Correspondence

A Trial of RSV Immune Globulin in Infants and Young Children: The FDA's View

N Engl J Med 1994; 331:203-205July 21, 1994

Article

To the Editor:

We write regarding the article (Nov. 18 issue)1 reporting the results of a multicenter clinical trial that appeared to demonstrate the efficacy of immune globulin with a high titer of antibodies against respiratory syncytial virus (RSV immune globulin), given intravenously, in preventing RSV infection of the lower respiratory tract in high-risk children. The article was accompanied by a very positive editorial2. A scant two weeks later, the Blood Products Advisory Committee of the Food and Drug Administration (FDA) voted unanimously, on the basis of the same data, against recommending that this product be licensed. Although we cannot speak for the members of the advisory committee, the concern that developed at the FDA during the review of this application (which included the evaluation of protocol documents, raw data, correspondence between the FDA and the sponsor about the conduct of the ongoing trial, the results of site audits, and other materials) may interest readers of the Journal. For reasons of confidentiality, under the regulations our remarks are restricted to matters already presented for discussion at the public meeting.

The published report stated that the treatment assignment was made randomly, but it did not describe the method of randomization. In fact, the approach to randomization was developed and implemented separately at each site. Since treatment was not blinded, the randomization list showed clearly which treatment was assigned to each patient, and the person maintaining the list would have been aware of the upcoming treatment assignments. At the Denver site, where the most patients were enrolled and the most RSV infections occurred, this person was a study nurse who was involved in screening patients for the study and obtaining informed consent from parents. Although there was no specific evidence that the nurse's knowledge of treatment assignments influenced the randomization process, the method of randomization did not protect against such influence, conscious or unconscious.

The published report stated that all analyses followed the intention-to-treat principle. The material submitted to the FDA specifically stated that all children who underwent randomization and for whom written informed consent was obtained were followed for outcome assessment. In fact, this was not the case. Eight children who were identified as study participants were not followed, because of failure to obtain written informed consent or ineligibility discovered after entry into the study. Seven of these eight were assigned to an active-treatment group. Seventeen children were known to be eligible and to have had written consent provided, but they were not followed and therefore were not included in the report. All 17 children had been assigned to one of the two active-treatment groups. Although the numbers are small, these children appeared to differ from the children remaining in the study in a number of ways. They tended to be younger, had lower birth weights, were more likely to be female and nonwhite, had fewer prior infections, and in almost all cases had bronchopulmonary dysplasia or premature birth. Furthermore, the dropouts were predominantly from two of the five study sites. All these imbalances, particularly the imbalance in treatment assignment, suggest that the dropouts were not a random subsample of the originally randomized group, hence the assumption that they were prognostically similar to the other children may not be valid.

One way to protect against potential bias in this situation would be to perform an intention-to-treat analysis, assessing the outcomes of all randomized patients regardless of assigned treatment. We cannot perform such an analysis for this study, since we do not have reliable outcome data for the dropouts. Therefore, the possibility that the reported results are to some extent biased cannot be excluded.

To explore the potential effect of such bias, we performed a sensitivity analysis that considered only patients assigned to the group receiving a high dose of intravenous RSV immune globulin (nine dropouts) or to the control group (no dropouts). In this analysis, we examined the effect on the treatment comparison of the number of occurrences of infection of the lower respiratory tract among the nine dropouts from the high-dose group. If three or more of the children who dropped out had infections, statistical significance at the 0.05 level would be lost. (The investigators have told us that one of these dropouts is known to have had an RSV infection of the lower respiratory tract.)

The consistency of the treatment effect across sites was not examined in the published report, except for a brief acknowledgment that the controls at the Denver site had a higher incidence of RSV infection of the lower respiratory tract than the controls at the other sites. In fact, the overall results were strongly driven by the results in Denver. Although Denver had 28 percent of the overall patients, those patients accounted for 48 percent of the cases of RSV infection of the lower respiratory tract. The comparison of outcomes according to treatment group was statistically significant in the Denver site alone, with only a slight trend at best in the remaining sites -- the 95 percent confidence interval for the relative risk of RSV infection in the high-dose group as compared with the controls, at all sites other than Denver, ranged from 0.21 (a 79 percent reduction in incidence) to 1.64 (a 64 percent increase). Potential explanations for the disparate results in Denver were considered (higher altitude, more virulent strains of RSV, and regional variations in the attack rate), but they were all speculative and in any case did not diminish concern about the generalizability of the study results.

As the report made clear, six children in this study died, none from RSV infection. All six had been assigned to one of the two active-treatment groups. None of these deaths were clearly attributable to treatment, although in two cases treatment could have exacerbated the fluid overload that contributed to the deaths. Although the distribution of the deaths among the three study groups is not inconsistent with chance (P = 0.15), the trend is sufficient to demand attention. Conventional levels of statistical significance are not generally required for there to be concern about the potential for harm.

Because five of the six deaths occurred in children with congenital heart disease, the sponsors earlier began a second controlled study of this product in such children and requested licensure at this time only for children with bronchopulmonary dysplasia or prematurity. If there is truly a safety issue with this product, however, it is not clear that it would be limited to children with heart disease. The case with the strongest plausible connection between treatment and mortality was the single death in an infant with bronchopulmonary dysplasia and no congenital heart disease.

We believe that all these issues contributed to the advisory committee's determination that the efficacy and safety of intravenous RSV immune globulin in these high-risk populations have not been clearly established.

