SEOM has recently updated its clinical guidelines for febrile neutropenia (FN) [1]. However, as surprising as it may seem, the usefulness of clinical judgment as a guarantee and central axis of decision-making in FN is being questioned by some experts [2]. In light of this, the following ten clarifications highlight the importance of starting with clinical evaluation, as per SEOM and ASCO 2018 guidelines [1, 3], using scales solely as a complement when confronting doubts whether to intensify care:
1.
Most prognostic scales are designed to inform clinical judgment, but in no way supersede it. In the event of contradiction, the experienced clinician’s opinion should prevail [4]. As George Box said, “Essentially all models are wrong, but some are useful.”
2.
To base decision-making merely on summative models flies in the face of clinical guideline consensus [1, 3], and of the general principles governing evaluation of sepsis.
3.
The uncertainty of each FN episode is inversely proportional to its severity. In serious episodes, clinical presentation often speaks for itself and doesn’t require prognostic models that incur delay, precisely when treatment and intensive monitoring is of the utmost urgency. Paradoxically, it is precisely the low risk patients with FN (the most common) who, due to the risk of misclassification, become the most vulnerable if ambulatory management is chosen.
4.
There is currently no evidence that the MASCC score is able to stratify clinically stable patients according to their risk of complications. For example, a single-center, retrospective series of outpatients screened patients with FN eligible for ambulatory management based on exclusion criteria (e.g., clinical stability, normal organ function, no comorbidity, etc.) [5]. A high readmission rate (21%) was found, revealing how the inability of immunocompromised patients to generate inflammatory reactions diminishes the clinical expression of infection in the early stages. This justifies having models that support clinical data when said data alone are insufficient to detect seriousness. If, as suggested in the article, the MASCC score had been used instead of clinical evaluation, results would not have improved, since MASCC classified 98% of the outpatients as low risk and was, therefore, incapable of detecting those at risk for complications [5]. Curiously enough, the authors come to the opposite conclusion, after misinterpreting their own data, possibly because of wishful thinking.
5.
The FINITE prospective series of 1133 similar patients indicate that both MASCC and the Talcott model failed to predict 2 out of every 3 complications (Fig. 1a, b) [6]. Given the structure of the three models, it is expected that these results will be repeated in successive international validations [7, 8].
×
6.
Applying models to patient populations other than the ones from which they were derived is reckless, since the model’s critical variables may not be present in the entire population (e.g., all without hypotension). MASCC and Talcott include acute leukemia or hematopoietic transplant, which should not be mixed with solid tumors, given their different microbiological profiles and longer and deeper aplasia; in short, they require specific management.
7.
The MASCC model’s most serious mistake and, oddly, its strength (high positive predictive value), is that hypotension is both the most heavily weighted predictor, as well as the most common outcome; thereby, predicting itself [9, 10]. Patients with shock should be admitted without having to administer a model.
8.
Unlike MASCC, the Clinical Index of Stable Febrile Neutropenia (CISNE) is not only a score, but the sum of clinical judgment plus a score. It is applied to ascertain whether the apparent clinical stability is real, to prevent early discharge of a patient potentially at risk for complications (high risk CISNE classification) [11, 12].
9.
In contrast to MASCC/Talcott, CISNE does not seek to select patients for outpatient management, but to safeguard them from inappropriate discharge until favorable evolution is confirmed. Therefore, CISNE informs classification without increasing risk or adding complexities to the diagnostic process. Thus, according to the usual cutoffs, CISNE is more sensitive than other models to predict complications in seemingly stable patients (Fig. 1c). Since CISNE was conceived for clinically stable patients with solid tumors, its application in unstable patients with acute leukemia is reckless and methodologically incorrect, as some authors mistakenly suggest [13].
10.
Under the precautionary principle and despite apparent clinical stability, CISNE III patients (high risk, > 30% complications) should initially be treated as inpatients. Early discharge could provoke unnecessary risk to clinically stable patients who may, in fact, not be so stable.
Antibody–drug conjugates (ADCs) are novel agents that have shown promise across multiple tumor types. Explore the current landscape of ADCs in breast and lung cancer with our experts, and gain insights into the mechanism of action, key clinical trials data, existing challenges, and future directions.