviously, the model will not give the right results after a poorly formulated question.

Finally, CISNE's aim is to delay the premature discharge of patients with potential risk until the apparent stability has been confirmed as real. Instead, some authors try to use it to directly select low-risk subjects for reductions in the level of supportive treatment, which is not the focus of the calculator.

What recommendations do you have for doctors once they have applied the CISNE? Are there any adjustments or updates you would make to the score based on new data or practice changes?

First, keep in mind that febrile neutropenia involves two distinct problems. You must get an idea of the risk of unexpected serious complications, but on the other hand you should also consider the probability of strains of resistant or unusual pathogens. These are two absolutely different factors, but they must be integrated into the decision-making process at the same time, since there is a risk of turning a low-risk individual into a high-risk patient as a result of failure to categorize either of them.

Second, I recommend that the starting point should be the general principles of febrile syndrome assessment of the immunocompromised patient in the ED. For me, the best guide is the ASCO clinical practice guideline (Flowers 2013), which has a concise table with all the clinical criteria to systematically consider whether a patient is at high or low risk.

Third, CISNE can be used as a clinical aid, but not as a unique decision-making tool. It should also be noted that the CISNE calculator was not designed to select patients for outpatient management, but rather to delay the early discharge of doubtful patients until the apparent stability is confirmed as real.

How do you use the CISNE in your own clinical practice? Can you give an example of a scenario in which you use it?

In my routine practice, I use the CISNE Score when, through other criteria, I have come to the conclusion that the patient may be a candidate for some type of reduction in supportive treatment. In that case, I use CISNE's high-risk criterion as a screening tool, to delay that early discharge decision until I have checked the negativity of blood cultures, and have ensured through hospital observation for a reasonable period of time that the apparent stability is not fictitious. On the contrary, I would never use it as an exclusive criterion for decision making, or in patients who, for additional reasons, would not be candidates for reduction in support.

What are your thoughts on using CISNE over the MASCC Risk Index for febrile neutropenia?

My personal thought is that people evaluate the MASCC model and leave it written in the medical record, but decisions are almost never made based on its prediction. The MASCC score has been validated on multiple occasions, and is recommended by most international scientific societies. However, it is virtually useless in providing additional information about these patients because it was not designed to be useful. Among its recognized limitations, the heaviest predictor is hypotension, which coincides exactly with the most frequent endpoint (also hypotension), making the prediction result useless (the variable oddly predicts itself). If hypotensive subjects are eliminated, the sensitivity drops to about 30%.

As if that were not enough, two other predictors (solid tumor and ambulatory episode) are not useful for medical oncologists who evaluate subjects in the ED, as the variables are always present. The MASCC sample was not truly representative of patients with solid tumors because it included a very high proportion of patients with bone marrow transplant or acute leukemia in induction therapy. Nor does it serve to select candidates for outpatient management because many of them had very serious clinical conditions from the beginning, so they would never have been candidates for the therapeutic strategy that the model ultimately claims.

The “burden of disease” variable is subjective; between a 59- and 60-year-old patient, we don't think there is any real difference, etc. With these premises, it does not even make sense to consider the predictive values of this model. Instead, CISNE tries to provide useful information that cannot actually be obtained in any other way at present, and includes variables that may make sense, such as stress hyperglycemia, a biomarker of poor clinical outcomes known since Claude Bernard's time.

The risk percentages for complications quoted in the FINITE study are slightly different from the ones on the official CISNE calculator. Where did that data come from?

The method followed in the construction of the official CISNE calculator is explained in an article in the British Journal of Cancer published after the publication in the Journal of Clinical Oncology, with data from the original FINITE study.

Basically, the small differences are due to the fact that to build the nomogram, the models' coefficients were updated and retrained, bearing in mind the entire FINITE series (n = 1,133). Subsequently, the results were validated by Dr. Ignacio Matos in an external registry of the Complejo Asistencial Universitario de Salamanca. This model provides a continuous probability of risk. However, the predictions are basically the same. A probability of serious complications above 12-13% in the official calculator is equivalent to the simplified high risk criterion on MDCalc, and decision-making should be equivalent.

Any other research in the pipeline you're particularly excited about?

At ESMO 2017, we learned that a British group is trying to evaluate an algorithm based on pragmatic eligibility criteria, which relies on the CISNE Score only as the clinical aid for the decision. We agree with this criterion, and believe that like any other tool, the degree of certainty will depend on other independent groups progressively contributing their experience in this regard. In any case, all experiences are interesting if you can learn from them, and they will be welcomed.

Alberto Carmona-Bayonas, MD, PhD, is a medical oncologist in the department of hematology and medical oncology at Hospital General Universitario Morales Meseguer in Murcia, Spain. Dr. Carmona-Bayonas’ research interests include gastrointestinal cancers and febrile neutropenia.

Paula Jiménez Fonseca, MD, PhD, is a medical oncologist at the Hospital Universitario Central de Asturias in Oviedo, Spain. She has authored or co-authored dozens of studies in solid tumor oncology, particularly in the area of febrile neutropenia. Dr. Jiménez Fonseca is a member of the Spanish Society of Medical Oncology (SEOM).