INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The pediatric pulmonary and cardiovascular complications of vertically transmitted human immunodeficiency virus (HIV) infection have been investigated in a multicenter natural history study (P²C² HIV, or P²C²) funded by the National Heart, Lung, and Blood Institute. The pulmonary arm of the protocol was conducted to elucidate the pathophysiology, incidence, course, and outcome of lung involvement in children, both infected and uninfected, born to HIV-infected women.

The design and methodology of the study have been previously reported (1). Briefly, two groups of children were enrolled. Infants and children with confirmed vertically transmitted HIV infection were enrolled after 28 d of age (Group I). Children of HIV-infected women were enrolled before birth or within 28 d of birth, before their HIV status was determined (Group II). Group II provided an opportunity to study the earliest features of the disease, and because most children remained free of infection, to study the effects of prenatal exposure to HIV infection. Children who remained HIV uninfected (Group IIb) served as a comparison group for children who became infected (Group IIa).

The aim of this portion of the P²C² study was to test the hypothesis that infants who were exposed to HIV before birth but remained uninfected have normal airway function as manifested by forced expiratory flows, measured at functional residual capacity (V'max,_FRC). To create a comparison group, we reanalyzed previously reported data from a normal population of infants between 60 and 90 cm in length (2).

Beginning in May 1990 through January 1994, 805 patients were enrolled at five centers and followed until January 1997. Group I contained 205 patients and Group II contained 600. In Group II, 93 (17%) were determined to be infected (and assigned to Group IIa), and 463 (83%) were uninfected (assigned to Group IIb). In the remaining 44 children, HIV status was never determined because of death or loss to follow-up; these subjects were excluded from this report.

Many of our Group IIb subjects were not followed for the entire duration of the study. After the P²C² study began, the rate of HIV transmission was found to be lower than anticipated, producing an excess of Group IIb subjects. We extended the study to recruit more infected children. Approximately half of the HIV-uninfected cohort was randomly selected to remain on study as a control group and the remainder was randomized off study.

Unfortunately, financial constraints made it impossible to enroll a reference sample that was similar to our Group IIb but that had matured in an HIV-uninfected environment. In searching for an appropriate reference group, we obtained data from both Hanrahan and coworkers (2) and Tepper and Reister (3), who granted permission for us to reanalyze the data. We selected subjects between 60 and 90 cm in length because this height range was well represented in our sample. The analyses showed that the data from the two studies were similar and produced virtually identical regression lines for flow over height. We selected the Hanrahan set for comparison because it included subjects within the age and size range of most of our subjects, and was a longitudinal study like ours. In addition, the Hanrahan study used techniques similar to ours and provided important demographic and smoking data that were not available for the Tepper group. Also, one of the investigators in this study helped to implement the Hanrahan study.

Of the 463 Group IIb subjects 150 were lost to follow-up or randomized off study prior to infant pulmonary function tests (IPFT) testing and 14 had unacceptable readings. To keep our group as comparable as possible with the Hanrahan group, we excluded 14 additional patients from the study; 12 were of ethnic origin other than black, Hispanic, or white, the 3 groups represented in the Hanrahan study, and 2 had no studies when the infants were longer than 60 cm. As a result, we included 285 patients with analyzable data. We excluded 9 studies from these 285 patients because they were taken when the subjects were not within the height range; this resulted in 500 studies; 46% of the patients had one measurement, 34% had two measurements, 19% had three measurements, and 1% had four measurements. The demographic characteristics of the subjects are presented in Table 1.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The institutional review board at each center approved the study, and informed consent was granted by the legal guardians of all participants. HIV status was determined by HIV culture: (1) HIV positive, two cultures positive at less than 6 mo (cord blood excluded); (2) HIV negative, two cultures negative with at least one at 5 or more mo of age (4). Data were sent electronically or on hard copy to the coordinating center (the Cleveland Clinic Foundation, Cleveland, Ohio), where the data were analyzed. Regular conference calls and meetings were conducted to ensure that methods were uniform.

