Original Paper Neonatology 2017;111:376–382 DOI: 10.1159/000455036 Received: September 29, 2016 Accepted after revision: December 12, 2016 Published online: February 1, 2017 Three-Dimensional Imaging-Based Web Application for Predicting Tracheal Tube Depth in Preterm Neonates Raksa Tupprasoot a Dean Langan b J. Ciaran Hutchinson a, b Hannah Barrett a, b Michael R.J. Sury a, b Owen J. Arthurs a, b a Great Ormond Street Hospital for Children NHS Foundation Trust, and b UCL Great Ormond Street Institute of Child Health, University College London, London, UK Abstract Background: Positioning a tracheal tube (TT) to the correct depth in preterm infants is challenging. Currently, there is no reliable single-predictor model for neonates applicable to the whole range of size or age. Objective: In this study, we used post-mortem magnetic resonance imaging (PMMRI) of preterm infants to measure tracheal dimensions and to develop a clinical guide for TT positioning. Methods: We measured tracheal length (TL) and tracheal diameter (TD) in a cohort of normal neonates and foetuses that underwent PMMRI (cause of death unexplained). The distance between the lips and the mid-tracheal point, i.e., the mid-tracheal length (mid-TL), and the TD measurement were obtained. We produced univariate prediction models of mid-TL and TD, using gestational age (GA), foot length (FL), crown-rump length (CRL) and body weight (BW) as potential predictors, as well as multiple prediction models for mid-TL. Results: Tracheal measurements were performed in 117 cases, with a mean GA of 28.8 weeks (range 14–42 weeks). The best linear association was between mid-TL and FL (mid-TL = FL × © 2017 S. Karger AG, Basel E-Mail email@example.com www.karger.com/neo 0.914 + 1.859; R2 = 0.94), but was improved by multivariate regression models. We developed a prediction tool using only GA and BW (R2 = 0.92), and all four predictors (GA, BW, FL and CRL; R2 = 0.94) which is now available as a web-based application via the Internet. Conclusion: Post-mortem imaging data provide estimates of TT insertion depth. Our prediction tool based on age and BW can be used at the bedside and is ready to be tested in clinical practice. © 2017 S. Karger AG, Basel Introduction Insertion of a tracheal tube (TT) to the correct depth is essential to avoid misplacement into the bronchus or the pharynx, and, the smaller the infant, the more challenging this becomes. Ideally, the TT tip should be placed at the mid-point between the larynx and carina and although its position can be checked by chest X-rays , repositioning is frequently necessary . Methods used to investigate the correct or ideal TT depth have involved either imaging with conventional chest radiographs or post-mortem (PM) autopsy [3–6], and several formulae or rules have been published to help predict the safe insertion depth accurately. Studies have Dr. Owen J. Arthurs, Consultant Radiologist Great Ormond Street Hospital for Children NHS Foundation Trust London WC1N 3JH (UK) E-Mail owen.arthurs @ gosh.nhs.uk Downloaded by: University of Missouri-Columbia 22.214.171.124 - 10/26/2017 9:53:51 PM Keywords Airway · Trachea · Post-mortem magnetic resonance imaging · Foetus Color version available online Fig. 1. Airway measurements. Example of multiplanar reconstruction (MPR) of PMMR sequence and tracheal measurements – from mouth to carina (left and centre, top and bottom rows), mouth to epiglottis (right, top, and bottom rows). shown that airway length and tube insertion depth have linear relationships with body weight (BW) , gestational age (GA) , foot length (FL) , and size such as crown-rump length (CRL) , or crown-heel length [4, 10]. The European Resuscitation Council has recommended that the estimation of TT depth should be based on GA  even though, in practice, the difference between BW and GA may not be appreciable . However, most of the published studies have involved too few infants born very preterm (28–32 weeks) or extremely preterm (<28 weeks), and the relationships change or become non-linear when infants 1 kg are included [8, 13, 14]. Currently, there is no reliable single-predictor model for neonates applicable to the whole range of size or age. Modern, 3-dimensional (3D), cross-sectional imaging can be used to measure airway and tracheal dimensions, and is more accurate than simple 2D chest radiography. 