Predicted body weight relationships for protective ventilation – unisex proposals from pre-term through to adult

The first observation from this analysis is that the traditional IBW/PBW formulae of Devine [19] and Stewart [13] are not closely aligned with median body weight across a broad young population, as represented by the 4 data sources [21, 22, 24] (Fig. 1). In addition, as has been reported previously [15], the Devine and Stewart curves are highly dissimilar to each other: the Stewart relationship over-estimates body weight by up to 30% compared with the Devine prediction. The Devine prediction appears to be the most widely used in current practice, and which is associated with the ARDSNet improved mortality outcomes [1].

The rationale behind using PBW in lung-protective ventilation rather than actual body weight might be summarized as:

Thus, when configuring a safe and adequate tidal volume, use of ‘predicted’ or ‘lean’ body weight may better reflect both metabolic requirements and lung size than actual body weight [15]. Even though critical illness and lung injury will likely weaken height-based associations with ventilatory demand and respiratory system characteristics, this approach has proven valuable in improving outcomes [1, 7, 8] when combined with the other elements of lung-protective ventilation.

In this context, the Devine formulae have been applied as a coarse but convenient surrogate for lean body weight [16]. However, as can be seen from Fig. 1, the Devine predictions appear to be not at all ‘lean’ when compared to population median weight over the relevant height range. This disparity may be due to the populations represented by each curve. The Devine prediction is considered to offer reasonable approximation to the body weights of mature adults (aged >18 years) who achieved good longevity [16, 30]. In contrast, the population median represents healthy subjects during growth up to the age of 20 years, surveyed between 1963 and 1994 in the CDC surveys [22]. It seems likely that the growing children and young adults in the population data might not have ‘filled-out’ as much as the healthy older adults of similar height. It is not known if the younger reference population had less muscle or less fat or lighter bones than the Devine population, but the distinction may not be important from the point of view of ventilatory demand. Recent magnetic resonance imaging (MRI) research in adults has revealed that resting energy expenditure is not uniform through all fat-free (lean) mass, but is dominated by highly metabolically active organs compared with bone mass or skeletal muscle [31]. If it is reasonable to assume that organ mass of a growing individual is similar to a mature adult of the same height, the (leaner) population median data may offer a better weight-from-height prediction for the purposes of predicting ventilatory demand than that of a mature adult of similar height. Also of relevance is that as lean body mass increases with height, the metabolically dominant proportion of that mass (organs) does not increase in direct proportion [31]. Hence resting metabolic demand may not be linearly proportional to lean body mass (height) predictions. If so, the use of lean body mass to infer resting ventilatory demand may have some inherent inaccuracy, which will be emphasized in models with steeper mass-per-height gradient, such as the Devine predictions.

Even if lean body mass was adequately correlated with ventilatory demand, in pediatric patients it remains to be clinically assessed whether using lean body mass for setting tidal volume might minimize volutrauma. Respiratory system compliance (C_rs) may be correlated with height [28, 29, 32], but this relationship is highly non-linear. It may be speculated that if the non-linear relationship between lean body weight and height was of similar nature to the non-linear relationship between respiratory compliance and height, then scaling tidal volumes to lean body mass in pediatric ventilation may indeed minimize volume trauma throughout the height range. Such a comparison is presented in Fig. 6, using predictive formulae for normal respiratory system compliance [32] compared to population median body weight. Coarse correlation is evident throughout the height range, with enough similarity to suggest crude proportionality between lean body weight and C_rs. If so, within normal lungs, the concept of scaling tidal volume to lean body weight may be sympathetic not only to ventilatory demand, but also to appropriate driving pressures. However it must be emphasized that the height-based associations discussed here derive from healthy subjects, not critically ill individuals with injured lungs. Clinical evaluation is needed to determine whether a tidal volume scaled to lean body weight predicted from height can improve mortality while also delivering adequate ventilation.