A pulmonary physical examination including pulse oximetry, respiratory rate, and lung auscultation was performed every 3 mo from birth to 18 mo and every 6 mo thereafter. Height and weight were measured using a standard infant scale and stadiometer. IPFT were performed at 6-mo intervals starting at age 6 mo. At age 18 mo, as predetermined by the study protocol, follow-up was discontinued for some subjects, and the remaining continued to have IPFT until age 30 mo. Prior to pulmonary function tests subjects were examined and their caregivers questioned to ascertain that the child was free of lower respiratory infections for at least 2 wk before testing.

For IPFT, the subjects were sedated with 80-100 mg/kg of chloral hydrate (maximum dose 1,000 mg) and were studied in the supine position using the Sensormedic 2600 Infant Pulmonary Function system (SensorMedics Corp., Yorba Linda, CA). A face mask of the appropriate size was used; a leak-proof seal was created with a soft air-filled collar around the mask perimeter (Vital Signs, Totowa, NJ). The face mask was connected to a pneumotachograph of the appropriate size (10, 30, or 160 LPM, with volume ranges of 0-127 ml, 0-255 ml, and 0-2048 ml, respectively; Hans Rudolph Inc., Kansas City, MO). Flow was digitally integrated to obtain volume, and all measurements were displayed continuously on a computer screen. In this study, we report data on forced expiratory flows, measured at functional residual capacity (V'max,_FRC). To keep the scope of this report reasonable, we will separately report details of the passive mechanics in this sample, including total respiratory system compliance (Crs), resistance (Rrs), and time constant (rs) by the single occlusion technique (5), and measurements of functional residual capacity (FRC) by nitrogen washout (6).

Partial forced expiratory flows were obtained by rapid thoracic compression (RTC) (7). The flow was expressed as the maximal flow at FRC (V'max,_FRC) (7). The infant was placed supine in an inflatable jacket (Hammersmith Hospital, London, UK) fitting snugly under the arms and around the abdomen and thorax. After two or three tidal breaths, the jacket was inflated through a solenoid valve connecting the jacket to a pressurized chamber, which was triggered by a computer at the end of inspiration. As a rule, the jacket pressure began at 40 cm H₂O. Pressure was increased in increments of 10-15 cm H₂O to a maximum of 100 cm H₂O or until no further increase or decrease in flow was observed, when flow limitation was assumed to have been reached. The face mask was removed after each maneuver, and at least 60 s was allowed between successive RTC maneuvers to limit the effect of volume history on flow.

The curves were obtained by technicians and investigators experienced in identifying "flow-limited" segments of the curves. The hardware stored all curves, but only those deemed technically acceptable were saved. Curves were technically acceptable if they were free of noise and obvious inspiratory effort over the flow-limited segment. The reported value of V'max,_FRC was the mean of the three highest acceptable curves in which flow limitation was judged to have been achieved. In the early stages of the study, the choice of the curves for analysis was based on increasing the jacket pressure until no further increase in V'max,_FRC was observed. In later phases of the study, software was developed to optimize the detail of the descending portion of the expiratory limb, the critical portion in determining flow limitation (10). The program allowed the volume axis to be expanded so that the slope of the descending portion of the expiratory limb was approximately 45°. It also permitted the user to move individual loops acquired at increasing pressures along the volume axis to superimpose the segments of interest at isoflow points. If the segments were tightly superimposable at two different pressures, and if the segment of each curve constituted at least 20% of the inspired volume flow, limitation was assumed to be present over the overlapping segments. This software was used only to assist in the analysis by alerting the researcher that flow limitation had been obtained. There was no attempt in this study to determine pressure transmission across the chest wall.

Quality Control

Hard copies of transmitted data were provided by the coordinating center to the study sites, where they were compared with the original data. The reproducibility of the analyses for the various tests was assessed by the coordinating center, which repeatedly sent each center data disks containing that center's data as well as data from other centers for reanalysis. A full review of the quality control methods is being prepared as a separate report.

Statistical Analysis

The appropriateness of the normality assumption for mean V'max,_FRC was checked using a normal probability plot and by plotting the data relative to the normal probability density function. These plots suggested the normality assumption was reasonable for V'max,_FRC and hence parametric analyses were performed.