3D imaging of airway structures is only rarely indicated in live, preterm infants, but since recently, PM magnetic resonance imaging (PMMRI) is being used routinely to investigate the cause of death . Our institutional autopsy imaging database provides 3D data on the airway dimensions in a wide range of foetuses and neonates, and it could prove useful to develop a mathematical model for the bedside. Furthermore, our database includes foetuses younger than 22 weeks gestation that, although being too preterm to survive, may be of future interest. The aim of this study was to use 3D-detailed anatomy derived from foetal PMMRI to measure airway parameters, and also to develop a bedside mathematical tool to predict optimal TT insertion depth. 3D MRI Application for Predicting TT Depth Neonatology 2017;111:376–382 DOI: 10.1159/000455036 Methods 377 Downloaded by: University of Missouri-Columbia 126.96.36.199 - 10/26/2017 9:53:51 PM Recruitment and Criteria We evaluated PMMRI of all foetuses (miscarriages and stillbirths) with a GA of <44 weeks GA referred to our institution from February 2012 to September 2015. Ethical approval was obtained Table 1. Summary of demographic details Mean (SD) Median Min Max IQR Predictors Gestational age, weeks (n = 117) Foot length, cm (n = 117) Crown-rump length, cm (n = 116) Body weight, kg (n = 117) 28.8 (8.3) 5.4 (2.1) 25.9 (8.4) 1.50 (1.3) 26.0 4.9 24.2 0.72 14.0 1.5 10.0 0.05 42.0 9.0 41.4 4.8 21.0 – 37.0 3.6 – 7.7 18.4 – 34.2 0.38 – 2.65 Tracheal dimensions Upper airway length, cm (n = 117) Total airway length, cm (n = 117) Mid-tracheal length, cm (n = 117) Tracheal diameter, cm (n = 58) 8.1 (2.4) 5.6 (1.6) 6.8 (2.0) 0.28 (0.102) 7.4 5.3 6.4 0.29 3.3 2.4 2.8 0.09 12.7 9.0 10.8 0.51 6.1 – 10.4 4.3 – 7.1 5.2 – 8.8 0.20 – 0.38 Table 2. Univariate linear models of mid-TL Adjusted R2 Predicted below carina Predicted above glottis 0.293 + 0.227 × GA 1.859 + 0.914 × FL 0.844 + 0.231 × CRL 4.72 + 1.403 × BW 7.001 + 1.740 × loge (BW) 0.89 0.94 0.93 0.86 0.91 9 (7.7%) 3 (2.6%) 5 (4.3%) 10 (8.5%) 4 (3.4%) 3 (2.6%) 2 (1.7%) 3 (2.6%) 7 (6.0%) 8 (6.8%) for the analysis of PMMRI, and written informed consent was obtained from the parents. The bodies were stored in a mortuary at 4 ° C until the examination. Cases were excluded if the airway was abnormal on either PMMRI or subsequent autopsy, or where image quality was inadequate to permit measurements. Demographic data acquired from the clinical notes included GA, BW, FL, and CRL. Magnetic Resonance Imaging Imaging was performed on a 1.5-T scanner (Avanto, Siemens Medical Solutions, Erlangen, Germany) with a conventional phased array head coil. Conventional 3D T1- and T2-weighted sequences were examined by a paediatric radiologist for clinical purposes . T2-weighted isotropic sequences of the head and chest were used to create 3D multiplanar (sagittal, coronal, and axial) datasets. Tracheal Measurements Reformatted images (Fig. 1), using a Centricity Web DX Viewer (Centricity WebPACS system, 2006; GE Healthcare, Chalfont St Giles, UK) were used to measure and calculate the following: • Upper airway length (uAL) = the distance from the lips to the glottis. The position of the glottis was defined as that part of the airway at the level of C5/C6 intervertebral disc space because this has a close relationship with the cricoid cartilage • Total airway length (totAL) = the distance from the lips to the carina 378 Neonatology 2017;111:376–382 DOI: 10.1159/000455036 • Tracheal length (TL) = totAL – uAL • Mid-tracheal length (mid-TL) = the distance between the lips and the mid-tracheal point, and calculated as uAL + 1/2 TL; this is equivalent to TT depth • Internal luminal tracheal diameter (TD) was measured at the mid-tracheal point All measurements were made to the nearest millimetre by a single observer (R.S.). Twenty datasets were selected at random, and a second observer (O.J.A.) repeated the measurements, so as to assess the inter-observer variability. Statistical Analysis Univariate linear regression models were fitted for both