Incorporating Genetic Potential When Evaluating Stature In Children With Cystic Fibrosis

Abstract

Objective

The 2002 Cystic Fibrosis Foundation (CFF) practice guidelines recommend adjusting for genetic potential when evaluating height status in children with CF. However, there is paucity of data to support this recommendation. We compared three methods of classifying short stature: unadjusted height percentile < 10^th, Himes adjusted height percentile < 10^th, and unadjusted height below the CFF target height lower bound.

Patients and Methods

Data from 3306 children with parental heights documented in the 1986–2005 CFF Patient Registry were analyzed.

Results

Mean height percentile of CF children (33^rd) was lower than their parents’ (mothers’ 53^rd, fathers’ 57^th), and 80% of CF children were below the average of their parental height percentiles. In children with short parents, Himes adjusted height percentile was significantly higher than unadjusted height percentile (27^th vs. 8^th), whereas the opposite was found in children with tall parents (Himes adjusted at 18^th vs. unadjusted at 49^th). Consequently, the prevalence of short stature decreased from 52% to 22% in children with short parents and increased from 8% to 34% in children with tall parents after Himes adjustment. In children with discrepant classification on short stature before and after Himes adjustment, percent predicted forced expiratory volume in one second was negatively associated with unadjusted height percentile but positively associated with Himes adjusted height percentile. In children with short parents, the CFF method underestimated the prevalence of short stature (9%) compared to the Himes method (22%).

Conclusion

Without adjustment of genetic potential, the prevalence of short stature is underestimated and the association between height and lung function is biased.

INTRODUCTION

Height is both a hereditable trait and a feature of growth that is profoundly impacted by nutrition and disease. It is important that the genetic contribution to height be considered when evaluating the influence of nutrition and disease on attained height, especially for children with chronic diseases such as cystic fibrosis (CF). According to the 2004 Cystic Fibrosis Foundation (CFF) Patient Registry Annual Report, 15% of CF children had heights below the 5^th percentile without adjusting for their genetic potential (1). The prevalence of short stature in CF children is likely to be different if the contribution of genetic potential is accounted for. One possibility is that the prevalence of short stature in CF children may be overestimated, because parents of CF patients may be shorter than normal adults, as reported by a recent Italian study (2). Alternatively, the prevalence of short stature in CF children would be underestimated if their parents have normal/tall stature, as revealed in our previous analysis using data from a single CF center (3). In either case, it is important to separate the effects of genetic potential versus disease impact in order to provide optimal clinical care.

The genetic potential for height is commonly estimated from parental stature. However, methods for utilizing parental stature to adjust the child’s stature vary (4–10). The 2002 CFF consensus report (10) recommended a simple method to estimate genetic potential, namely, calculation of target height range based on parental heights, when evaluating height status in children with CF (7). The concept of this method is that a healthy and well-nourished child’s attained adult height shall reflect his/her genetic potential. The CFF target height method (7, 10) is intuitive to interpret and convenient to use, but has not been validated in the CF population. Clinical applications also reveal limited use because the CFF target height range is very large (e.g., for boys, 176 ± 10 cm spans the 7^th to the 93^rd percentile on the CDC growth chart, ref 11). This means that even if a child’s height is substantially below the target height, he/she most likely would remain above the lower bound of the target height range and thus be considered as meeting his/her genetic potential.

Another method, developed by Himes et al (8), also utilizes parental heights to adjust the child’s height. This method (8) is based on statistical modeling of age-specific relationships between mid-parental heights and children’s heights using data from the Fels Longitudinal Study (12). Nevertheless, Himes method (8) requires the use of large reference tables to calculate an “adjusted height”, making it impractical for use in routine clinical settings.

The objectives of this study are to utilize the 1986–2005 CFF Patient Registry to: 1) compare the difference between unadjusted height to Himes adjusted height percentiles (8) and their associations to lung function, and 2) examine the agreement between the CFF target height method (7, 10) and the Himes adjusted height method (8) in classifying short stature.

SUBJECTS AND METHODS

Study Population

The CFF Patient Registry documents the diagnosis and follow-up evaluations of patients with CF who are seen at accredited centers in the United States (13). Data from 3510 children older than 2 years of age who had self-reported parental heights available were identified from the 1986–2005 CFF Patient Registry. Of these, 204 patients with parental heights less than 100 cm (likely due to inch-centimeter conversion or recording errors) were excluded from analysis. The most recent height measurement between age 2 to 18.5 years for each patient was used for analysis. Sex- and age-specific percentiles and z-scores for height were calculated by using the reference values from 2000 CDC growth charts (11).