Comparisons between centers for mean V'max,_FRC were performed using ANOVA. Plots of V'max,_FRC versus height and log (V'max,_FRC) versus log (height) were made. Linear regression models were fit to both scatterplots. The resulting fits were very similar. The log (V'max,_FRC) versus log (height) model was back-transformed, and the fitted values of the two models were compared. The transformed model proved to be linear except for a small drop at the lower end of the height range. Thus, the reported data in Figure 2, as well as all subsequent analyses, were done using the untransformed data.

View larger version (22K): [in this window] [in a new window]   Figure 2.   V'max,FRC (ml/s) versus height for infants exposed to HIV and for healthy infants (Hanrahan). Each point represents the mean of the three highest V'max,FRC values from each measurement session. Open circles (with dashed regression line) show data from HIV-exposed children. Solid circles (with solid regression line and 95% prediction intervals) are from the Hanrahan cohort of healthy children. The relationship between V'max,FRC and height (slope of the regression line) is similar in the two groups (p = 0.22), but the HIV-exposed group had significantly lower intercepts at a height of 60 cm (128 and 164, respectively; p < 0.001). Regression equations for HIV-exposed children: V'max,FRC = 128 + 7.16 × (length  60); for Hanrahan data: V'max,FRC = 164 + 8.59 × (length  60).

Longitudinal regression analyses were done using linear random effects models with random intercepts only to help model the correlation among repeated observations on a given child over time (11). To test the equality of intercepts and slopes, a hierarchical approach was used. First, a test of commonality of slopes was done by fitting unrestricted regression models to all groups of interest. If no significant differences between slopes were found, then a common slopes model was fit to test equality of intercepts across groups of interest. The p values for these tests are reported in the tables. Models studying the effects of race/ethnicity on the V'max,_FRC and height relationship are first reported, and then multiple predictors are adjusted for, including birth weight, sex, mother's smoking status during pregnancy, her drug use during pregnancy, and gestational age.

All hypothesis tests were two sided. A p value  0.05 was considered to indicate significance.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Flow Values

Flow at FRC (V'max,_FRC) increases with height in a linear fashion (Figure 1). However, mean V'max,_FRC differed significantly between centers. These differences were apparent: (1) when mean values of flow within specified height ranges were compared (Table 2); and (2) when regression lines of flow versus height were compared by center (Table 3). Two centers (numbers 3 and 4) had mean flows that were significantly lower than the other three centers, except for measurements in infants 80-90 cm long (Table 2), and had significantly lower intercepts at 60 cm height (Table 3). The slopes of the regression lines varied between centers, but these differences did not reach significance (p = 0.19; Table 3). Within-center variability was also substantial, with coefficients of variation ranging from 29% to 76%. To identify possible contributors to the observed variability, we examined the effects of the magnitude of thoracic compression (applied jacket pressure), race/ethnicity, sex, birth weight, gestational age, and maternal smoking.

View larger version (21K): [in this window] [in a new window]   Figure 1.   V'max,FRC (ml/s) versus height for infants studied at 5 clinical centers. Each point represents the mean of the three highest V'max,FRC values recorded for the session. As shown by the regression lines, centers 3 and 4 had lower V'max,FRC values.

Contributors to Variability

Jacket pressures used to obtain maximal expiratory flow varied (Table 4). To identify low jacket pressures that might have been insufficient to reach flow limitation, we identified all flow-volume curves obtained throughout the 5 yr of the study at jacket pressures of less than 60 cm H₂O and reexamined them to verify that jacket pressures were indeed increased to a level at which no further increase in flow was observed. Eleven questionable studies were found and removed from subsequent analysis.

Also, the pressures within height intervals were analyzed by year of study (Table 4). Within each interval, jacket pressures increased significantly over the years (particularly between the first and second years) in all but the largest children (12). However, within each height group, the increasing pressure did not affect the value of the flow, and mean flows within the height groups did not increase significantly over the years of the study (data not shown).

