Nomograms are statistical formulas that use variable inputs (i.e., information about a patient) to develop predictions about outcomes and effects of treatment. They are useful as evidence-based tools to guide physicians and their patients in making informed treatment decisions. Typically a nomogram is developed from a data source, such as a population study or clinical trial, using a regression model based on the values of risk factors and patient outcomes. The nomogram is then tested and its validity confirmed by applying it to a separate, independent data source to demonstrate its general applicability.
Most nomograms in oncology are developed to predict standard outcomes such as overall survival, progression-free survival, or cancer recurrence. However, for certain patients at high risk for competing non-cancer mortality events (so-called competing risk populations), such models do not work well for selecting patients for given treatments, because they ignore the impact of competing events. That is, patients with an identical survival prognosis may have a very different competing risk prognosis, and thus a different expected benefit from therapy. This is a particular problem for older patients with cancer, patients with significant underlying cardiopulmonary disease (e.g., lung or head/neck cancer), and patients with low or moderate risk of disease progression relative to other health problems (e.g., prostate or uterine cancer). Comogram models are designed to account directly for the impact of competing events on patients' prognosis and likelihood to benefit from treatment, by modeling the effects of risk factors on the ratio of cancer to non-cancer events (AKA Omega Ratio - see below).
An "Omega Ratio" refers to a patient's proportional overall event risk that is attributable to cancer. It is a number on the scale from 0-1 (0-100%). Patients with very high values (close to 1) have a high risk of cancer progression relative to mortality from other causes, indicating a greater likelihood to benefit from intensive cancer therapy. Patients with a very low value (close to 0) have a higher risk of mortality from other causes relative to cancer progression, indicating a lower likelihood to benefit from intensive cancer therapy; such patients may benefit from alternative strategies directed toward their competing health risks.
This online tool is designed to provide a simple interface for patients and providers to guide treatment decisions and discussions about prognosis. For a given set of inputs, the model will predict a patient's Omega Ratio (see above). In some studies, higher Omega Ratios have been linked to a greater benefit of intensive treatment. The tool can be especially useful for patients with relatively indolent cancers but no or few competing health problems, and for patients with relatively aggressive cancers but many competing health problems or advanced age, where the benefit of intensive cancer therapy is unclear.
In the process of developing nomograms, researchers are required to access several different data sources to test their algorithms in a particular context. Unfortunately, not all data sources have collected precisely the same variables, in precisely the same manner, for precisely the same type of patients. Thus, even within a given patient population, there may be multiple nomograms, each giving slightly different predictions. In such cases, we have developed a composite (ensemble) nomogram taking all of the inputs together to give one final score, and note when the different algorithms agree or disagree, particularly where a treatment recommendation is considered.
Caveats
Nomograms are only as good as the data and models used to generate them. All statistical models make simplifying assumptions that may or may not hold for a given patient or cohort. Omega Ratios are continuous values on the scale from 0-1, but treatment decisions are binary; thus, specific threshold values for the Omega Ratio that define when a patient is or is not expected to benefit from intensive therapy are scientifically validated, but should be used as a guide and not a substitute for clinical judgment. Caution should be exercised when interpreting values near a treatment decision threshold. Thresholds are specific to the particular group under study. What constitutes "intensive" treatment is based on the context of the disease and specific treatment options, as well as the studies used to compare treatments.
Comorbidity refers to the existence of other health problems unrelated to cancer. For example, a patient with esophageal cancer may also have underlying heart and lung disease from previous smoking, which may affect their tolerance to treatment and expected benefit from intensive cancer therapy. Different indexes have been developed to measure the extent of comorbidity in a given patient. In some cases these indexes will differ based on which health conditions constitute a comorbidity, its severity, and the methods used to assess the presence or absence of a particular condition.
Competing risks refer to patients' simultaneous risk for other adverse health outcomes unrelated to cancer, in addition to the risk for cancer recurrence. Such outcomes might include severe side effects from therapy, or mortality from non-cancer causes, or from a newly developed different (i.e., secondary) cancer.