Susan S. Ellenberg, Ph.D.
Jay S. Epstein, M.D.
Joseph C. Fratantoni, M.D.
Dorothy Scott, M.D.
Kathryn C. Zoon, Ph.D.
Food and Drug Administration, Rockville, MD 20852

2 References

1. 1

   Groothuis JR, Simoes EAF, Levin JM, et al. Prophylactic administration of respiratory syncytial virus immune globulin to high-risk infants and young children. N Engl J Med 1993;329:1524-1530
   Full Text | Web of Science | Medline

2. 2

   McIntosh K. Respiratory syncytial virus -- successful immunoprophylaxis at last. N Engl J Med 1993;329:1572-1574
   Full Text | Web of Science | Medline

Author/Editor Response

The authors reply:

To the Editor: Although the advisory committee expressed concern about the method of randomization at Denver, a thorough audit revealed no evidence of bias in treatment assignment by the study nurse. Also, except for a family history of hay fever or asthma, we found no differences (P ≤ 0.15) with regard to 21 demographic variables and risk factors for RSV between treatment groups among the children randomized at Denver.

Our report analyzed all 249 children described by the investigators, whether or not the children received any treatment. While our article was in press, an FDA-requested review identified 25 dropouts who had been randomized but not included in the analysis. Written consent had been provided for 17 who were enrolled, 9 in the group receiving high-dose intravenous RSV immune globulin and 8 in the low-dose group; 4 in each group received some treatment. For all 25 children, we reviewed the hospital records or interviewed the parents or physicians, and we found that there had been two additional hospitalizations related to RSV, both in the high-dose group.

An intention-to-treat analysis of all 274 randomized patients yields a frequency of hospitalization with RSV of 8 of 92 (9 percent) in the high-dose group, 10 of 92 (11 percent) in the low-dose group, and 18 of 90 (20 percent) in the control group (P = 0.03 for high dose vs. control). The high-dose group also had a 59 percent reduction in the number of hospital days (P = 0.03) and a 97 percent reduction in days in the intensive care unit (P = 0.03). Thus, the efficacy of high-dose treatment remains significant after the inclusion of all the dropouts in the intention-to-treat analysis. An intention-to-treat analysis of all 257 randomized, eligible, enrolled children yields a 65 percent reduction in hospitalization with RSV in the high-dose group (P = 0.01).

Since our study had an overall power of 80 percent, it was not expected to demonstrate statistically significant efficacy either at any one study site or in the sites that remained after the exclusion of the site where the greatest efficacy was demonstrated. As expected, the frequency and severity of RSV disease varied among sites. As shown in Table 1Table 1Effect of High-Dose Intravenous RSV Immune Globulin, According to Study Site., the control patients in Washington and Boston did not have sufficiently severe RSV infections for reductions to be detectable either in the frequency of RSV infection of the lower respiratory tract (defined by a respiratory score ≥ 2) or in the severity of RSV infection. The three sites where such reductions were detected had reductions in RSV infections of the lower respiratory tract among patients receiving the high dose (Table 1); two sites also had reductions in the mean severity of infection. All four sites with hospitalizations for RSV had fewer hospitalizations in the high-dose group and fewer hospital days in that group. Thus, intravenous RSV immune globulin was found consistently to have efficacy at all sites where the frequency or severity of RSV disease allowed a treatment effect to be observed.

Space limitations in our article did not permit us to describe the extensive investigation of the deaths of study patients that we presented to the FDA. A review of case histories by two independent pediatric cardiologists yielded no temporal association or plausible link between the study infusion and death in five patients with cardiac disease. The treated and control groups with heart disease did not differ during the study in regard to changes in severity of disease, scores for congestive heart failure, or the requirement for open-heart surgery. In the second controlled study of 253 patients with congenital heart disease, referred to by Ellenberg et al., 3 children in the high-dose group and 8 children in the control group have died to date.

The infant with bronchopulmonary dysplasia mentioned by Ellenberg et al. had multiple episodes of respiratory failure, one of which was temporally associated with the study infusion and concomitant transfusions. This infant died of bronchopulmonary dysplasia, acute bronchopneumonia, and cor pulmonale three months after the infusion. In a subsequent open-label trial of high-dose intravenous RSV immune globulin in 68 infants with bronchopulmonary dysplasia or prematurity, no deaths have occurred to date.

We continue to believe that intravenous RSV immune globulin is safe and effective in preventing severe RSV disease in infants at high risk.

Jessie R. Groothuis, M.D.
Denver Children's Hospital, Denver, CO 80218

Val G. Hemming, M.D.
Uniformed Services University of Health Sciences, Bethesda, MD 20889

George R. Siber, M.D.
Massachusetts Public Health Biologic Laboratories, Boston, MA 02130

Franklin H. Top, Jr., M.D.
Medimmune, Inc., Gaithersburg, MD 20878

Citing Articles (4)

Citing Articles

1. 1

   James H. Ware. 2005. Statistical Review for Medical Journals. .
   CrossRef

2. 2

   Vance W. Berger. (2005) Quantifying the Magnitude of Baseline Covariate Imbalances Resulting from Selection Bias in Randomized Clinical Trials. Biometrical Journal 47:2, 119-127
   CrossRef

3. 3

   SUSAN M BURKE. (1996) Respigam for Protection Against RSV. The American Journal of Maternal/Child Nursing 21:5, 258
   CrossRef

4. 4

   O. Ruuskanen. (1995) Respiratory syncytial virus—is it preventable?. Journal of Hospital Infection 30, 494-497
   CrossRef