• meningococcal septicaemia
• therapeutic filtration
• septic shock

From 1995 to 1999 we pursued a policy of early therapeutic filtration for purpura fulminans with shock (presumed meningococcal septicaemia).

Patients with deteriorating blood pressure, accelerating tachycardia, or worsening acidosis despite aggressive resuscitation, received venovenous extracorporeal plasmafiltration, 1:1 exchanging 3–4 plasma volumes over four hours. After exchange, oliguric or anuric patients were supported with haemofiltration. Plasma exchange was repeated if clinical improvement was not observed or sustained.

In the three years prior to this policy, the crude mortality for meningococcal septicaemia on the paediatric intensive care unit (PICU) was 28% (n = 60). During the 3.5 year period of the filtration policy, 101 children with meningococcal disease were admitted to the PICU. Eighty five patients were in shock. Of these, 62 were not filtered, of whom five died, including four patients with maximal Glasgow meningococcal septicaemia prognostic scores (GMSPS) who died within four hours of admission (negating any opportunity to complete a plasma exchange). Twenty three patients were filtered (six deaths). The overall mortality for meningococcal septic shock was 12.9%.

Plasmafiltration was performed in 21 cases and haemofiltration alone in two. The mean time interval from the onset of symptoms to the start of filtration was 44.3 hours in the deaths whereas that in the survivors was 27.9 hours (difference between means 16.4 hours, standard error 7.07, 95% confidence interval (CI) 1.69 to 31.1) Table 1 gives details of the filtration. The worst morbidity in the survivors occurred in those with the longest time interval between the onset of symptoms and commencing filtration.

The mortality for meningococcal disease ranges from 5% to 11% whereas that for associated septicaemic shock is 24–30%. The historical mortality rate was comparable to those series.1 ,2 The mortality rate for meningococcal septicaemia may be falling; however, the UK Public Health Laboratory Service Communicable Disease Surveillance Centre 1992–1998 mortality rates for septicaemia (11–16%) are artificially low as a result of inclusion of patients without septicaemic shock but with positive blood cultures.

The GMSPS relating to the time of first contact between intensive care doctor and patient were worse for the shocked patients who were filtered. The difference between the mean scores was 4.7 points (95% CI 3.03 to 6.37) on a scale that ranges from 0 to 15. A GMSPS score of 12 or above predicted mortality with 100% sensitivity but gave an expected mortality rate of 25%. A maximal score of 15 however predicted death with a predictive power of 0.69 (negative predictive power 0.97, sensitivity 0.82, and specificity 0.94) irrespective of filtration. Experience of filtration in human meningococcal septic shock appears to mirror animal models with lower mortality than predicted in treated groups and reduced or no effect if plasma exchange is delayed.

The low mortality (44% of predicted) associated with an intention to treat with filtration could be a result of confounding factors. However, we suggest a causal relation. A randomised controlled trial involving 244 patients could detect an effect size as large as implied with an alpha of <0.05 and a beta of <0.05.3