In children with cystic fibrosis (CF), maintenance of normal growth rate is associated with superior pulmonary outcomes (1,2). Similarly, maintenance of body mass index (BMI) in adults with CF is correlated with improved pulmonary function (3). Nutritional compromise impairs growth and weight maintenance, and thus nutritional support has become a central component of the care of patients with CF. Although the pathophysiology of nutritional compromise in patients with CF is complex, it probably results from an imbalance in energy losses, energy expenditure, and energy intake (4,5).
Studies have shown that gastrostomy tube (GT) placement improves BMI in patients with CF. To date, these studies have not shown a significant improvement in pulmonary function as a result of GT placement (6,7). Because survival is dependent on pulmonary function, the efficacy of GT placement remains in question (8).
We performed a retrospective analysis of a large number of patients with CF who received GTs to determine whether GT placement was associated with a long-term improvement in nutritional status and pulmonary function. Our objectives were to assess the efficacy of GT placement on nutritional status, as measured by BMI percentile (pBMI) for age and sex, and pulmonary function, as measured by percent-predicted forced expiratory volume in 1 second (ppFEV1). Our hypothesis was that GT placement would be associated with significant improvement in pBMI and this improvement would be associated with a significant improvement in ppFEV1.
PATIENTS AND METHODS
The University of Minnesota institutional review board reviewed and approved the study protocol. Data were collected using the Minnesota CF Database. The Minnesota CF Database is a 30-year encounter-based database for our center. It includes data not gathered in the national CF database, including expanded treatment and laboratory data. Database entries from the first GT placement noted from 1989 to 2007 were reviewed.
Subjects were identified in the database by the presence of a GT. Patients were included in the study if they had at least 5 ppFEV1 measurements and at least 1 pBMI before and after GT placement. All of the patients received a diagnosis of CF according to Cystic Fibrosis Foundation standards (9). Subjects with transplantation (liver or lung) before GT placement or with other causes of malabsorption (eg, severe short bowel, celiac disease) were excluded.
Because CF treatment has changed during the 20 years the data spanned, each subject served as his or her own control. For each subject, data were obtained from all of the outpatient visits before and after GT placement. Inpatient visit data were not collected to avoid transient pulmonary and weight changes associated with acute exacerbation of pulmonary disease. For each visit, the subject's age, height, weight, FEV1, and National Health and Nutrition Examination Survey ppFEV1 were collected. Growth measurements were generally performed by experienced personnel at the time of pulmonary function testing using well-maintained scales and a stadiometer; because the same laboratory performs both adult and pediatric pulmonary function testing, the measurements were generally from the same devices for both groups. Although some variations are inevitable, the use of multiple measurements over time (>3000 BMI points in 46 subjects) reduces the effect of any 1 (possibly) inaccurate measurement. Pulmonary function testing was done by experienced personnel because they are generally able to gauge compliance with the process, even in young children. The requirement for at least 5 measurements of ppFEV1 before GT placement ensured that subjects of all ages would have experience with the procedure.
BMI was calculated for all of the subjects and BMI percentile (pBMI) for age and sex was calculated using the Centers for Disease Control and Prevention equation (10) for subjects 3 years or older and 18 years or younger. To allow continuous comparison of the subjects, for subjects 18 years and older, Centers for Disease Control and Prevention pBMI was calculated with the subject's age set at 18 years.
To assess the effect of GT placement on nutritional status and the preservation of this effect, the median pBMI values were compared during 4 time periods: 24 months immediately before GT placement, 12 months after GT placement, 12 to 24 months after GT placement, and 36 to 48 months after GT placement. For each patient, the median pBMI during each of the 4 time periods and the change in median pBMI from the baseline time period until each of the 3 follow-up periods were calculated. The null hypothesis that the median change in pBMI for all of the patients from the baseline time period until each of the 3 follow-up periods was zero was tested using a nonparametric signed rank test. If the pBMI change for all of the patients was significant, a nonparametric Kruskal-Wallis test was used to determine whether the changes differed by sex and age group.
