2. CAPÍTULO 2: METODOLOGÍA
2.4. TIEMPO HISTÓRICO EN QUE SE HA CONSTRUIDO LA CIUDAD
2.4.5. Consolidación de la ciudad comercial
These findings are unlikely to be due to biases arising from the study design. First, in order to reduce errors due to misclassification, estimation o f gestational age for the classification o f birthweight status was based on sonographic assessments before 20 weeks gestation, which are currently considered to be the most accurate means o f estimating gestation and hence birthweight centiles. In addition, matemal reports o f post-natal smoking were confirmed by cotinine assay o f matemal saliva and infant urine samples collected at time o f test. Thus, random errors arising from misclassification o f birthweight or smoking status were minimised, as these errors might lead to an underestimation o f the strength o f the association between birthweight status and airway function. Although the possibility o f residual confounding (discussed in section 5.10.1) may still be present, reasonably accurate adjustments for matemal smoking were made in the model.
Secondly, lung function tests were performed prior to any lower respiratory illness (LRI), thus enabling respiratory function in SGA and AGA infants to be compared without confounding due to LRI.
Thirdly, the use o f the raised volume technique to measure airway function allowed measures of forced expiration to be compared between infants over an extended volume range (Le Souëf et al. 1996).
Finally, this study had sufficient power to detect clinically important differences in airway function between SGA and AGA infants. The calculations for power of study suggested that a sample size of approximately 40 per group (SGA non-smoking, SGA smoking, AGA non-smoking and AGA smoking) would provide 80% power to detect a 10% difference in adjusted estimates of forced expiratory flows and volumes between birthweight groups, significant at the 5% level. Although, as with any statistical testing, for a given total sample size, power is maximised if group sizes are similar, it is not essential that they are absolutely equal, and any power calculation can only ever be an approximation. Details on the power o f study and sample size calculations are given in Appendix I. This estimate was based on knowledge from previous studies conducted in our laboratory, which has shown that age, length, sex, matemal smoking status and family history o f asthma explain a high proportion of variability in airway function at this age (Dezateux et al. 1999; Dezateux et al. 2001). In this study, retrospective analyses have shown that 47% o f the variability observed in FEV0.4 was accounted for by variables such as body size, age, matemal smoking and matemal social class.
According to power calculations, this should provide 80% power to detect a (0.05) 5% o f the total variability in FEV0.4 that can be explained by group differences due to birthweight status and/or smoking status (Figure 5.1). A difference o f this magnitude is equivalent to 9% (0.05/0.53) o f the variability remaining after accounting for known confounding factors, i.e. this would account for the power to detect tme difference between SGA and AGA infants, with and without smoking mothers.
Figure 5.1 Chart showing proportion variability of FEV0 .4 due to birthweight
status and other confounding factors
Sample size 40 AGA, 40 SGA
•§
I
.14 .13 . 12 .1 0 .09 .08 .07 .06 .05 .04 .03 .02 .01 0.00 0.0 1 2 ,3 .4 .5 .6 ,7 ,8 .9 1.0 Power to detect 95% 90% 85% 80% 75%Proportion variability accounted for by other variables
In reality this equates to an average (95% Cl) reduction in FEV0.4 of 8 (1, 16) mL in SGA compared to AGA infants after allowing not only for matemal smoking, but for differences in length, age and matemal social class (Table 4.20). Thus, as total variability of FEV0.4 in this study population was 159 mL (ranged 44 - 203 mL) and being SGA was associated with on average a reduction in FEV0.4 of 8 mL, this is equivalent to 5% of total variability observed (as predicted). After adjustment for other confounding factors, for an infant of average length (54 cm) and age (6 wk), FEVo.4 = - 138- 8 (SGA) + 4 (length, cm) + 2 (age, wk) (Table 4.20). Thus, for such an SGA infant, FEV0.4 predicted is 82 mL while that for an AGA infant with similar characteristics is 90 mL. This actually equates to a 9% reduction in FEV0.4 when comparing between an SGA and AGA infant of similar body size and age. A similar reduction was observed in FVC. Therefore this study proved to be adequately powered to detect variability in airway function attributable to low birthweight of a magnitude that is clinically and aetiologically relevant (see Section 5.7).
In this study population, the total between subject variability for MEF25 was 320 mL.s'* (ranged 56 - 377 mL.s*^), the variability o f which was virtually twice that for FEV0.4. The increased between subject ‘unexplained’ variation in MEF25 will to some extent reflect the increased ‘noise’ associated with flow parameters, as reflected by a much higher within subject coefficient o f variation for MEF25 compared to that for FEV0.4 (MEF25 vs. FEV0.4, mean [95% Cl]: 8.8 [7.9, 9.6] vs. 3.7 [3.4, 4.1]) as shown in Table 4.32. In this study, although being bom SGA was also associated with an average reduction in MEF25 o f 9% (22 mL.s'^) compared to an AGA infant o f similar characteristics after adjustment for sex and family history o f asthma, this difference just missed statistical significance (p = 0.07; Table 4.21). This adjusted difference in MEF25 according to birthweight status did in fact become significant when an additional 40 infants were recruited after completion o f this thesis (bringing the final total to 91 SGA and 132 AGA infants) in order to attain the numbers required for subsequent follow-up o f respiratory function in this cohort (Dezateux et al. Lancet; submitted). This illustrates the practical difficulties in estimating power o f study prospectively when using several outcome measures, with differing degrees o f total and unexplained variability, especially when new techniques are being used, with little prior evidence on which to base such estimates. The results from this study indicate that significant differences in airway function according to birthweight status were detectable with fewer subjects when FEV0.4 rather than MEF25 was selected as a primary outcome measure, due to the lower between subject variability o f the former. Nevertheless, additional important information regarding other determinants o f airway function (e.g. influence o f sex and matemal history o f asthma. Table 4.21) were obtained by including assessments