The slope of the regression line did not vary by subject sex, although the intercept for females was slightly but not significantly higher (p = 0.10, Table 3). When birth weight (below 2,500 g), gestational age (< 37 wk), maternal smoking during pregnancy, and maternal drug use during pregnancy were included in the model, the intercept for females was still somewhat but not significantly higher than for males (p = 0.095).

The racial/ethnic distribution within centers differed substantially (Table 1), which may have contributed to the between-center variability. A linear random effects model showed that race/ ethnicity produced no significant differences in slopes (p = 0.21) but did produce significant difference in intercepts: Hispanic children had a lower intercept than either blacks or whites (p = 0.006; Table 3). Because of the widely uneven racial/ethnic distribution of subjects between centers, and resulting small numbers, a meaningful analysis of racial influence on flow within a center could not be carried out. Flows were significantly lower in Hispanic children even after accounting for birth weight (< 2,500 g), gestational age (< 37 wk), maternal smoking during pregnancy, and maternal drug use during pregnancy.

Comparison to Data in the Literature

Our subjects had a regression line intercept significantly lower than the intercept of our comparison group selected from the data of Hanrahan and coworkers (Figure 2; 128 [95% CI 109, 148] versus 164 [95% CI 136, 192]; p < 0.001). However, the slopes were similar (7.2 [95% CI 6, 8.3] versus 8.6 [95% CI 6.7, 10.6]; p = 0.22). These observations for intercept and flows were true for both males and females.

Our study group contained more black children than did the Hanrahan group, which was largely white and Hispanic (Table 1). When we compared our white subjects with Hanrahan's white subjects (Figure 3), the intercepts and slopes were virtually identical (Group IIb intercept 132 [95% CI 85, 179] versus Hanrahan 144 [95% CI 108, 180]; p = 0.31; Group IIb slope 9.9 [95% CI 6.7, 13] versus Hanrahan 10.8 [95% CI 8.6, 13]; p = 0.63). However, the Hispanic infants in our group had an intercept that was approximately half of the intercept of the Hispanic infants in the Hanrahan group (91 [95% CI 57, 125] versus 196 [95% CI 155, 237]; p < 0.001). In addition, the slope was steeper in our Hispanic subjects (8.0 [95% CI 5.9, 10] versus 4.7 [95% CI 1.5, 7.9]; p = 0.09). As the prevalence of blacks in the two groups was so different, we excluded the black subjects and reanalyzed the data, with similar results.

View larger version (15K): [in this window] [in a new window]   Figure 3.   V'max,FRC (ml/s) versus height by racial or ethnic group in infants exposed to HIV and the healthy comparison group. Within the Hanrahan group of healthy children, the slopes of the regression lines for the white and Hispanic subjects differ significantly. However, the regression lines for the whites in the HIV-exposed group and the Hanrahan group appear to be similar. Hispanic children in the P2C2 study group had a lower mean V'max,FRC at 60 cm than did the Hispanic children in Hanrahan's group. Those three regression lines approach each other as height increases. Regression equations for P2C2 (IIb) data: HispanicV'max,FRC = 91 + 8.01 × (length  60); White V'max,FRC = 132 + 9.90 × (length  60); BlackV'max,FRC = 144 + 6.40 × (length  60). Regression equations for Hanrahan data: HispanicV'max,FRC = 196 + 4.73 × (length  60); WhiteV'max,FRC = 144 + 10.83 × (length  60).

Other demographic factors were also examined. Group IIb had a larger proportion of low-birthweight subjects (17% for IIb versus 6% for Hanrahan; p = 0.009). In contrast, the proportion of children with a gestational age less than 37 wk was larger in Hanrahan's group (30% versus 16%, respectively; p = 0.002). In Group IIb, the percentages of subjects with gestational age less than 37 wk and with birthweight under 2,500 g were fairly evenly distributed by race (gestational age figures were 15%, 19%, and 18% for Hispanic subjects, whites, and blacks, and low-birthweight figures were 15%, 9%, and 19%). In contrast, in Hanrahan's study, the frequency of subjects with gestational age younger than 37 wk was the same in white and Hispanic subjects (30% and 33%, respectively), but 11% of whites and no Hispanic infants had birth weights less than 2,500 g.