Pulmonary Function Analysis
We next wanted to examine the effect of GT placement on ppFEV1. Because patients with CF are expected to experience a decrease in pulmonary function with time, estimated at a decrease in ppFEV1 of −1.5% to 3%/year (2,3), a different strategy was necessary. We examined the slope of deterioration of lung function (as represented by ppFEV1) before GT placement compared with the slope after GT placement. Longitudinal mixed model analysis was used to assess the effect of GT placement on pulmonary function, measured as ppFEV1. In this approach, each patient's measurements were modeled as a linear function of time with both a step change in the intercept (difference between the estimated ppFEV1 immediately before and immediately after GT placement) and a slope change (difference between the rate of ppFEV1 change before and after placement). The rate of deterioration of lung function and the magnitude of the immediate (step) and gradual (slope) changes at GT placement were allowed to differ by age group and sex. Each patient's rate of deterioration and step and slope changes were allowed to vary somewhat around the mean for their age group and sex via patient-specific random effects. The final model equation was
where Yij is the outcome of interest (lung function) for the ith patient on the jth measurement, adulti is the age status of the ith patient at placement (1 if older than 18 years of age, 0 otherwise), malei is the sex of the ith patient (1 if male, 0 otherwise), yearsij is the number of years since tube placement (negative if before placement), indgt0ij is an indicator variable that equals zero before placement and 1 afterward, years gt0ij equals zero before placement and equals “years” afterward, β0 through β3d are the (fixed) model coefficients for all of the patients, α0i through α3i are the(random) model coefficients for the patient-specific effects for the ith patient, which are assumed to be normally distributed with mean zero, and εij is the residual error.
The null hypotheses that there is, on average, no step change in lung function at placement (β2 = 0 for all 4 groups, ie, boys 3 years or older and 18 years or younger, girls 3 years or older and 18 years or younger, men older than 18 years, and women older than 18 years) and that there is, on average, no change in the rate of lung function deterioration (β3 = 0 for all 4 groups) were tested using t tests.
This model allowed comparison of the rate, or slope, of pulmonary deterioration before and after GT placement. Including a step change in the analysis enabled us to determine whether there is a significant immediate change in ppFEV1 at the time of GT placement. With this analysis, we would capture both long-term change in ppFEV1 associated with GT placement and any significant deterioration associated with the surgical procedure itself.
We also tested the possibility that the changes in ppFEV1 after GT placement may be correlated with the level of ppFEV1 at GT placement (eg, subjects with better lung function at GT placement have better results thereafter) using regression analysis. To determine whether ppFEV1 at the time of GT placement was a significant predictor of pulmonary improvement after placement, the estimated step and slope changes for the 46 patients were regressed against their ppFEV1 levels at placement.
Our final goal was to determine whether the change in ppFEV1 after placement of a GT was correlated with the change in physical growth. The change in physical growth after GT placement for a given patient was taken as the difference between their median pBMI during the 2 years immediately before GT placement (baseline) and their median pBMI during the second year after GT placement. The possibility that changes in ppFEV1 after GT placement may be correlated with changes in physical growth was tested by regressing the estimated ppFEV1 step changes and slope changes separately against the change in median pBMI and testing whether the pBMI change was a significant predictor. The data analysis was carried out in SAS 9.2 (SAS Institute, Inc, Cary, NC). Significance level was set at a P value of 0.05.
There were 78 subjects with CF and a GT identified in the Minnesota Cystic Fibrosis Database. Four subjects had no date given for GT placement. Twenty-eight subjects were excluded for having <5 ppFEV1 measurements before and/or after GT placement, generally because they were too young to perform pulmonary function testing. There were 46 subjects who met all criteria for inclusion in the study (Fig. 1).
The patients ranged from 5 to 50 years of age at the time of GT placement. In the group, 28 of 46 were male patients. The median observation time before GT placement was 2.2 years (range 0.8–6.4 years) and after GT placement was 4.5 years (range 0.4–13.9 years); however, only the BMI data from 2 years before GT placement was used. Five subjects underwent lung transplantation at variable periods after GT placement. BMI and FEV1 data after transplantation were omitted from this analysis. Demographic data are summarized in Table 1. For the 46 subjects, there were a total of 3086 ppFEV1 measurements and 3091 pBMI measurements.