The CFF target height method

This method (7, 10) uses parental heights to predict the genetic potential for a child’s adult height, referred as the target height, which is calculated by mid-parental height plus 6.5 cm for boys or mid-parental height minus 6.5 cm for girls. A 10 cm-range above (upper bound) and below (lower bound) the target height for boys (9 cm for girls) is then applied to define the range of normal variation for target height. If the child’s height is below the lower bound of his/her target height, he/she is considered to be below genetic potential. The procedure to calculate the CFF target height and range (7, 10) is described in detail in the Appendix. In an example illustrated in Figure 1, the child’s CFF target height is 166.5 cm (7^th percentile), with a lower bound at 156.5 cm (0.2^th percentile). His unadjusted height at age 15 (5^th percentile) is above his target height lower bound (0.2^th percentile) and therefore he is meeting his genetic potential.

Figure 1.

An example illustrating unadjusted height, Himes adjusted height, and the CFF target height and lower bound for a 15 year-old boy with unadjusted height of 157 cm (= 5^th percentile) and mid-parental height of 160 cm.

The Himes adjusted height method

This method (8) does not directly predict the child’s genetic potential for height. Instead, it attempts to eliminate the influence of tall and short parental stature on the child’s stature by generating an “adjusted height”, which represents the child’s height as if his/her parents had average stature. Therefore, Himes adjusted height presumably reflects the impact of nutrition and disease on the child’s height. The procedure to calculate Himes adjusted height (8) is described in detail in the Appendix. In the example illustrated in Figure 1, Himes adjusted height percentile is 22^nd. If 10^th percentile were used to define short stature, this boy would be classified as “short” by unadjusted height percentile (5^th) but “normal” by Himes adjusted height percentile (22^nd).

Assessment of agreement between the CFF method and the Himes method

Direct comparisons between the CFF (7, 10) and the Himes (8) methods are not possible because the CFF method does not give an adjusted height. Since our purpose of utilizing a parental height adjustment method is to identify short stature, it is logical to compare the CFF (7, 10) and the Himes (8) methods based on their agreement in classifying short stature. When we compared the CFF lower bounds with the 2000 CDC growth charts reference values at age 20, we found that 10(9) cm lower bound corresponds to the 7^th percentile cutoff point, when target height at age 20 is at the population mean, i.e., 50^th percentile. Therefore, we applied two cutoffs, < 5^th and < 10^th, to Himes adjusted percentile (8) to define short stature and compared each of these two cutoffs to the CFF target height lower bound (7, 10).

Association of height percentile to lung function

The associations of unadjusted height and Himes adjusted height (8) to the lung function parameter, percent predicted forced expiratory volume in one second (%FEV₁), were evaluated. %FEV₁ was calculated according to the Wang equations (14). For this analysis, only patients older than 6 years of age and having %FEV₁ data were included.

Statistical analysis

Statistical analyses were performed by using SAS (version 9.13, SAS Institute, Inc, Cary, NC) and R (http://www.r-project.org). Group differences were assessed by t-tests for continuous variables and by chi-square tests for proportions. All analyses involving height percentiles were performed on the basis of Z-scores but reported as percentiles because the latter are displayed on the 2000 CDC growth chart (11) and more commonly used in routine clinical setting.

Agreement between any two of the three methods (unadjusted height, CFF target height range, Himes adjusted height) in classifying short stature was assessed by kappa statistic (15, 16). A kappa of < 0.4 is considered “poor” agreement, between 0.4 and 0.6 “fair” agreement, between 0.6 and 0.8 “moderate” agreement, and > 0.8 “good” agreement.

RESULTS

Height status of CF children and their parents

Table 1 shows that, in our study population, CF children were shorter than normal children (mean height percentile at 33^rd), but their mothers (164.1 cm, 53^rd percentile calculated based on the CDC 2000 growth charts at age 20 years) and fathers (178.4 cm, 57^th percentile) were slightly taller than the normal population (176.5 cm for adult men and 162.3 cm for adult women, ref 17). Consistent with these results, mean height percentile of CF children was below the average of their parental height percentiles (i.e., 55^th) by an average of 22 percentile points, and 80% of CF children were below the average of their parental height percentiles (Table 1). Two-thirds of children had “average” parents (mean of parental height percentiles between 26^th–74^th), 22% had “tall” parents (mean of parental height percentiles ≥ 75^th), and 11% had “short” parents (mean of parental height percentiles ≤ 25^th).

Table 1.

