Correspondence

Intravenous Immune Globulin to Reduce Nosocomial Infections

N Engl J Med 1994; 331:678September 8, 1994

Article

To the Editor:

The recent article by Fanaroff et al. on the use of intravenous immune globulin to reduce nosocomial infections in very-low-birth-weight infants (April 21 issue)1 is the largest published study of the use of this intervention in this population. The study was conducted in two phases; phase 1 was blinded, and phase 2 was not. The authors report on outcomes such as the incidence of confirmed nosocomial infection and septicemia in the treated and control groups in each phase and in both phases combined. They also report on the statistical analysis of the two phases combined.

Because the study design changed between the two phases of the study, we performed separate analyses of each phase. In phase 1 there were statistically significant reductions in the immune globulin group for both confirmed nosocomial infection (relative risk, 0.75; 95 percent confidence interval, 0.58 to 0.98) and septicemia (relative risk, 0.71; 95 percent confidence interval, 0.53 to 0.94). In phase 2 the rates of nosocomial infection and septicemia increased in both treated and control groups as compared with the rates in phase 1. However, in phase 2 there was no statistically significant difference in either outcome between the groups; nor, as reported by Fanaroff et al., was there a reduction in these outcomes with treatment in both phases combined. We found the difference in results between the two phases of the study (which appeared to have identical methods except for blinding) interesting. We wonder whether the authors think that the lack of blinding in phase 2 resulted in any bias.

Janet B. Lacy, M.Sc.
Arne Ohlsson, M.Sc., M.D.
University of Toronto, Toronto, ON M5S 1B2, Canada

1 References

1. 1

   Fanaroff AA, Korones SB, Wright LL, et al. A controlled trial of intravenous immune globulin to reduce nosocomial infections in very-low-birth-weight infants. N Engl J Med 1994;330:1107-1113
   Full Text | Web of Science | Medline

Author/Editor Response

Dr. Fanaroff replies:

To the Editor: As part of its monitoring plan, our study of the use of intravenous immune globulin to prevent nosocomial infections was statistically designed for multiple looks. Hence, although Lacy and Ohlsson suggest that the nominal 95 percent confidence interval for the primary outcome in phase 1 (intravenous immune globulin vs. placebo) was statistically significant, they have not adjusted for the multiple looks and have therefore underestimated the width of the confidence interval. Indeed, it was this favorable but not significant trend that persuaded the Data Safety and Monitoring Committee to recommend continuing the trial after it was halted midway because of concern about necrotizing enterocolitis.

Lacy and Ohlsson raise the possibility of bias during the latter half of the trial. This would be more plausible if the trial had concluded that infusions of immune globulin prevented nosocomial infections. However, during phase 2 there was an increased rate of nosocomial infections at all centers and in both weight categories. A test of homogeneity revealed no significant difference in the estimated odds ratios. We therefore stand by the conclusion that it is the content of the various batches of immune globulin, rather than bias, that accounts for the lack of effectiveness of intravenous immune globulin.

Avroy A. Fanaroff, M.D.
Rainbow Babies and Children's Hospital, Cleveland, OH 44106

for the National Institute of Child Health, Human Development Neonatal Research Network