Clinical utility of blood cultures drawn from indwelling central venous catheters in hospitalized patients with cancer

Background: Because of concern about low specificity, the American College of Physicians guidelines and expert opinion discourage the use of a central venous catheter when obtaining blood for culture for bacteremia or fungemia. However, data on the reliability of cultures done with blood obtained from a central venous catheter are conflicting. Objective: To determine the sensitivity, specificity, and positive and negative predictive values of cultures done with blood obtained through a central venous catheter compared with peripheral venipuncture. Design: Retrospective cohort study of hospitalized patients with cancer in whom samples for paired cultures were drawn through a central venous catheter and peripheral venipuncture. Setting: Tertiary care, university-affiliate medical center. Patients: 185 patients hospitalized on a hematology-oncology ward between August 1994 and June 1996. Measurements: Blinded assessments of culture results done by infectious disease experts were used as the gold standard. Sensitivity, specificity, and positive and negative predictive values were compared for culture of blood from central catheters and culture of blood from peripheral venipuncture. Results: Of 551 paired cultures, 469 (85%) were catheter-negative/venipuncture-negative. 32 (6%) were catheter-positive/veni puncture-positive, 17 (3%) were catheter-negative/venipuncture-positive. and 33 (6%) were catheter-positive/venipuncture-negative pairs. For the 82 paired cultures with at least one positive result, blinded determination of true bacteremia or fungemia was made by two infectious disease specialists. For catheter draw compared with peripheral venipuncture, sensitivity was 89% (95% CI, 79% to 98%) and 78% (CI, 65% to 90%) (difference, 11 percentage points [CI, -6 to 28 percentage points]), specificity was 95% (CI, 93% to 97%) and 97% (CI, 96% to 99%) (difference, -2 percentage points [CI, -5 to 0.2 percentage points]), positive predictive value was 63% (CI, 50% to 75%) and 73% (CI, 60% to 86%) (difference, -10 percentage points [CI, -26 to 5 percentage points]), and negative predictive value was 99% [CI, 97% to 100%]) and 98% (CI, 96% to 100%) (difference, 11 percentage point [CI, -0.5 to 3 percentage points]). Conclusions: In hospitalized hematology-oncology patients, culture of blood drawn through either the central catheter or peripheral vein shows excellent negative predictive value. Culture of blood drawn through an indwelling central venous catheter has low positive predictive value, apparently less than from a peripheral venipuncture. Therefore, a positive result from a catheter needs clinical interpretation and may require confirmation. However, the use of a catheter to obtain blood for culture may be an acceptable method for ruling out bloodstream infections.