Height status of children with CF and their parents

Number of CF children	3306
Gender
Male	1649 (50%)
Female	1657 (50%)
CF children’s most recent height measurement
Percentile	33 ± 28
Z-score	−0.62 ± 1.08
Age at the most recent height measurement
Mean ± SD	10.6 ± 5.0
Mother’s height (Mean ± SD)
Centimeters	164.1 ± 7.2
Percentile (at age 20 on the 2000 CDC growth chart)	53 ± 30
Z-score	0.13 ± 1.12
Father’s height
Centimeters	178.4 ± 7.9
Percentile (at age 20 on the 2000 CDC growth chart)	57 ± 30
Z-score	0.22 ± 1.10
Average parental height percentile	55 ± 23
Average of parental height Z-scores	0.17 ± 0.85
Child’s height percentile minus average parental height percentile	−22 ± 27
Number (percentage) of children with height percentile below     their average parental height percentiles	2631 (80%)
Number (percentage) of children with:
Short parents (average parental height percentiles ≤ 25th)	388 (11%)
Average parents (average parental height percentiles 26th–74th)	2204 (67%)
Tall parents (average parental height percentiles ≥ 75th)	714 (22%)

^*Data from 1986–2005 CF Foundation Patient Registry

Comparison between unadjusted height and Himes adjusted height

As shown in Figure 2, overall mean unadjusted height percentile (27^th) was significantly higher than Himes adjusted height percentile (21^st), p < 0.0001. The difference between unadjusted and Himes adjusted height percentiles varied greatly by parental stature. Among children with average parents, small differences were found between unadjusted and Himes adjusted height percentiles. However, among children with short parents, Himes adjusted height percentiles were substantially higher than unadjusted height percentile (27^th vs. 8^th, p < 0.0001), while the opposite was found among children with tall parents (18^th vs. 49^th, p < 0.0001).

Figure 2.

Comparison of unadjusted height percentiles and Himes adjusted height percentiles in children with CF. Short parents: average parental height percentiles ≤ 25^th; average parents: average parental height percentiles between 26^th and 74^th; tall parents: average parental height percentiles ≥ 75^th.

Differences between unadjusted and Himes adjusted height percentiles led to different prevalence estimates of short stature. The prevalence of short stature determined by Himes adjusted height < 10^th percentile was 5% higher compared to that determined by unadjusted height < 10^th percentile (Table 2). However, this does not fully reflect the discrepancy between these two methods, because Himes method increased the prevalence of short stature among children with tall parents (from 8% to 34%) but decreased it among children with short parents (from 52% to 22%), Table 2. Consequently, 30% of children who had short parents and 26% of children who had tall parents had discrepant classification on normal vs. short stature comparing these two methods (Figure 3).

Table 2.

Comparison of the prevalence of short stature defined by different methods

	Overall	By parents’ stature*
		Short Parents	Average Parents	Tall Parents
No. of CF children	3306	388	2204	714
Prevalence of short stature defined by:
Unadjusted height percentile < 5th	16%†	35%†	16%†	4%†
Unadjusted height percentile < 10th	26%†	52%†	27%†	8%†
Himes adjusted height percentile < 5th	18%†	13%†	18%†	22%†
Himes adjusted height percentile < 10th	31%†	22%†	32%†	34%
CFF lower bound method	24%	9%	23%	39%

^*Short parents: average of mother’s and father’s height percentiles ≤ 25^th

Average parents: average of mother’s and father’s height percentiles between 26^th and 74^th

Tall parents: average of mother’s and father’s height ≥ 75^th

^†p < 0.05, pair-wise comparison to the CFF lower bound method

Figure 3.

Pair-wise comparisons of the discrepancy among four methods of determining the prevalence of short stature in CF children. Short parents: average parental height percentiles ≤ 25^th; average parents: average parental height percentiles between 26^th and 74^th; tall parents: average parental height percentiles ≥ 75^th.

Impact of discrepancy between unadjusted and Himes adjusted height on their associations to lung function

The subgroup of children with discrepant classification between unadjusted height < 10^th percentile and Himes adjusted height < 10^th percentile were examined. As shown in Figure 4, the association of height to %FEV₁ at age 15 (n = 217) changed from negative (regression coefficient −5.5% and correlation coefficient −0.11, panel A) when unadjusted height was used to positive (regression coefficient +7.3% and correlation coefficient 0.14, panel B) when Himes adjusted height was used, p = 0.04. This is because, in children who were classified as short by unadjusted method but normal by Himes method (primarily because their parents were short), unadjusted height percentiles were low while Himes adjusted height percentiles were high (red points in Figure 4). In contrast, in children who were classified as normal by unadjusted method but short by Himes method (primarily because their parents were tall), unadjusted height percentiles were high while Himes adjusted height percentiles were low (blue points in Figure 4). Similar results were observed at age 10 (n = 334), 17 (n = 164) and 18 (n = 128) years (data not shown). Taken together, these results indicate that unadjusted height percentiles masked the association between height and lung function in children with short and tall parents.

