MP541
MULPA: A MULTICOMPONENT INDEX FOR A QUICK DIAGNOSIS OF PERITONITIS IN PERITONEAL DIALYSIS PATIENTS

INTRODUCTION AND AIMS: Peritoneal infection (PI) is very frecuent and negatively impacts on survival of peritoneal dialysis technique. Currently, the gold standard for the diagnosis of "typical" bacterial peritonitis is based on cloudy peritoneal effluent, leukocyte cell count greater than 100 leukocytes/microliter (L/μl) and more than 50% polymorphonuclear cells (PMN). To reduce complications derived from peritonitis, the rapid onset of treatment is recommended. This study develops the construction of a multi-component model (MUL+PA) for the diagnosis of PI, to be useful in clinical practice.

METHODS: For our purpose, a cohort was used for the construction of the model, and later it was validated with another cohort, independent of the first one. The model construction cohort was generated by collecting information on effluent samples for the first 6 months of 2015. Using Multistix ® 10 SG Siemens test strips for leukocyte detection (value 0 = 0-15 L / μl, value 1 = 16-70, value 2 = 71-125 L / μl and value 3 = 126-500 L / μl). Subsequently, each sample was evaluated according to the gold standard: number of leukocytes, percentage of PMN, and a microbiological culture. Other variables collected were: age, sex, number of previous peritonitis, cloudy fluid (no/yes/doubtful), self-reported abdominal pain (yes/no) and diabetes (yes/no). The construction of the MUL+PA model, was based on the high association between the Multistix strips and the PI. To the values of the modified chromatic scale (MULTISTIX [0-1-2-3]) it was added one point if the patient reported pain (PAin +1). So, MUL+PA took values from 0 to 4. During the second 6 months of 2015, a new cohort was created, similarly and with identical variables. The MUL+PA model was applied to each sample of peritoneal effluent, and leukocyte cell count and percentage of PMN were determined.

RESULTS: The model construction cohort included 134 samples, 34 of them had infection (25.4% [17.6-33.1]). Mean age 64.6±13.0 years, 66% men. Positive samples (peritonitis) showed significantly more pain, more cloudy effluent and more intense color in the Multistix strips. In the construction cohort, samples with a MUL+PA value greater than 1, presented a sensitivity and specificity of 100%. The validation cohort included 100 samples with 16 infections (16% [8.3-23.7]). Both cohorts were similar, with a lower presence of diabetes in the validation cohort. In the validation cohort assuming as positive a sample with a value of MUL+PA greater than 1, a sensitivity of 100% and a specificity of 95.2% were obtained. Area under the ROC curve in the construction and validation cohort were respectively 1 and 0.997 (0.991-1).

CONCLUSIONS: The multi-component model MUL+PA applied in the construction cohort, showed a perfect separation of the positive and negative populations, typical of a gold standard test. All the positive patients presented a score equal to or greater than two, while all the negative patients had a score equal to or less than one. In the validation cohort, the MUL+PA model presented a sensitivity of 100% and a specificity of 95.2%. This is an acceptable error, taking into account that in daily clinical practice "the perfect diagnostic test does not exist".