The groups did not differ in length for age or weight for age (Figure 4). Further analysis showed that whites in Hanrahan's study had slightly but not significantly higher heights for age than whites in our study (p = 0.06). There were no differences observed in the Hispanics.

View larger version (20K): [in this window] [in a new window]   Figure 4.   (A) Height versus age in HIV-exposed infants and healthy infants. Open circles (with dashed regression line) represent the P2C2 data; solid circles (with solid regression line) represent Hanrahan's data. The regression lines are essentially superimposable; no significant differences were found between the intercepts (p = 0.5), the linear slopes (p = 0.71) or the quadratic slopes (p = 0.61). (B) Weight versus age in HIV-exposed infants and healthy infants. Again, the regression lines are almost identical; no significant differences appear between the intercepts (p = 0.32), linear slopes (p = 0.49), or quadratic slopes (p = 0.49).

We also compared the intercept differences after adjusting for birthweight and gestational age, while maintaining a common slope. Like the earlier analysis, this one showed that intercepts were similar for whites in the two groups, but intercepts were markedly lower for Group IIb Hispanic subjects than for the Hispanic subgroup in the Hanrahan group.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The major finding in this study is that forced expiratory flow in infants whose prenatal environment includes infection with HIV differed only slightly (about 20% less) from values in infants from a healthy comparison group.

We found marked variability among our five centers in height-adjusted values of V'max,_FRC. To understand the sources of this variability, we focused on methodologic issues, biological variability, ethnic differences, and fetal exposure to smoking, drugs, and other environmental factors that can alter lung development and enhance measurement variability. In this discussion, we will first deal with between- and within-center variability and then discuss how our sample differs from the Hanrahan sample.

A number of explanations of the observed between-patient variability are possible. An inherent problem of the RTC technique is the lack of control over lung volume. The degree of sedation and the timing of maneuvers within the sedated period varied, and these variations may have influenced functional residual capacity (FRC). After the mask was removed from the child's face between studies, the technicians waited until they thought FRC had stabilized, but the stabilization could not be confirmed before the next thoracic compression. This may have affected results in three ways. First, repeated thoracic compressions can reduce FRC, and this in turn can reduce values of forced expiratory flow at FRC. Such shifts in FRC have a particular impact on V'max,_FRC in younger infants (13). Second, the volume history was undoubtedly variable, as the lungs were not inflated to 20 or 30 cm pressure (14, 15). Finally, flows obtained at low lung volumes (V'₅₀, V'₂₅) are more variable than those obtained at higher lung volumes even when the extremes of lung volumes can be relatively fixed.

We established strict methodological guidelines and carefully standardized equipment between centers to control measurement errors, calibration differences, and inadequate frequency response of equipment. Quality control meetings and discussions were held throughout the study, producing increasingly stricter and better defined protocols over the first 2 yr together with improved software. From the onset, we used uniform equipment, services, and software. The technicians performing the studies were uniformly trained and attended refresher courses and reviews in the first 3 yr of the study. Of the five participating centers, four had from 2 to 10 yr of previous experience with the RTC technique. Although the data obtained from the least experienced center appear to be no different than data from other centers, the importance of previous experience remains unanswered and should be addressed a priori in future studies. This is particularly important in large epidemiological studies, because the center with the greatest expertise with IPFT may not always be coupled with the population that is best suited to answer the research issues. As a quality control measure, sample studies were reanalyzed by centers and cross-analyzed between centers, but we did not perform site visits to ascertain uniformity in technique. In retrospect, we agree with the conclusions from studies in adults (16) and the recent U.K. extracorporeal membrane oxygenation (ECMO) study (17) that site visits, although costly, would have enhanced quality assurance and might have reduced within- and between-center variability.

