False positives Seven out of the 267 MTB culture-negative MK0683 specimens were initially hyplex® TBC PCR positive and considered as false-positives. Assessment of these samples by CTM PCR gave negative results with all seven samples. Five of these samples were also clearly negative on repeat with the hyplex® TBC PCR, while in two samples, the positive values of the first runs were confirmed on repeat.
One of these two specimens gave a positive culture for M. intracellulare, the other one showed no mycobacterial growth on HSP targets culture. Together, based on merged PCR data and culture results, two out of 267 MTB culture negative specimens (0.75%) were finally classified as false-positive hyplex® TBC PCR results (Table 3). Positive and negative predictive values Positive (PPV) and negative (NPV) predictive values largely depend on the prevalence of a disease. In particular, see more with low prevalence, the specificity of a test has to be very high, otherwise the PPV of the test will be poor. The proportion of TB samples (52%) included in this study was rather high and did not reflect the real situation of a TB diagnostic laboratory. In our laboratory, real time PCR (CTM PCR) yields between 7.0% and 9.5% positive results, depending on year and season. Assuming a mean rate of 8% of TB positive samples and a number of approximately
3000 PCR assays per year, the PPV of the hyplex® TBC test would be calculated to 90.4%, and the NPV to 98.5% Urease (Table 4), based on the sensitivity and specificity values found in this study (83.1% and 99.25%). Table 4 Predictive values at cut-off values 0.400 and 0.200 cut-off 0.400 cut-off 0.200 PCR pos b PCR neg b TOTAL a PCR pos b PCR neg b TOTAL a TB pos (n) 199 41 240 221 19 240 TB neg (n) 21 2739 2760 414 2346 2760 TOTAL (n) 220 2780 3000 635 2365 3000 PPVc (%) 90.4 34.8 NPVc (%) 98.5 99.1 a Based on the assumption
of a mean rate of 8.0% true TB positive specimens and a total number of 3000 specimens in a routine TB laboratory per year, the resulting numbers of TB positive and negative samples were calculated. b Based on the specificity and sensitivity values found in this study, the numbers of expected PCR positive and negative results among 3000 were calculated. Resulting numerical values were rounded. c Positive and negative predictive values were deduced from calculated PCR positive and negative results. Finally, the validity of the hyplex® TBC test was determined using an OD cut-off value for positive results of 0.200 as given in the instructions of the manufacturer. Using this value, the sensitivity of the test would rise to 92% while the specificity would decrease to 85% (data not shown). The PPV and NPV deduced from these sensitivity and specificity estimates would be calculated to 34.8% and 99.1%, respectively (Table 4). Thus, the PPV of the hyplex® TBC test is dramatically decreasing when the cut-off is changed to OD 0.200, meaning that out of 1000 PCR-positive results only 348 truly indicate TB.