Does Placement of a GT Acutely Worsen Pulmonary Function in Patients With CF?
The immediate effect of GT placement on ppFEV1 was measured by the step change. Longitudinal analysis of the ppFEV1 data yielded a mean estimated step change at GT placement of 2.16% (P = 0.5209) for men, 0.43% (P = 0.9167) for women, 0.99% (P = 0.6518) for boys, and −0.91% (P = 0.7373) for girls. In our population, on average and at subgroup analysis, there were no significant immediate positive or adverse consequences of GT placement on ppFEV1.
Is Placement of a GT Associated With Improvement in BMI Percentile in Patients With CF?
Nonparametric signed rank tests of the change in pBMI from the 2 years before to the 12 months after GT placement in all 46 subjects showed a median pBMI change of 6.3% (P = 0.0007). For the 39 subjects who had pBMI measurements in both the 2 years before placement and the second year after placement, the median pBMI change in the second year after GT placement was 13.3% (P ≤ 0.0001). For the 29 subjects who had pBMI measurements in both the 2 years before placement and the fourth year after placement, the median pBMI change in the fourth year after placement, compared to baseline, was 8.9% (P = 0.0067). The pBMI changes are summarized in Table 2. Our data show that GT placement improves pBMI in patients with CF, and that the improvement is sustained at least 4 years after placement.
To determine whether the pBMI improvements after GT placement differed by sex or age group, we performed subgroup analysis (Table 3). For the first year after GT placement, the pBMI change was positive for 36 of the 46 subjects (78%), including for 12 of the 13 girls, 16 of the 20 boys, and 7 of the 8 men, but was negative for 4 of the 5 women (Table 3). Mean raw BMI for the 13 adults increased from 18.2 to 18.6. The median pBMI change was positive for 3 of the 4 groups, but negative for women. The differences by sex/age group are significant (Kruskal-Wallis test, P = 0.0130).
For the second year after GT placement, the pBMI change was positive for 31 of the 39 patients (79%), including 10 of the 12 girls, 15 of the 19 boys, and 5 of the 5 men, but was negative for 2 of the 3 women. The median pBMI change was again positive for 3 of the 4 groups, but negative for women. Raw BMI data for the 8 adults at year 2 was 18.6, compared with 18.2 before GT placement. The differences by sex/age group are of borderline significance (Kruskal-Wallis test, P = 0.0667), perhaps because of the small number of women.
For the fourth year after GT placement, the pBMI change was positive for 19 of the 29 patients (66%), including 8 of the 10 girls, 8 of the 15 boys, and all 3 men, but was negative for the 1 woman with data at 4 years. The median pBMI change was again positive for 3 of the 4 groups, but negative for the 1 remaining woman. Raw BMI data for the 4 adults at year 4 was 18.7, compared with 18.2 before GT placement. The differences by sex/age group are not significant (Kruskal-Wallis test, P = 0.3777).
Although the number of women in our study was small, it appears that pBMI improves after GT placement for children and for men, but not for women.
What Are the Reasons for Recurrence of Poor Weight Gain After GT Placement?
We reviewed the charts of subjects who had at least 1 pBMI after GT placement that was lower than their pre-GT pBMI. Twenty-one subjects (10 female, 11 male) had at least 1 pBMI during one of the follow-up periods that was lower than the mean pBMI during the baseline period. Two of these 21 subjects developed significant gastroesophageal reflux, necessitating a change of the gastrostomy tube to a gastrojejunostomy tube. Two subjects encountered multiple feeding pump malfunctions after GT placement. Adherence to feeding regime was noted to be poor in 3 subjects. Two subjects died approximately 18 months after GT placement. Only 2 subjects had outpatient clinic visits with a gastroenterologist during the periods in which BMI decreased after GT placement.
Is GT Placement Associated With an Improvement in ppFEV1 Slope?
