PEDIATRICS Vol. 121 No. 4 April 2008, pp. 866-867 (doi:10.1542/peds.2007-3716)
LETTER TO THE EDITOR |
Bilirubin-binding Capacity in Low Birthweight Premature Infants
Antony F. McDonagh, PhDDivision of Gastroenterology
University of California
San Francisco, CA 94143
To the Editor.—
Bender et al1 have collected a wealth of interesting data on the bilirubin-binding capacity of blood in premature infants. However, clarification of several issues related to the data set, as outlined below, would be helpful in understanding the stated conclusions.
1. Figure 1 of the article depicts the relationship between unbound bilirubin (UB) and total bilirubin (TSB) in 3 risk groups of premature infants: low (L), moderate (M) and high (H). This is a key figure underlying a principal finding of the article: that UB is higher in unstable than in stable neonates. Figure 1 is presented as representing the data listed in the accompanying Table 3 and refers to bilirubin measurements performed at 5 days of age and risk assessments at 24 hours of age. Table 3 shows that there were 31 (29%), 38 (36%), and 36 (34%) infants in the L, M, and H groups, respectively, for a total of 105 samples. However, examination of an enlarged PowerPoint slide of Fig 1 shows that, in disagreement with Table 3, there are many >105 data points on the plot. I counted at least 135, apparently distributed as follows: 30% in L, 33% in M, and 37% in H. The same number (135) is the theoretical sample size the authors calculated as necessary for reliable statistical separation between groups.
In contrast to the data presented in Table 3 of the article, the data in "Methods" indicate that the number of samples with UB and TSB measurements for which risk assessment was reliable was 142, and that 105 (the number in Table 3) was the lower number of risk-classed samples for which binding capacity measurements were also available. On that basis, the number of data points to be expected in Fig 1 would be 142. Of course, if some points in Fig 1 represent identical data from 1 or more infants, the number of visible points on the graph would be <142.
These apparent inconsistencies in the relationship between the data in Table 3 and Fig 1 raise a number of questions.
- Exactly how many data points are plotted in Fig 1, and how many correspond to each of the risk categories?
- Why is the number of data points in Fig 1 much greater than the number of patients given in Table 3?
- Are the mean values for UB and TSB in Table 3 based on the 105 patients in that table or on the larger number of patients plotted in Fig 1?
- Why is the number of data points in Fig 1 apparently less than the number of patient samples (142) indicated in "Methods"?
2. Each of the 3 regression lines in Fig 1 passes through the origin when extended linearly. Although this is theoretically what should be observed, it seems unlikely experimentally, and suggests that the 0.0 point may have been included incorrectly, when the regression lines were determined. In addition, the r2 values for each group listed in Fig 1 are different from the values reported in the text of the article (although only slightly). More puzzlingly, when the digitized points on an enlargement of Fig 1 were replotted, an identical array of points was observed but with regression lines that do not pass exactly through the origin when extended and an r2 value for the high-risk group that is considerably smaller (0.271, higher variance) than the value (0.463) given in Fig 1.
If correct, these observations raise the following questions:
- Are the regression lines and their axis intercepts correct as depicted if the origin is not included in their calculation?
- Is the r2 value shown for the high-risk group correct?
3. The risk classifications used in Fig 1 of the article were based on 24-hour assessments. A comparable graph for the 48-hour risk assignments was not shown because "Results were the same whether the first 24-hour data or average 48-hour data were used." It is not clear what is meant by "average 48-hour" data. If that refers to all of the UB/TSB data points, then the graphs for 24-hour and 48-hour data sets would not only be comparable but necessarily identical because the same single set of UB/TSB measurements, measured at 5 days, would be used in both plots. What is unfortunately missing from the article is a plot similar to Fig 1 for the 48-hour risk assignments. Important questions are:
- What were the r2 values, slopes, and intercepts when the 48-hour data were plotted?
- How did these compare to the data shown in Fig 1?
- What was the distribution of infants across the 3 risk groups at 48 hours compared with that at 24 hours?
These important details, not included in the article, would be of interest and highly relevant to the overall conclusions.
4. Because UB concentrations can change rapidly, and the risk assessment of patients is also time dependent, it is unclear why UB/TSB measurements on blood drawn at 5 days should be relevant to risk assessments performed at 24 and 48 hours.
5. The authors found a roughly linear correlation between UB and TSB. That is not surprising. However, the low r2 values that they reported, reflecting high variance, are not encouraging with respect to the practical value of measuring UB values and TSB in such patients. The relative value of TSB:albumin molar ratios, which are often readily available and more easily measured than UB, was not considered. Interestingly, the data in Table 3 of the article show a gradual increase in that ratio across the 3 groups (L: 0.37; M: 0.42; H: 0.44), although it is unclear that the increase was statistically significant.
REFERENCE
- Bender GJ, Cashore WJ, Oh W. Ontogeny of bilirubin-binding capacity and the effect of clinical status in premature infants born at less than 1300 grams.
Pediatrics. 2007;120
(5):1067
–1073
[Abstract/Free Full Text]
PEDIATRICS (ISSN 1098-4275). ©2008 by the American Academy of Pediatrics
CiteULike 
Connotea 
Del.icio.us 
Digg 
Facebook 
Reddit 
Technorati 
Twitter What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