Figure 4.

Comparison between unadjusted height percentile (panel A) and Himes adjusted height percentile (panel B) on their associations to lung function parameter percent predicted forced expiratory volume in one second (%FEV₁) in CF children with discrepant classification on short stature at age 15 (n = 217).

Comparison of the CFF method to the Himes method in classifying short stature

The CFF method classified significantly fewer children with short stature than Himes adjusted height < 10^th percentile, but more children with short stature than Himes adjusted height < 5^th percentile (Table 2). More importantly, the differences in the prevalence of short stature between the Himes method and the CFF method were much greater among children with short or tall parents (Table 2). It should be noted that, even if we choose a cutoff for the Himes method that matches exactly to the CFF lower bound for average parents (i.e., 7^th percentile), these two methods still would not agree in children who have tall or short parents (data not shown). This is because the CFF lower bound values are fixed and chosen based on a confidence interval centered around the population mean; these lower bound values are too large for children with short parents. Consistent with these findings, the disagreement between the CFF method and Himes method was also largest among CF children who had short parents (Figure 3).

DISCUSSION

A recent study from an Italian population reported for the first time that young CF adults are shorter than their healthy peers because their parents are also short (2). CF children in our study population are also short, but their parental statures are similar to normal adults (Table 1). One possible explanation for this discrepant observation is the difference in sample size. In the Italian study (2), sample size of the CF patients/parents was not reported, thus it’s unclear whether that sample is representative of the entire Italian CF population. One the other hand, our study included a large sample size drawn from the CF Foundation Patient Registry of the US.

Two findings from our study support the importance of accounting for genetic potential when evaluating stature in children with CF. First, the overall prevalence of short stature in our study population, regardless of what cutoff point is used, is underestimated without adjustment for genetic potential. This is because the majority of CF children’s height percentiles are lower than those predicted based on their parental heights, most likely due to malnutrition associated with CF. Therefore, the overall distribution of CF children’s heights is likely to shift toward shorter stature, when compared to the distribution of the heights of their parents, who do not have CF. Consequently, CF children’s height distribution is likely to differ before and after adjusting for genetic potential/parental height, which supports the need for such adjustment. Most importantly, without adjusting for parental heights, the prevalence of short stature in children with short parents is overestimated while prevalence of short stature in children with tall parents is underestimated (Table 2).

Second, our study demonstrated that adjustment for genetic potential based on the Himes method (8) changed the direction the association between height and lung function. To our knowledge, this is a novel finding. Our analysis showed that, in CF children who had discrepant classification before and after Himes adjustment, unadjusted height percentile is negatively associated with %FEV₁, which contradicts biological expectation. Our discovery that Himes adjusted height percentile is positively associated with %FEV₁ uncovered the underlying explanation to this contradiction. It should be emphasized that, without adjustment of genetic potential, the positive association between height and lung function would not only vanish but a negative association would appear.

Several implications can be drawn from the above findings. At the individual patient’s level, unadjusted height masks the true impact of CF-associated malnutrition on linear growth. That is, unadjusted height percentile is likely to misclassify a CF child with tall parents to be “normal stature” and a CF child with short parents to be “short stature”. Therefore, treatment decisions based on unadjusted height may be misguided. A negative association between unadjusted height percentile and lung function may also mislead CF clinicians that improving height status would have little impact on lung function. This finding needs to be validated with other measures of pulmonary status such as chest radiography.

At the population level, the difference in prevalence estimates between using unadjusted and Himes adjusted height percentile for classifying short stature are large enough to be of concern, particularly considering that height percentile is incorporated into CFF’s nutrition outcome measure (3, 10, 18) that is used as an indicator to monitor treatment progress over time in the national CF population, as well as for comparison of center differences (18–20). In addition, quality improvement projects implemented by many CF centers often aim at reducing the rates of malnutrition that are defined not only by weight but also height status of CF children (18).

Based on the above evidence, it is essential to account for genetic potential when evaluating height status of CF children. However, we showed that current methods in use, namely, the CFF target height method (7, 10) and the Himes method (8), do not agree. In addition, both methods have disadvantages. The CFF method (7, 10) is easy to understand and use, but the target height lower bounds are invalid, particularly for children who have short parents. The Himes adjusted height method (8), although not a gold standard either, is more accurate because it is derived based on observed relationships between parent and child statures. Nevertheless, Himes method (8) is complicated and inconvenient to use in routine clinical settings. A new method, one that retains the simplicity of the CFF method (10) but agrees with the accuracy of the Himes method (8) is needed.