The rate of decline of ppFEV1 before and after GT placement was compared to determine whether GT placement altered the rate of ppFEV1 decline in CF. The mean estimated slope change, or difference between the rate of change in ppFEV1 before and after GT placement, was 5.01%/year (P = 0.0159) for men, 4.48%/year (P = 0.0712) for women, 1.49%/year (P = 0.2297) for boys, and 4.02%/year (P = 0.0107) for girls (Table 4). There was a significant improvement in the mean rate of ppFEV1 decline in men and girls, and a trend toward significant improvement in women. The mean rate of decline improved also in boys, but the improvement was not significant. Note, however, that for boys the initial rate of decline before GT placement was low, −1.13%, and was not significant (P = 0.3453), in contrast to the initial rates of decline of the other groups.
The rate of ppFEV1 decline improved after GT placement in 37 subjects but worsened in 9 subjects, that is, the lung function decreased more rapidly after placement of the GT, compared with before GT placement. Three developed diabetes at variable periods after GT placement, 1 had diabetes at the time of GT placement, and 5 did not have diabetes. One died of congenital cardiac disease unrelated to CF. No liver disease was diagnosed in these 9 individuals. Two of these subjects died within 2 years of GT placement secondary to respiratory failure (one at 7 months and one at 20 months after GT placement).
Are Changes in ppFEV1 After GT Placement Correlated With Level of ppFEV1 at the Time of GT Placement?
We next determined whether the subject's ppFEV1 at the time of GT placement affected the change in ppFEV1 after GT placement. Regressing the estimated immediate (step) changes in lung function against the level of lung function at GT placement revealed an estimated coefficient for the ppFEV1 level at placement of −0.110 (P = 0.0119), suggesting that the largest immediate improvements in ppFEV1 occurred in patients who had the lowest lung function levels at placement; however, this result was entirely dependent on 1 patient, who had ppFEV1 at placement of 128% but a large negative step change in lung function after tube placement. This observation was deemed a potential outlier because it was poorly fit, with a studentized residual of −3.8, which translates to a difference of >3 standard deviations between this observation and the value predicted by the model, and because it was greatly influential, because removing this observation from the analysis changed the fitted value by −2.1 standard deviations. Based on these characteristics, the poorly fit and greatly influential observation was removed. Refitting the regression model without this observation revealed an estimated coefficient for the ppFEV1 level at placement of −0.041 (P = 0.2802), indicating that the immediate (step) changes in ppFEV1 after GT placement are not correlated with the level of ppFEV1 at the time of GT placement.
Regressing the slope changes in ppFEV1 against the level of ppFEV1 at GT placement revealed an estimated coefficient for the ppFEV1 level at placement of −0.0046 (P = 0.8380), indicating that the trend changes in ppFEV1 after GT placement are not correlated with the level of ppFEV1 at the time of GT placement.
Previous investigators have suggested that a ppFEV1 <25% at GT placement predicts a poor result (6). In our study, 4 patients had a ppFEV1 <25% at the time of GT placement. The model-estimated step and slope changes in ppFEV1 were positive for all 4 of these patients.
Do Changes in ppFEV1 After GT Placement Correlate With Changes in pBMI?
Results of regressing the estimated ppFEV1 step and slope changes against the change in median pBMI to the second year after GT placement were reviewed to determine whether the change in pBMI was a significant predictor of ppFEV1 improvement. Regression analysis of the ppFEV1 step changes yielded an estimated coefficient for the change in pBMI that was slightly negative but not significant (P = 0.6000). Similarly, regression analysis of the change in ppFEV1 slope gave an estimated coefficient for the change in pBMI that was slightly negative but not significant (P = 0.3103). Thus, the changes in ppFEV1 that we observed after GT placement were not significantly correlated with change in pBMI.
The effect of inadequate nutrition on the progress of CF pulmonary disease and survival is well documented (11,12). Morison et al (12) conducted a cross-sectional study using the UK CF registry and found that mean height and weight standard deviation scores in children with CF were significantly lower than those seen in the general population and that the decline continued with age. In a study comparing 2 CF populations, Corey et al (11) observed that the median survival was 9 years longer in Toronto compared with Boston. CF care was similar in the 2 clinics except with respect to nutritional management, resulting in patients with CF in Toronto having better growth parameters, specifically height and weight. The researchers suggested that the higher survival rate in the Toronto CF population was secondary to superior nutritional status.
