You are the attending in a busy liver clinic. The gastroenterology fellow has just seen a 25-year-old male with chronic hepatitis C (HCV). The patient has liver enzymes that are two to three times the normal level, an elevated HCV RNA, and an HCV genotype of 1b. The fellow expresses a desire to treat the patient aggressively with interferon-based therapy because "he has a worse prognosis than patients with other HCV genotypes." When you ask him to justify his comment, he produces a recent article by Roffi and colleagues in which the influence of HCV genotype on the clinical outcome of liver disease was assessed in 2,307 patients with chronic HCV (Roffi et al, 2001).
Prognosis studies should seek to explain possible outcomes of a disease and define the probability with which they occur. Knowing this information is important for clinicians because it will affect their treatment thresholds and also aid them in discussing options with patients. Studies about prognosis are almost always cohort studies in which patients with similar risk factors for the outcome of interest are identified and followed over time to identify which patients develop the outcome, or case-control studies in which differences are investigated between similar groups with one of the groups consisting of subjects with the outcome of interest. Randomized controlled trials also may provide important prognostic information for each treatment arm.
The first consideration in considering an article about prognosis is determining whether or not the sample of patients included is representative of all patients to whom the results may apply. Referral bias is an example of where this concept may be important. Consider a study about the prognosis of irritable bowel syndrome (IBS). Patients with
IBS seen at tertiary care facilities may represent the most refractory and symptomatic patients among the entire spectrum of IBS patients. As such, these patients may have a worse "prognosis" for continued IBS symptoms compared with the majority of patients with IBS who never seek medical attention for their symptoms. Individual patients in a study about prognosis should be reasonably homogeneous in terms of the "stage" of their disease and risk of the outcome of interest. Additionally, other potentially confounding factors (eg, age, gender, or habits) should be analyzed and, if found, adjusted for by the investigators. Follow-up should be sufficiently long and complete to minimize the chances of missing the development of any cases of the outcome of interest. If the outcome of more than 10 to 20% of a study population is unknown, the validity of the results may be compromised. Lastly, the outcome of interest should be as objective as possible, and all subjects should be examined for this outcome equally. If more subjective outcome criteria are used, individuals measuring outcomes should be blinded to the outcome and hypotheses of the study.
Upon reviewing the abstract and methods sections of the study by Roffi and colleagues (2001), you determine that it meets the validity criteria for a study about prognosis (see Tables 1-6 and 1-7). The sample of patients in this trial was representative of patients with chronic HCV who were sufficiently homogeneous with respect to their risk of objective outcomes, such as cirrhosis, hepatocellular cancer (HCC), and death. Investigators took explicit steps to account for potential confounds, such as age, alcohol and IV drug use, methods of HCV transmission, prevalent cirrhosis, and duration of infection in their analysis. Follow-up was sufficiently long and complete at 1,482 cumulative years, so that important outcomes were unlikely to be missed. The outcomes that were measured were done so in an objective and unbiased manner equally in all patients.
Was this article helpful?