We studied twelve parameters (physical appearance, mucin clot, fibrin clot, white cell count, differential count, red blood cell count, gram stain for bacteria, crystal microscopy, aerobic bacterial culture, anaerobic bacterial culture and ratio between synovial sugar and blood sugar) in over 300 samples of synovial fluid from patients with a variety of suspected pathologies (e.g. infection, inflammatory disease, infection adjacent to a joint, aseptic loosening of a prosthesis).

The diagnosis of infection was further established using clinical signs, radiological features, full blood count, C-reactive protein and iron profile. Many of the patients came to surgery. This of course created further opportunity to establish or rule out the diagnosis of infection with greater certainty. Nine of the features of synovial fluid were analysed statistically, including turbidity, diminished viscosity, mucin clot, fibrin clot, total white cell count, polymorphs greater than 60%, bacteria observed on direct microscopy, bacteria yielded by culture and concentration of synovial sugar less than 40% of the simultaneous blood sugar. The positive or negative features of infection were determined to be true or false in the light of the cumulative overall features of infection. The data so obtained was analysed to establish sensitivity, specificity, positive predictive value, negative predictive value and accuracy.

The mass of data so obtained cannot be meaningfully expressed in such a brief abstract. Important examples are when culturing synovial fluid there were 44% false negatives or no growth and 56% true positives. Looking at the ratio between synovial sugar and blood sugar we found that taking 40% as the critical value, this was 62% sensitive, the specificity was 89%, the accuracy was 73%, the positive predictive value was 89%, the negative predictive value was 62.4%. However we went further and separated those who were definitely infected or probably infected i.e. Groups 4 & 5 from those who were probably or definitely NOT infected according to the sum of clinical laboratory and radiological parameters.

When thus separated the predictive value of a positive result was 100% in Group 4 & 5 and 0% in Group 1 & 2. The predictive value of a negative result in Group 1 & 2 was 98.7% accurate and 22.4% in Group 4 & 5.