Growth retardation observed in patients with CF is not inherent to the CF transmembrane conductance regulator mutation, but rather is secondary to an energy imbalance. Energy imbalance in CF is manifested by excessive energy losses and decreased enteral intake in the setting of increased energy expenditure, putting patients with CF at particular risk of malnutrition (13). GT placement addresses energy imbalance. It has been demonstrated that GT placement improves weight and growth in patients with CF and malnutrition. In a study involving 54 patients with CF and a GT, Williams et al (7) showed a significant improvement in BMI after GT placement, but failed to show a significant improvement in pulmonary function, as measured by ppFEV1 and forced vital capacity, after the GT was placed. A study by Efrati et al (6) involving 21 patients with CF and a GT confirmed an improvement in BMI and found a trend toward pulmonary function stabilization. Truby et al (14) reviewed the effect of GT placement on 14 patients with CF for 2 years, showing modest improvement in BMI z scores and stabilization of FEV1 in the 7 subjects able to perform pulmonary function testing. No study directly answers the question of whether placement of a GT in patients with CF and malnutrition results in significant improvement in pulmonary function, a necessity if survival is to be improved.
Our study extends these observations with a well-documented population, with growth and pulmonary data for a median length of 2.2 years before and 4.5 years after GT placement. This extensive body of data allowed us to confirm that GT placement in patients with CF improves pBMI and to demonstrate that GT placement is associated with stabilization in lung function, measured as ppFEV1. Improved pBMI was seen in boys and girls and in men. The improvement was durable, continuing out to 4 years where data were available. In subgroup analysis, only women failed to achieve pBMI improvement after GT placement. The reason for this is uncertain. The number of women receiving GTs was small. Body image issues may have made women more reluctant to use the GT fully, resulting in poorer weight gain (15). Without further study, this remains speculative.
Reduction in the rate of decline in ppFEV1 after GT placement was significant in men and in girls and trending toward significance in women. In our study population, neither the initial decline in ppFEV1 before GT placement nor the improvement after GT placement was significant on average in boys. Although the changes in ppFEV1 that we observed after GT placement were not significantly correlated with pBMI, it is possible that pBMI is not the ideal variable to assess the relation of the effect of improved nutrition on pulmonary function; body composition may be a better tool to assess these effects. King et al (16) showed that in adults with CF, fat-free mass depletion was not consistently predicted by BMI and that fat-free mass index was independently associated with FEV1. Future studies of nutritional interventions in CF should evaluate the correlation of changes in body composition with changes in pulmonary function.
In 9 subjects, GT placement was associated with a worsening in ppFEV1 trend after placement. We found no marker for failure of a GT to improve ppFEV1. In previous studies of children with CF and placement of a GT, investigators have suggested that low preoperative ppFEV1 predicts mortality after GT placement (17). In our study, subjects with ppFEV1 <25% before GT placement did not experience deterioration in ppFEV1 slope after the procedure, and ppFEV1 at the time of GT placement did not correlate with the ppFEV1 slope change after GT placement. At present, we see no contraindications to GT placement based on preoperative lung function alone.
A retrospective study presents a number of limitations. The present study includes data collected over several years, when changes in CF care occurred that may affect both pBMI and ppFEV1; however, the effect of these changes is limited by using each patient as his or her own control. There may be confounding variables not accounted for contributing to improved nutrition and pulmonary function. We cannot demonstrate, for example, that improved pBMI is solely the result of GT placement, or even that our subjects used their GT. Another potential limitation of our study is that it was conducted at a single center using 1 population of patients with CF. With the extensive heterogeneity of phenotypic expression of CF, it is unknown whether the findings that we obtained by studying the CF population in the state of Minnesota can be applied to the CF population as a whole. Ideally, the next step to confirm these findings would be a prospective trial assessing the effects of GT placement in this population. Development of a reliable means of monitoring GT use would be essential to such a study.
In conclusion, we were able to demonstrate that GT placement in patients with CF results in significant improvement in both pBMI and ppFEV1. Women patients with CF are an important exception in that they do not have improved pBMI after GT placement. The change in lung function after GT placement is not dependent on the level of lung function at placement. It is reasonable to speculate that the stabilization of pulmonary function seen after GT placement would be associated with prolonged survival.