outcome variables (mid-TL and TD) using 4 predictors (GA, BW, FL, and CRL). Two multivariate regression models were fitted for midTL, using (1) the 2 most readily available predictors (GA and BW) and (2) all 4 predictors. These prediction models were developed into a web application for clinical practice. For each regression model, subjects were identified in whom the model would have predicted a mid-TL that would have resulted in a TT being inserted either not deep enough or too deep (i.e. the TT tip would be above the glottis or below the carina). Bland-Altman limits of agreement were calculated to describe the inter-observer variability of mid-TL and, using a regression approach , to account for a relationship between variability and the mid-TL itself. All analyses were carried out in R v3.3.0. Tupprasoot/Langan/Hutchinson/Barrett/ Sury/Arthurs Downloaded by: University of Missouri-Columbia 188.8.131.52 - 10/26/2017 9:53:51 PM GA, weeks (n = 117) FL, cm (n = 117) CRL, cm (n = 116) BW, kg (n = 117) Loge BW, kg (n = 117) Formula 10 8 6 4 2 10 15 20 25 30 35 40 Color version available online Airway to mid-trachea length, cm Airway to mid-trachea length, cm 12 12 10 8 6 4 2 45 2 4 12 10 8 6 4 2 10 15 20 25 30 35 Crown-to-rump length, cm 6 8 Foot length, cm Airway to mid-trachea length, cm Airway to mid-trachea length, cm Gestational age, weeks 40 12 10 8 6 4 2 0 1 2 3 Post-mortem weight, kg 4 5 Tracheal measurements were performed in 117 foetuses (mean GA 28.8 weeks, range 14–42 weeks; 17 foetuses were <22 weeks; Table 1). The smallest weighed only 50 g, and had a CRL of 10 cm. Mid-TL ranged between 2.8 and 10.8 cm (Table 1). All predictor variables had a strong linear relationship with mid-TL. FL had the highest adjusted R2 of 0.94 (Table 2) and produced the fewest predictions of TT tip positioning below the carina (3; 2.6%) or above the glottis (2; 1.7%; Table 2; Fig. 2). BW had the lowest adjusted R2 of 0.86, but our results suggest that this may be because of a non-linear relationship, particularly at a low BW (log transformation R2 = 0.91; Fig. 2). The multivariate regression model using all 4 predictors had only a marginally better fit than the multivariate model with only GA and BW (an adjusted R2 of 0.94 and 0.92, respectively; Table 3). Formulae for these models were made accessible through a web-based application (https://chpredict.shinyapps.io/shinyapp/; Fig. 3). TD was only measurable in 58 (50%) of the foetuses. Univariate prediction models for TD all had an adjusted R2 of 0.51–0.53 and multivariate regression modelling was not undertaken. 3D MRI Application for Predicting TT Depth Neonatology 2017;111:376–382 DOI: 10.1159/000455036 Results 379 Downloaded by: University of Missouri-Columbia 184.108.40.206 - 10/26/2017 9:53:51 PM Fig. 2. Scatter plots of mid-TL against the 4 predictor variables, GA (top left), FL (top right), CRL (bottom left), and PM weight (bottom right). Regression lines (from Table 2) are plotted in red. Vertical lines represent absolute TL in each case, and those in red represent where predicted mid-TL falls outside of this range. Color version available online Fig. 3. Screenshot of web-based applica- tion. Formulae for both multiple prediction models of airway to mid-tracheal length are currently accessible through a web-based application (https://chpredict. shinyapps.io/shinyapp/). Table 3. Multivariate linear models of mid-TL Predictors All predictors (n = 116) 0.94 (constant) GA loge (BW) FL CRL GA and BW (n = 117) 0.92 (constant) GA loge (BW) β (SE) 95% CI for β p value Predicted below carina above glottis 1.752 (0.693) –0.026 (0.025) 0.045 (0.205) 0.642 (0.141) 0.090 (0.032) 0.379 to 3.125 –0.075 to 0.023 –0.362 to 0.452 0.362 to 0.921 0.027 to 0.154 2.529 –1.047 0.217 4.550 2.814 0.0129 0.2974 0.8283 <0.0001 0.0058 3 (2.6%) 3 (2.6%) 4.350 (0.656) 0.090 (0.022) 1.088 (0.168) 3.051 to 5.650 0.046 to 0.133 0.755 to 1.421 6.632 4.055 6.474 <0.0001 <0.0001 <0.0001 3 (2.6%) 5 (4.3%) The variability (agreement between observers) of the mid-TL increased as it increased; the 95% limits of agreement were ±0.25 and ±0.75 cm for an average mid-TL between observers of 4 and 10 cm, respectively. Discussion We used 3D foetal PMMRI images to measure midTL, in order to produce a mathematical model to help predict the ideal depth of TT insertion in preterm infants. 