FIG. 14:  This Figure shows the results from a study that demonstrated the impact of prevalence on positive and negative predictive value.^[8]  The data are based on results published initially in a paper by Grossman and colleagues who used a BladderChek® test to assess 1300 patients with hematuria.^[9] 

As seen in the Figure, overall about 8% of men and 3% of women in this cohort had bladder cancer, so given the disease prevalence, the positive predictive value was dramatically higher in men than in women. Yet the negative predictive value was impacted only slightly, because most patients in the study did not have cancer.

Looking more closely at the data, we see that because patients without hematuria rarely had cancer, the positive predictive value was zero; the marker never performed well in that population, because even when it was positive, no patient had cancer (the prevalence was zero). 

On the other hand, patients at much higher risk, such as smokers with gross hematuria, had very, very high positive predictive values – and almost every patient with a positive marker did, in fact, have cancer.  However, although the negative predictive value was impacted very significantly in men, it was not impacted so much in women (because the disease prevalence varied).  So even though the marker sensitivity and specificity was identical in all cases, how we interpret the positive and negative results really depends on whether or not the patient is likely to have cancer.