Our study protocol did not specify the applied jacket pressure. The applied jacket pressure increased over time in this study, possibly reflecting the technicians' increased confidence over time. Our analysis of the influence of pressure upon flow values does not, however, support the hypothesis that lower applied pressures were related to increased variability, because lower jacket pressures were not associated with significantly lower flows in the first year of the study. Additionally, they were not associated with lower flows in the center that used the lowest pressures in the first year.

The protocol defined flow limitation during forced expiration as curves that could be superimposed at stepwise increasing pressures, regardless of the pressure. We did this to avoid the problem of negative effort dependence (or pressure dependence) (18, 19). As a result, it is possible that we increased within-center variability. However, when we reexamined data from subjects in whom the curves were obtained at pressures less than 60 cm H₂O we found that flow limitation had been achieved by our criteria in all but 11 studies, which were subsequently excluded. As would be expected from the increasing stiffness of the chest wall with increasing age, higher jacket pressures were required to obtain flow limitation in older, larger subjects. A further analysis showed that pressures were increased with each length increment of 10 cm. This observation was true for the study as a whole as well as within each year of the study. It is likely that young infants with high chest compliance (20) will need only modest bag pressures to achieve flow limitation, whereas higher pressures will be required in older children with stiffer chest walls.

Thus, flow limitation was achieved in studies with lower applied pressures. Because of this, and because older children should require higher applied pressure to achieve flow limitation, we did not fix the applied jacket pressure or reject studies that met all the other criteria for flow limitation. Similar observations were published in the recent UK ECMO study (17), which compared infant PFT data between two study centers and found remarkable agreement in V'max,_FRC even though the London center was using jacket pressures that were almost twice those used at the Leicester center [mean (SD) 4.8 (1.7) kPa versus 2.6 (0.9) kPa, p < 0.01]. We think it unlikely that differences in applied jacket pressures account for the observed between-center differences in flow or for the differences between our flow values and those of others in the literature (Table 5).

After our study began, recommendations to measure the pressure transmitted across the chest wall were published (21). These recommendations were implemented in the U.K. ECMO study (17), but not in a more recent study determining expiratory flows from elevated lung volumes in a large sample of normal subjects (22). It seems reasonable that such measurements would improve quality control and reduce variability, although they are probably not indispensable for determining flow limitation. However, measuring transmitted pressure does not increase values of flow for a reference population (23) (Table 5).

Our statistical analyses suggest, but do not prove, that some of the between-center variability stems from differences in the racial and ethnic distributions of the samples. Small differences in pulmonary function measurements between ethnic groups have previously been reported in older children (24, 25). The data on ethnic and racial differences on various pulmonary indices in the neonatal period and infancy are inconclusive but suggest that these differences may be reflected in time to peak tidal expiratory flow as a proportion of total expiratory time (TPTEF:TE) (26), total respiratory compliance (Crs) (27), and airway resistance (Raw) (28). A single study examined V'max,_FRC in 56 premature infants at the time of discharge from the neonatal intensive care unit (26) and found no difference between black and white infants.

We found no significant sex differences in flow. Higher flows for females have been detected in some reports (9, 23) but not others (2, 3, 8). The large variability with the increase in growth may have obscured such differences, and this study may not have been powered to detect them. Previous wheezing, for which we did not control, may also have an impact on the sex differences (15, 23, 29).

The major finding of this study was the small, about 20%, difference between our sample and the reference sample in values of V'max,_FRC. The meaning of this difference in height-adjusted flow is not clear, but our values are well within the range of values reported for other groups of children of similar size (Table 5).

A number of variables that could differ between our group and Hanrahan's might have affected flow values. These include maternal smoking and drug abuse, increased rate of prematurity and perinatal/infantile morbidity, malnutrition, race, socioeconomic status, and effect of maternal HIV infection.

Maternal smoking, reduces infant flow values markedly at birth (30), and accounts for a 16% (31) to 17% (23) reduction in flow at 1 yr of age (75 cm). In our sample, maternal smoking and drug use were self-reported, and we did not verify smoking by measuring urinary cotinine as Hanrahan did (2); thus, both smoking and drug use may have been more common than reported.