It is also accepted that any theoretical ‘improvement’ to PBW models may be irrelevant to adult ventilation if the clinical practice built around the ARDSNet findings must be strictly adhered to (i.e. that of basing initial tidal volumes on the Devine body weight predictions, multiplied by 5–8 mL/kg). Even so, for statures outside the ‘proven’ region of the Devine formulae (below 5 ft/152 cm), there may be scope for weight predictions based on population median data.

The curve fitting employed here minimizes relative error (expressed as a proportion of the reference body weight), rather than absolute error as typical in traditional least-squares regression curve fitting. The intent is that the weight predictions maintain fidelity even at the smallest body sizes. As a result, these PBW models may be considered if applying the lung-protective ventilation strategy to children. Across all models, a common piecewise curve is employed up to a height of 105 cm. The PBW models diverge at larger body sizes, reflecting the contextual nature of what might be considered the optimal ‘reference’.

The PBWmf + MBW curves (Fig. 2, Table 1) are entirely consistent with prevailing acute ventilation practice, in that they fully adhere to the male and female Devine formulae, and in fact extend them down to around 130 cm (they were traditionally considered valid above 5 ft/152 cm [15, 16]). Below 130 cm, the PBWmf + MBW curves merge to fit the population median data. The contribution of this model is primarily that of extending down to the smallest body sizes without disturbing current ventilation practice. However, the value of retaining sex-specific body weight predictions may be questioned when other sources of variation embodied in the final tidal volume calculation are considered (see later).

The PBWu + MBW curve (Fig. 3, Table 2) removes patient sex from the model in the interests of simplification. This comes at the expense of a +4.9% (female) or −4.6% (male) deviation from the established PBW formulae across the region where these are generally considered valid (above 5 ft/152 cm). Of the three unisex PBW models, this offers the least percent error relative to both PBW Male and PBW Female. Therefore, this curve may be useful for applications where close conformance to current practice is important but specifying sex is impractical or inconvenient. Yet it prompts the question: how much discrepancy in tidal volume due to deviating from the ‘proven’ PBW formulae might result in a clinically-relevant change in outcome? A precise answer to this is unknown, but we can put it in the context of other sources of error.

One source of error is the accuracy of estimating height. True height can vary throughout normal daily activity by up to 2% [33, 34], which might be considered a baseline accuracy. In the hospital setting, standard methods of estimating patient height include: asking the patient (not always possible), dedicated height measurement devices, measuring height/length in-situ using a tape, estimating height from a more convenient body part, and basic visual estimation. Visual estimation is common, leading to average PBW errors of 10% in one adult study, while the best-performing height estimation method (the Chumlea knee-heel approach) led to an average PBW error of 5.8% [35].

Other contributors to variation in eventual tidal volume also need to be taken into account. A common protective tidal volume recommendation is 6–8 mL/kg_PBW, offering a discretionary variation of 25–33%. This scaling factor was found to be protective when compared to 10–15 mL/kg_PBW, but it is less clear how protective such a scaling factor is compared to intermediate volumes (8–10 mL/kg_PBW). Furthermore, the major trials tended to compare two clinical’approaches’, rather than just two different tidal volume factors [4]. More fundamentally, recent analysis suggests that tidal volume may be less critical than driving pressure within a heterogeneous lung, emphasizing the importance of all elements of the lung-protective ventilation bundle rather than just tidal volume [2].

As a worst-case error contribution, the use of actual body weight for volume setting is still widespread in adults. Use of actual body weight in average height adults can result errors of +30% relative to established PBW, and more than +35% for shorter statures [35]. Even in centers where lung-protective ventilation is accepted, adults of short stature may be less likely to receive volumes consistent with protective recommendations [10].

In the context of the above observations, sex-specific weight predictions may be of questionable value to ultimate tidal volume selection. A unisex simplification to existing formulae would be a variation of at most 7% predicted weight (very short female). More generally, systematic deviations of 5–10% from the established PBW formulae may ultimately have minimal impact on initial tidal volume settings when considered amongst all other sources of clinical variation, particularly the ml/kg discretion and height estimation. Such generalizations are not intended to diminish the importance of setting safe tidal volumes during initial ventilator configuration, but rather to probe the ‘evidence’ supporting the established PBW formulae, in the interests of simplicity. As eloquently stated by Linares-Perdomo et al. in their adult PBW standardization proposal [15], “While it is not possible to identify a “true” or “correct” PBW, it is possible to choose a reasonable PBW equation that will eliminate this source of unwarranted variation in clinical research and practice”.

