Use of GRADE for assessment of evidence about prognosis: rating confidence in estimates of event rates in broad categories of patients.

TLDR: This article covers studies answering questions about the prognosis of a typical patient from a broadly defined population and considers how to establish degree of confidence in estimates from such bodies of evidence.

Abstract: Introduction The term prognosis refers to the likelihood of future health outcomes in people with a given disease or health condition or with particular characteristics such as age, sex, or genetic profile. Patients and healthcare providers may be interested in prognosis for several reasons, so prognostic studies may have a variety of purposes,1–4 including establishing typical prognosis in a broad population, establishing the effect of patients’ characteristics on prognosis, and developing a prognostic model (often referred to as a clinical prediction rule) (Table 1). Considerations in determining the trustworthiness of estimates of prognosis arising from these types of studies differ. This article covers studies answering questions about the prognosis of a typical patient from a broadly defined population; we will consider prognostic studies assessing risk factors and clinical prediction guides in subsequent papers. Knowing the likely course of their disease may help patients to come to terms with, and plan for, the future. Knowledge of the risk of adverse outcomes or the likelihood of spontaneous resolution of symptoms is critical in predicting the likely effect of treatment and planning diagnostic investigations.5 If the probability of facing an adverse outcome is very low or the spontaneous remission of the disease is high (“good prognosis”), the possible absolute benefits of treatment will inevitably be low and serious adverse effects related to treatment or invasive diagnostic tests, even if rare, will loom large in any decision. If instead the probability of an adverse outcome is high (“bad prognosis”), the impact of new diagnostic information or of effective treatment may be large and patients may be ready to accept higher risks of diagnostic investigation and treatment related adverse effects. Inquiry into the credibility or trustworthiness of prognostic estimates has, to date, largely focused on individual studies of prognosis. Systematic reviews of the highest quality evidence including all the prognostic studies assessing a particular clinical situation are, however, gaining increasing attention, including the Cochrane Collaboration’s work (in progress) to define a template for reviews of prognostic studies (http://prognosismethods.cochrane.org/scope-ourwork). Trustworthy systematic reviews will not only ensure comprehensive collection, summarization, and critique of the primary studies but will also conduct optimal analyses. Matters that warrant consideration in such analyses include the method used to pool rates and whether analyses account for all the relevant covariates; the literature provides guidance on both questions.6 7 In this article, we consider how to establish degree of confidence in estimates from such bodies of evidence. The guidance in this article is directed primarily at researchers conducting systematic reviews of prognostic studies. It will also be useful to anyone interested in prognostic estimates and their associated confidence (including guideline developers) when evaluating a body of evidence (for example, a guideline panel using baseline risk estimates to estimate the absolute effect of Summary poIntS