A second variable affecting flow values might have been prematurity and birthweight. A previous publication from the P²C² study showed that our group had a high prevalence of prematurity and low birthweight, but little perinatal morbidity and essentially no bronchopulmonary dysplasia (32). The incidence of respiratory infections in the first year of life was low, and similar to that in uninfected age-matched infants (33). We separately analyzed the subjects in this study who had the lowest flows (V'max,_FRC of 100 ml/s or less at heights 75 cm or longer), and found that clinical morbidity was not increased in this subgroup (data not presented). Thus, respiratory tract infections are unlikely to have caused the decreased flow in our cohort.

No differences were observed for weight for age and height for age between our group and the Hanrahan one, and thus malnutrition cannot explain the differences.

Our cohort did differ substantially from Hanrahan's in racial and ethnic distribution, primarily by having more black children. However, we found that results in the white children in the two groups were comparable, suggesting that the methodologies in the two studies were comparable. Thus, the differences between the study groups are likely to reflect not different techniques but true population differences. We speculate that the differences observed between Hispanic children in our study and in Hanrahan's reflect the imprecise definition of the term "Hispanic," which can encompass a number of racial and ethnic designations. It is certainly possible that Americans of Mexican origin are ethnically and racially different from Puerto Ricans or people of South American origin, yet all might be labeled "Hispanic." A recent study examining spirometry of African-American and African-Jamaican children found markedly lower flows in both groups when compared with standard prediction equations for African-American children (34). This suggests that within such heterogeneous groups, some subgroups may have different paces of lung growth, which would affect the slope of flow for height.

Socioeconomic status (SES) may be linked to differences in flow values. School children from lower socioeconomic categories have values of FVC and FEV₁ about 8% lower than those of children of more advantaged socioeconomic status (35). A study of adults (36) suggested a similar relationship. The cause of this difference has not been determined.

Although neither we nor Hanrahan and coworkers recorded the socioeconomic status of our patients, the groups do appear to be different in this respect. Most subjects in our study group were inner-city residents, were the offspring of single mothers, and were extremely poor. In contrast, most of Hanrahan's subjects were from blue-collar working families, probably of a higher socioeconomic status. Our results may suggest that the socioeconomic differences in lung function apply to infants as well. SES may therefore be one of the multiple factors contributing to lower V'max,_FRC in this study population.

We conclude that the effect of intrauterine HIV infection on infant forced expiratory flows cannot be assessed with certainty but is likely to be modest. We are unable to disentangle the effect of the intrauterine HIV environment from other modifying influences. To do so, it would be optimal to compare HIV-exposed infants to a matched population of infants born to mothers unaffected by HIV infection. The ideal comparison group would also be matched for all other factors that could affect V'max,_FRC, such as race, birthweight, socioeconomic status, maternal smoking, and possibly maternal drug use. Such a control group was not feasible at the inception of the study because of cost as well as concerns about the risks of sedating healthy infants. Whether the detected reduction in flow is transient cannot yet be determined. Our cohort should be further studied with conventional lung function tests as the children grow up and reach ages in which reference values have been better established.

A broader conclusion from this study reemphasizes the problem of establishing reference values when large enough databases are not available (37, 38). Our study suggests that the limited reference values for V'max,_FRC may not represent the true variability of the general population, and that comparisons between subpopulations are difficult in view of the many variables that may differ between them. The development of techniques to fix lung volumes and volume history in infant lung function studies (14, 15) should be encouraged; such techniques are likely to reduce the variability of the resulting reference values (22). It is, however, doubtful that such techniques will eliminate the variability in measurements of forced expiratory flow in this young population, because these differences may stem from racial, ethnic, socioeconomic, and biological differences. Large multicenter studies of healthy subjects, using uniform techniques, are necessary to provide an adequate reference population.

Furthermore, a recent publication (22) showed that quality control can be improved and variability reduced by applying the raised volume technique and using improved equipment and software that saves all data for on- and off-line analysis. The European Respiratory Society/American Thoracic Society Task Force is coordinating an international effort to standardize measurement of infant respiratory functions and to urge industry to adopt these recommendations.