Article Text

PDF

A 2 factor model helped to rule out early stage necrotising fasciitis
  1. Allison J McGeer, MD
  1. Mount Sinai Hospital Toronto, Ontario, Canada

    Statistics from Altmetric.com

    
 
 QUESTION: How accurate is a 2 factor model in differentiating early stage necrotising fasciitis (NF) from other non-necrotising soft tissue infections (NNFs)?

    Design

    The model was derived by using data from a previous case control study, and a retrospective cohort study was used to validate the model.

    Setting

    A university medical centre in Torrance, California, USA.

    Patients

    Data from 42 patients (mean age 39 y, 81% men, 50% with NF and 50% with NNF) were used for the derivation set. Data from 359 patients (mean age 44 y, 77% men, 9% with NF and 91% with NNF) admitted to hospital between April 1998 and March 1999 with a primary diagnosis of NF or NNF infection at discharge were used for the validation set. More patients with NF had a history of hepatitis (19% v 7%, p=0.03) and were intravenous (IV) drug users (71% v 30%, p<0.001).

    Description of prediction guide

    Derivation included comparison of vital signs on admission, findings on physical examination, laboratory measurements, and radiographic studies. The decision tree model predicted NF in the derivation set if at admission a patient had a white blood cell (WBC) count >15.4 × 109/l or a serum sodium concentration <135 mmol/l, or both. The serum sodium concentratia measurement was adjusted with the following mathematical formula for patients with hyperglycaemia (serum glucose concentration >200 mg/dl): corrected sodium level = measured sodium concentration +0.016 (measured glucose concentration –100).

    Main outcome measure

    Surgery confirmed necrotic fascia or muscle.

    Main results

    In the validation set, 39% of patients with NF had ≥1 characteristic clinical finding, such as necrotic skin, bullae, or gas on radiography. The model predicted NF for 28 of the 31 patients (90%) with NF and 80 of the 328 patients (24%) with NNF. Of patients with NF, 16 met the criteria by using the WBC count alone, 1 by using serum sodium concentrations alone, and 11 by using both. The table shows the sensitivities, specificities, and likelihood ratios.

    Presence of white blood cell count >15.4 × 109/l or serum sodium concentration <135 mmol/l, or both, to predict early stage necrotising fasciitis*

    Conclusion

    In patients with soft tissue infections, a model using 2 factors (high white blood cell count or low serum sodium levels) helped to rule out early stage necrotising fasciitis.

    Commentary

    In this study, Wall et al develop and validate a prediction guide for NF, a rare soft tissue infection for which early diagnosis is important because of the rapid need for surgery. Diagnosis of NF is difficult, and previous preoperative diagnostic evaluations have not stood the test of time.1 New diagnostic proposals should therefore be viewed with caution. The model proposed in the study by Wall et al is not adequate for stand alone use. Although the authors suggest some practical means to use the model to improve diagnosis, the physical examination variables they suggest considering (tense oedema, purplish discoloration, and neurological deficit) were not kept in the derivation model. The authors admit that their suggestions to improve diagnosis may not work as well as hoped.

    In contrast to the findings of Wall et al, 3 recent studies in other populations found no difference in WBC count at presentation between NF and NNF infections.24 Although none of these studies considered serum sodium concentration at presentation, the WBC distribution suggests that the test would not have done nearly as well in these populations. Notably, 70% of the patients with NF were IV drug users, who have a distinctive portal of entry and perhaps a unique host response. This model by Wall et al appears useful to rule out NF in the setting of IV drug use. Low serum sodium concentrations may be a useful diagnostic clue in other populations as well, but prospective validation of the proposed model in other settings is needed.

    References

    View Abstract

    Footnotes

    • Source of funding: not stated.

    • For correspondence: Dr C de Virgilio, Department of Surgery, Harbor-UCLA Medical Center, 1000 West Carson Street, Box 25, Torrance, CA 90509, USA. Fax + 1 310 782 1562.

    Request permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.