One weakness in our study is that, unlike the Italian study (2) in which parental height was measured, parental heights were self-reported. Our results show that CF children’s parents are slightly taller than the normal population. This observation may be reflective of the higher proportion of white and black races (98%) compared to the general US population (93% white and black races, ref 21) and/or due to self-reporting bias. According to data from US population survey, adults over-reported their heights by an average of 1.22 cm in men and 0.68 cm in women (17). When we subtracted these average over-reported values, mean mothers’ and fathers’ heights became almost the same as the normal population mean (52^nd and 51^st percentile for fathers and mothers, respectively). Nevertheless, the discrepancy between the CFF method (7, 10) and the Himes method (8) in classifying short stature observed in our study remains present after accounting for potential over-reporting of parental heights.

An area of uncertainty with regard to using mid-parental height to determine a child’s target height is that mid-parental height may not be a good predictor for children with discordant parents, i.e., tall father and short mother or vice versa, because the genetic contribution to a child’s stature may be unequal from each parent. Further research is needed to develop appropriate methods to adjust for genetic potential in children with discordant parents. In fact, the questions that need to be addressed first are what is the definition of discordant parents, and is it necessary to develop different methods for this population? Himes method (8) assumes equal genetic contribution from each parent, with no dominance effect, which agrees with other evidence (22). Pubertal growth is another area concern for using mid-parental height to adjust genetic potential. Evidence shows that the relationship between mid-parental height and child’s height changes during pubertal growth (22), most likely due to large individual variations in the timing of pubertal growth spurt. Methods for adjusting genetic potential, as with any method to evaluate nutritional status, should be used in conjunction with indicators of pubertal maturation during the adolescent period.

In conclusion, findings from the present study provide strong evidence to support CFF’s recommendation that genetic potential must be incorporated when evaluating height status in CF children with short or tall parents. In addition, interpretation of stature with or without adjustment of genetic potential requires consideration of biological maturation of the individual child, particularly during pubertal growth. Himes adjustment method (8) is valid, based on its methodology and its association with lung function. On the other hand, the CFF target height method (7, 10) is inappropriate, because its fixed lower bounds are flawed (i.e., too large for short target heights and too small for tall target heights). Further research is needed to develop new methods that agree with the accuracy of the Himes method and are convenient to use in routine clinical settings.

Abbreviations

CF

cystic fibrosis

CFF

Cystic Fibrosis Foundation

FEV₁

forced expiratory volume in one second

APPENDIX

Procedure to calculate CFF target height and range (7, 10)

1. Average two parental heights to obtain mid-parental height. Calculate the child’s target adult height by adding 6.5 cm to mid-parental height for a boy, or subtracting 6.5 cm for a girl. Apply ±10 cm for a boy or ±9 cm for a girl to define the target height range.

2. Plot target height and range at age 20 years on the 2000 CDC growth chart (11) and estimate their respective percentiles.

3. Extrapolate the percentiles of target height and range at age 20 to the child’s current age.

4. Plot the child’s height on the 2000 CDC growth chart (11); if his/her height percentile is below the target height lower bound, he/she is considered to be below genetic potential.

In the example shown in Figure 1 (15 y.o. boy, unadjusted height = 157 cm = 5^th percentile, mid-parental height = 160 cm), the child’s CFF target height is 166.5 cm (7^th percentile), with a lower bound at 156.5 cm (0.2^th percentile). His unadjusted height percentile at age 15 (5^th) is above the target height lower bound percentile (0.2^th) and therefore he is meeting his genetic potential.

Procedure to calculate Himes adjusted height (7)

1. Calculate mid-parental height.

2. Based on the child’s sex, age, height and mid-parent height, find the adjustment value from the reference tables (8).

3. Apply the adjustment value to the child’s height to obtain adjusted height.

4. Plot adjusted height on the 2000 CDC growth chart (11) to obtain adjusted height percentile.

In the example shown in Figure 1 (15 y.o. boy, unadjusted height = 157 cm = 5^th percentile, mid-parental height =160 cm), Himes adjustment value is 7 cm and Himes adjusted height is 164 cm (22^nd percentile). If 10^th percentile were used to define short stature, this boy would be classified as “short” based on unadjusted height percentile (5^th) but “normal” based on Himes adjusted height percentile (22^nd).