Criteria abstracted from The Users' Guide to Medical Literature, from the Health Information Research Unit and Clinical Epidemiology and Biostatistics, McMaster University

Highlighted lines and questions below provide links to the pertinent description of criteria in The EBM User's Guide, now available at the Canadian Centres for Health Evidence

Article Reviewed:

Please visit the new Evidence Based Journal Club Reviews

Effect of the immunomodulating agent, pentoxifylline, in the treatment of sepsis in prematurely delivered infants: a placebo-controlled, double-blind trial.

Lauterbach R, Pawlik D, Kowalczyk D, et al.

Crit Care Med 1999; 27:807-14. [abstract]

Reviewed by Troy Dominguez MD, Children's Hospital of Philadelphia

Review posted November 2, 1999

---

I. What is being studied?:

The study objective:

To assess variability of plasma cytokine (TNF, IL-1, and IL-6) concentrations in the course of sepsis and evaluate the influence of pentoxifylline infusion on the synthesis of cytokines and clinical status (mortality and end-organ dysfunction).

The study design:

Prospective, randomized, double-blind, placebo controlled trial

The patients included:

All premature infants (gestational age, <36 wks) with suspected sepsis who were admitted to the neonatal intensive care unit of two university teaching hospitals in Poland between January 1, 1995, and July 30, 1996. Only neonates with "late onset sepsis" (diagnosed after the first week of life) were enrolled in the study.

Patients had to demonstrate physical and laboratory signs of infection or a rapid deterioration of respiratory and cardiovascular function by criteria described in detail in the method section. In addition, the patients also had to have a positive peripheral blood culture (2 positive cultures were required for S. epidermidis infections) after randomization to be included in the analysis.

The patients excluded:

Infants with major congenital malformations, intraventricular hemorrhage (Grade III or IV), or symptoms of a congenital infection were excluded.

The interventions compared:

Pentoxifylline 5mg/kg/hr for 6 hours daily for six days or equal volume saline placebo.

The outcomes evaluated:

The primary outcomes were differences in plasma cytokine levels during the six day course.

The secondary outcome was the patient's clinical status. Clinical status was determined by the presence of several conditions on any day throughout the six day course. These conditions included mortality, metabolic acidosis, disordered peripheral perfusion, oliguria/anuria, hepatic failure, necrotizing enterocolitis, disseminated intravascular coagulation, and hypotension.

II. Are the results of the study valid?

Primary questions:

1. Was the assignment of patients to treatments randomized?

Simple randomization of treatments was performed using computer-generated random numbers.

2. Were all patients who entered the trial properly accounted for and attributed at its conclusion?

Was followup complete?

Yes.

Were patients analyzed in the groups to which they were randomized?

This was not stated explicitly, but it appears as if the patients were analyzed to the groups they were randomized. However, 22% (22/100) of the patients were excluded from the final analysis since they had negative cultures. Since these patients were randomized, their subsequent outcomes should be included in the analysis to preserve the power of the randomization. Once these patients have been excluded, the distribution of known and unknown confounders that had been balanced by randomization may no longer be equal in the two groups. Because the authors have not analyzed their patients based upon an intent-to-treat, their results may be biased.

One could also question the concept that just because you don't recover the organism, the patient is not really septic. Assuming the study results are valid (internal validity), it is difficult for clinicians to predict which patients will culture positive and administer the treatment which raises concerns about the generalizability (external validity) of the study results. This is another reason to present the analysis of all randomized patients.

Secondary questions:

3. Were patients, health workers, and study personnel "blind" to treatment?

Identical bottles of the study drug and placebo were released to care givers and the code for the contents of the bottles was held by a pharmaceutical representative. Thus, medical care givers and patients were apparently blinded to the treatment allocation group. It is not explicitly stated that the study personnel or statisticians were blinded to the treatment groups.

4. Were the groups similar at the start of the trial?

The groups were fairly similar at randomization with regard to gestational age, birth weight, Apgar scores and cytokine levels. However, the placebo group had 2 more patients with symptoms of shock. Given the borderline statistical significance of some results and the small sample size, this uneven distribution of a potentially important confounding factor could bias the results.

