Impact of previous sepsis on the accuracy of procalcitonin for the early diagnosis of blood stream infection in critically ill patients

Background: Blood stream infections (BSI) are life-threatening infections in intensive care units (ICU), and prognosis is highly dependent on early detection. Procalcitonin levels have been shown to accurately and quickly distinguish between BSI and noninfectious inflammatory states in critically ill patients. It is, however, unknown to what extent a recent history of sepsis (namely, secondary sepsis) can affect diagnosis of BSI using PCT. Methods: review of the medical records of every patient with BSI in whom PCT dosage at the onset of sepsis was available between 1(st) September, 2006 and 31(st) July, 2007. Results: 179 episodes of either primary (n = 117) or secondary (n = 62) sepsis were included. Procalcitonin levels were found to be markedly lower in patients with secondary sepsis than in those without (6.4 [9.5] vs. 55.6 [99.0] ng/mL, respectively; p < 0.001), whereas the SOFA score was similar in the two groups. Although patients in the former group were more likely to have received steroids and effective antibiotic therapy prior to the BSI episode, and despite a higher proportion of candidemia in this group, a low PCT value was found to be independently associated with secondary sepsis (Odd Ratio = 0.33, 95% Confidence Interval: 0.16-0.70; p = 0.004). Additional patients with suspected but unconfirmed sepsis were used as controls (n = 23). Thus, diagnostic accuracy of PCT as assessed by the area under the receiver-operating characteristic curves (AUROCC) measurement was decreased in the patients with secondary sepsis compared to those without (AUROCC = 0.805, 95% CI: 0.699-0.879, vs. 0.934, 95% CI: 0.881-0.970, respectively; p < 0.050). Conclusion: In a critically ill patient with BSI, PCT elevation and diagnosis accuracy could be lower if sepsis is secondary than in those with a first episode of infection.