Assessing an Article on Prognosis

Prognosis refers to the possible outcome of the disease and the frequency with which it is expected to occur (e. g., death, survival, etc.). Frequently, there are characteristics of the patient or the underlying condition

Figure 11 Importance of cutoff values on test performance. As the cutoff value is moved to the left, the true positive rate (sensitivity) increases, but specificity decreases.


That we use to predict the patients eventual outcome. Prognostic factors need not cause the outcome but may be associated with it strongly enough to predict their development. Ideally, the best study design to identify the presence of an increased risk associated with a prognostic factor is a cohort study. In an ideal cohort study, investigators follow a group of individuals who have not yet had the adverse event and monitor the number of outcome events over a long period of time.

In assessing an article about prognosis, it is important to determine if a well-defined sample of patients at a similar point in the course of their illness were included (Fig. 12). There are likely to be systematic differences in the results of population-based and speciality clinic-based studies on the clinical course and prognosis of a disease as referral may be influenced by disease severity. A hospital's reputation may result from its particular expertise in a specialized area of clinical care. ''Referral bias'' results from the referral of patients with a particular condition to a

Is the evidence about prognosis valid?

Was a defined representative sample of patients assembled at a common point in the course of their disease?

Was follow-up of study patients sufficiently long and complete? Were objective outcome criteria applied in a ’blind’ fashion?

If subgroups were identified

• Was there adjustment for important prognostic factors?

Is the valid evidence about prognosis important?

How likely are the events over time?

How precise are the prognostic estimates?

Will the results help me care for my patient?

Were the study patients similar to my own?

Will this evidence make a clinically impact on decision about what to offer or tell our patient?

Figure 12 Assessing an article on prognosis.

Clinician or tertiary center with acknowledged expertise in the area concerned. This may increase the likelihood of adverse or nonfavorable outcomes. Inception cohorts at tertiary care centers yield useful information to other clinicians who work in such settings, but it may not be possible to generalize the results to the wider population. To allow a true assessment of outcome patients should be at a similar well-defined point in the course of their disease. The follow-up period of the study should be sufficiently long to detect the outcome of interest (e. g., late recurrence of tumor). Under ideal circumstances, the authors will report on all patients entered into the study, but in practice this rarely occurs. Patients may fail to return for follow-up for a wide variety of reasons (death, ill health, relocation, etc.). The larger the number of patients whose fate is unknown, the greater the treat to the study's validity. In general, fewer than 5% loss probably leads to little bias and greater than 20% probably threatens the validity of the study. The lower the risk of a prognostic outcome, the greater the potential effect of patients who are lost to follow-up (17).

The examination for important prognostic outcomes should be carried out by clinicians who were ''blind'' to the other features of these patients. This avoids bias on two counts. Firstly, a clinician who knows the patient has a prognostic factor may carry out a more detailed search for the relevant condition (diagnostic-suspicion bias). Secondly, clinicians (pathologists, radiologists, and surgeons) may have their judgment influenced by prior knowledge of the case (expectation bias). Ideally in a published report on prognosis, the diagnostic suspicion bias will have been avoided by subjecting all patients to the same diagnostic studies. In surgical studies on prognosis, the occurrence of death is frequently a cited outcome. Judging the cause of death is very prone to error (especially when based on death certification) and assigning a cause to death may be subject to both diagnostic-suspicion and expectation biases.

Articles on prognosis in surgery frequently claim an altered prognosis for different subgroups. These groups' prognostic factors may be age, gender, disease grade, TNM stage, or comorbidity. It should be remembered that these prognostic factors need not cause the outcome, they may only be associated with its development strongly enough to predict it. When a ''new'' prognostic factor is identified, there is no guarantee that it hasn't resulted from a ''quirk'' in its distribution between patients with different prognoses. This initial group is called the ''training set''. In assessing articles on prognosis, there should be a statement (in the methods section of a paper) of a prestudy intention to examine this specific possible prognostic factor (the test set). If this second independent study confirms the prognostic finding in the test set, one can feel more confident about the validity of the evidence. One final point in studying prognosis: There are very few disorders for which medical interventions do not interfere with studying the natural history of the condition, and it should be clear from the publication what interventions the patients underwent.

  • Contact
  • Category: Men's diseases