The PBWuf + MBW curve (Fig. 4, Table 3) also provides simplification while retaining consistency with the ARDSNet framework. In this case the single curve adheres to PBW Female formula, providing better alignment with lean body weight, while its adoption for males would result in under-volume rather than over-volume compared to the status quo. For a male patient, the result would be at most 10% less volume than if the PBW Male formula was used (at 5 ft/152 cm), or 6% less volume at an average male height (a discrepancy fully compensated for in tidal volume by a <0.5 mL/kg increase). Note that both male and female adult patients would receive volumes larger than if derived from population median weight. The PBWuf + MBW model is proposed for consideration as a standardized unisex PBW formula. It is offered as a practical compromise between simplification & conservative interpretation of ARDSnet practice, while also better reflecting adult lean body weight than established PBW formulae.

The MBW curve (Fig. 5, Table 4) offers an easily calculated indication of median population data which – if viewed in isolation – make it seem ideally suited to tidal volume titration in lung-protective ventilation. Compared to using the PBW Male formulae, direct replacement with the MBW curve would result in at most a 20% reduction in initial tidal volume, or 10% reduction compared to PBW Female. This lower volume would be fully compensated by an upward adjustment of less than 1 mL/kg_PBW. So MBW may also be considered for lung protective ventilation, if complete departure from the established Devine formulae was contemplated.

This analysis has a numerous limitations. It is emphasized that the PBW models presented here are specific to lung protective ventilation, and are not appropriate for pharmacology or assessment of healthy body weight. The focus here is on lean body weight, with up to 10% underestimation tolerated. Whereas in healthy body weight assessment, adopting a median weight for a given height has been judged inappropriate, and instead use of age- and sex-specific BMI is recommended [23, 36‐38]. It may also be questioned if median weight of modern populations should be used as a surrogate for lean body weight, given that increasing obesity may affect median values, particularly later in development. The WHO data sampled culturally and ethnically diverse populations, while the CDC population included growing children surveyed over 2 decades ago. Most importantly, the population median weights within the adult range were substantially leaner than those predicted by the prevailing relationships used in protective ventilation (i.e. Devine’s formulae describing healthy mature adults [19]). This suggests that the median population reference may be a better representation of lean body weight than the established PBW relationship. Another limitation is that direct height-weight data were not available at all statures, so age-based data were used to synthesize weight-from-height over these ranges. The age-mapping applied is equivalent to that of the McLaren method, which has 2 main limitations identified [23, 37, 38]: (1) it does not recognize age-related variation, of importance to nutrition assessment but less relevant to lean body weight estimation, and (2) it cannot offer prediction above the tallest median height, which is resolved in the MBW model by linear extrapolation at taller heights. The notion of mapping 50^th centile height to 50^th centile weight is broadly accepted, in contrast to doing so further from the median (as inherent to the Moore estimation method) [37, 38]. It can be seen in Fig. 1 that the median weights derived from each of the 4 data sets are in good alignment, despite the mix of direct and indirect height-weight information. Four nutritional studies [23, 36‐38] directly compared weight-for-stature and BMI-for-age, and found correlation moderately good around the 50^th centile. Two of these studies included small children, with one concluding the 2 approaches are similar below the age of 8 years [23], while the other found agreement to be poorer at age 4–5 years than at younger ages [36]. Any such weakness here is alleviated by the WHO 45–110 cm data set, which provides direct weight-from-height data extending to heights corresponding to a 4–5 year age. Finally, the clinical robustness of using predicted weight rather than actual weight in infants may be questioned: will any potential safety benefits be confounded by the particular challenge of measuring length in babies, or the complication of the prediction process? The analysis presented here does not touch on this question, but simply defers to the recent consensus recommendation [6] in favor of using predicted body weight in pediatric ventilation, and offers a means of doing so without reference to growth charts.