5. Aside from the experimental intervention, were the groups treated equally?

Apparently. The authors state that the routine medical management of sepsis in both groups was comparable. Several aspects of this management were discussed in general terms and suggested comparable care between the two groups. However, the descriptions of different aspects of care were often vague such as "... obligatory mechanical ventilation ..." and " ... the criteria for weaning from the ventilator were comparable ... ".

III. What were the results?

1. How large was the treatment effect?

There was a significant decline in plasma TNF comparing day 1 and day 6 in the pentoxifylline group, but not in the placebo group. A significant decline in IL-6 concentration was found comparing day 1 and day 6 in both groups. However, mean IL-6 levels were significantly lower in the pentoxifylline group on day 6. The fluctuations in plasma IL-1 concentrations in both groups were similar and did not significantly change over time. In summary, the authors found lower levels of TNF and IL-6 in the treatment group (no effect on IL-1) in an attempt to link clinical benefit with biochemical alterations.

The mortality rate was 2.5% (1 of 40) and 15.8% (6 of 38) in the treatment and control groups respectively. This gave an absolute risk reduction (ARR) of 13.3% and relative risk (RR) of 0.158. Another way of interpreting the RR is that the risk of mortality was 84.2% lower in the treatment group (1-RR or the relative risk reduction; RRR). Therefore, 7.5 patients (1/ARR) would need to be treated in order to prevent one death. This represents a very large treatment effect.

The authors perform a subgroup analysis of clinical outcomes stratifying on the presence of septic shock at randomization to determine if there is an interaction between the treatment and severity of illness. They neglect to present the main effect of the treatment on clinical outcomes and chose only to present a subgroup analysis of clinical outcomes in those infants without septic shock (Group B). They state there were not enough patients in the strata of patients with septic shock (Group A) to perform an outcome analysis and only presented mortality data for these patients and did not report a significant difference in mortality rates.

The clinical outcomes (prevalence of organ dysfunction, hypotension) for Group B patients were reported as absolute numbers rather than risk ratios, and the numbers in these categories were small. There was a lower rate of necrotizing enterocolitis in the treatment group 5.9% (2/34) compared to controls 26.7% (8/30). ). This gives an ARR of 20.8% and a RR of 0.221 (95% CI 0.051-0.959). The p values for the other outcomes reported are wrong since there is no way for the comparisons to all have the same p=0.05 when the percentages are changing and the denominator is fixed. For instance, 8.8% (3/34) and 30% (9/30) of the patients exhibited renal failure in the treatment and control group respectively (ARR=21.2% and RR=0.294 (95%CI 0.087-0.987)). The same p value was given for the comparison of metabolic acidosis in the treatment and control groups (17.6% (6/34) vs. 43.3% (13/30), ARR=25.7%, RR= .407 (95% CI 0.177-0.937)).

2. How precise was the estimate of the treatment effect?

Interestingly, the sample size is fairly small so the 95% confidence interval for the mortality risk ratio is 0.020-1.26, indicating the point estimate above of a RR of 0.158 could be as low as 0.02 or the treated group could be 1.26 times as likely as the controls to die (with 95% confidence). However, the mortality rates were reported to be significantly different (p=0.046). Why is this? The authors report the p value for the 2 x 2 table and appropriately use Fisher's exact test since some cells have small expected values. However, they report the one-sided p value instead of the two-sided p value (p=0.054, not significant). The appropriate test is a two-sided test since the null hypothesis is usually that the two groups are the same.

The width of the confidence interval for the risk ratio above suggests an imprecise estimate. The upper limit of the confidence interval exceeding 1 is in keeping with the non-significant two-sided p value.

IV. Will the results help me in caring for my patients?

1. Can the results be applied to my patient care?

No. The study population is outside the scope of my practice and the practice of most pediatric intensivists, and there is no other literature supporting the use of the drug in pediatrics. However, the large treatment effect is promising and deserves further research.

2. Were all clinically important outcomes considered?

No. Potential side effects of pentoxifylline were not assessed.

3. Are the likely treatment benefits worth the potential harms and costs?

Only the benefits were given. Potential harms and costs cannot be adequately assessed.

Comments

Back to the EB Journal Club Index