Likelihood level of evidence

The level or degree of plausibility of clinical evidence is a hierarchical system, based on evidence or research studies, that helps health professionals assess the strength or solidity of clinical evidence. tests associated with the results obtained from a therapeutic strategy, validity of a diagnostic test, risk factor for a disease, etc.

Since the late 1990s, any procedure performed in Medicine, whether preventive, diagnostic, therapeutic, prognostic or rehabilitative, has to be defined by its level of credibility of scientific evidence, a medical and scientific movement known as « fact-based medicine.

Likelihood levels and degrees of recommendation

According to the US Agency for Health Research and Quality:

Likelihood level

• Ia:The tests come from metaanalysis of controlled, randomized, well-designed trials.
• Ib: The evidence comes from at least a randomized controlled trial.
• IIa: The tests come from at least a well-designed controlled study without randomizing.
• IIb: The tests come from, at least, a not completely experimental, well-designed study, such as cohort studies. It refers to the situation in which the application of an intervention is beyond the control of the researchers, but whose effect can be evaluated.
• III: The evidence comes from well-designed descriptive studies, such as comparative studies, correlation studies or case and control studies.
• IV: The evidence comes from documents or opinions of expert committees or clinical experiences of prestigious authorities or case series studies.

Grade of recommendation

• A: The good truthfulness of scientific evidence suggests that the benefits of clinical service substantially exceed potential risks. Doctors should discuss the service with patients.
• B: Scientific verosilicity suggests that the benefits of clinical service are greater than potential risks. Doctors should discuss the service with patients.
• C: Scientific verosilicity suggests that there are benefits provided by clinical service, but the proportion between benefits and risks are too close to making general recommendations. Doctors should not offer the service unless there are individual considerations.
• D: Scientific verosimilitude suggests that the risks of clinical service are greater than potential benefits. Doctors should not routinely offer the service to asymptomatic patients.
• E: Scientific verosimilitude is poor, of poor quality, or in conflict, so that the risk proportion compared to the benefit cannot be assessed. Doctors should help patients understand the uncertainty surrounding clinical service.

Likelihood levels in oncology

For each cancer treatment, the results obtained can be classified in two different ways:

1. Solidity of study design.
2. Solidity of the final results.

The two rating scales together give an idea of the overall level of plausibility of the evidence.

1. Robustness of the clinical study design

Ordered from largest to smallest:

• 1. Random controlled clinical trial:

a. Double blind: Double-blind randomized controlled clinical trial 1.a is the gold pattern in the design of clinical studies. To achieve double-blind, the allocation of the study factor should not be disclosed to the doctor or patient before or after randomization. This design provides protection against bias in the assignment by the researcher and bias in the evaluation of results by both the researcher and the patient.

b. Not blind: In a clinical trial 1. b, the researcher and sometimes the patient know which group they belong (control or experimental). Unfortunately, most clinical trials in oncology cannot be double-blind after the allocation of treatment because the procedures or toxic effects often vary substantially between study assignments, so they are patent for both the health professional and the patient.

Meta-analyses of randomized studies offer the same level of strength of evidence as randomized studies, and not at a higher level.

• 2. Non- randomized controlled clinical trial: This category includes trials in which the allocation of treatments is performed without randomization but by another rule such as: date of birth, number of clinical history of the patient, day of visit to the office, availability of bed, or another strategy that would allow the researcher to know the assignment before obtaining the patient's consent. It is shown that a bias can occur in the allocation of treatments in such circumstances.
• 3. Case series: These clinical studies are the weakest form in the design of research studies, but they can be the only practical or available information that supports a therapeutic strategy, especially in the case of rare diseases or when the evolution of treatments is prior to the development of random study designs in medical practice. It can also be the only practical design when treatments in different branches of study are radically different such as amputation against surgery to save a member. However, these experiences do not have internal controls and should therefore be set on external experiences for comparisons. This always poses the problems of patient selection and the ability to compare with other populations. According to the possibility of extrapolation to other populations, case series are classified as:

a. Consequent population-based series.

b. Non-population consecutive cases.

c. Non-consecutive cases.

2. Solidity of the final results

• A. Total mortality or general survival from a point defined in time. This result is undoubtedly the most important for patients; it is also the one that is more easily defined and the one that is subject to a less bias by researchers.
• B. Mortality by specific cause from a point defined in time. Although specific mortality is very biologically important in interventions for certain diseases, it results more subjective than total and more risky mortality of researchers in their determination. Important effects of treatment that actually shorten overall survival can also be ignored.
• C. Careful evaluation of quality of life: This is a very important result for patients, so the careful description of this result within a solid design study is enough for most doctors to incorporate a treatment.
• D. Indirect results: All these studies are subject to the interpretation of the researchers and, more importantly, they do not translate into direct benefit for the patient, such as survival or quality of life. However, it is rational in many circumstances to prescribe a treatment that improves these indirect results while a more solid clinical trial is expected to support such a prescription. Indirect results are:

a. Survival free from disease.

b. Progressive-free survival.

c. Tumor response rate.

Since clinical studies are ranked by the strength of the research design and the importance of the results obtained, any study would have a double ranking. For example, 1.b.A for an unblinded randomized study showing a favorable overall survival outcome and 3.c.D.c for a phase II trial of selected patients using the response rate as the result. In addition, all recommendations must consider other characteristics that cannot be so easily quantified, such as:

1. Toxicity of the study factor.
2. Level of statistical confidence.
3. Observation time intervals.
4. Number of participants in the trial.
5. Quality assurance of testing.
6. Economic and human cost of rehearsal.