380 t statistic Neonatology 2017;111:376–382 DOI: 10.1159/000455036 Our best model to predict mid-TL uses 4 clinical variables, but a model using only GA and BW was almost as good. TD was not easily or accurately measured due to this being so small. Other investigators have used PM foetuses and neonates to measure ideal TT insertion depth. In 2001, Embleton et al.  dissected 39 specimens ranging from 24 to 43 weeks of post-menstrual age, and they showed that FL was a much better predictor of TT depth (R2 = 0.79) than BW (R2 = 0.67) and age (R2 = 0.58). Neonatal body dimensions, however, such as FL and CRL, are neither Tupprasoot/Langan/Hutchinson/Barrett/ Sury/Arthurs Downloaded by: University of Missouri-Columbia 220.127.116.11 - 10/26/2017 9:53:51 PM Adjusted R2 routinely measured at birth nor readily achievable in an emergency intubation setting. A prediction model combining body dimensions with BW and GA may be slightly more accurate, but is less practical in a clinical situation than a model using GA and BW only. Previous studies on live infants have developed formulae based on age and weight. The 7-8-9 Rule used BW to estimate TT insertion depth, defined as the distance from the lips to the level of the first or second thoracic vertebra on a chest radiograph . The formula derived was length = 1.17 × BW + 5.58, which approximates to 6 + each kilogram of body weight: this produces a TT depth of 7 cm for infants weighing 1 kg, 8 cm for those weighing 2 kg, and 9 cm for those weighing 3 kg. The data available in this study from infants weighing <1 kg, however, was sparse, and Peterson et al.  also reported that the formula gave TT depths that were too long in preterm infants <750 g. An internet tool (currently available at http://www.nicutools.org/) uses the formula TT depth (cm) = 1.1 × BW + 6.1, but only for infants >1 kg. For smaller infants, the TT depth is 5.5 cm for a BW <500 g, 6 cm for 550–700 g, and 6.5 cm for 700–999 g. Kempley et al.  reported that TT depth was not linearly related to BW, and that estimates based on GA reduced the need for TT repositioning. We also found that GA is not linearly related to mid-TL, especially in the smallest foetuses. Nevertheless, a clinical study randomising neonates to a TT depth, based on either GA or BW, suggested that there is no appreciable difference  and that neither predictor was reliable at achieving satisfactory positioning, with BW (the 7-8-9 rule) successful in only 25 of 49 infants (51%) and GA successful in 16 of 41 infants (39%) . In the light of these findings, our data and model may help to better predict the TT depth. Firstly, our data is based on 3D anatomy of tracheal and airway measurements from MRIs, rather than 2D radiographic imaging using vertebral body height as the reference level for the trachea. Secondly, we provide new high-quality data in the <22-week-old group, which increases the confidence in the mathematical model to predict mid-TL for potentially viable infants of 23–25 weeks of GA. Thirdly, our data supports the clinical findings of other studies that any single predictor of TT depth is not as reliable as a combination of predictors. Fourthly, by incorporating all our data, we have made a web-based application available which may be useful at the bedside. Whether using the data in this study improves TT placement accuracy remains to be determined in the appropriate clinical setting. The main limitation of our study is that we did not measure the effect of the position of the head and neck. Neck extension is known to lengthen the trachea [18–20], and so imaging in a defined neutral position would provide the most reliable predictions. There are physiological changes that occur after death, which may mean that our measurements were different from those for live infants. The trachea may be shorter PM because the diaphragm applies less traction , and therefore our formula may underestimate the mid-TL and TT insertion depth for live infants. Collapse of the upper airway in a dead infant may account for a small degree of measurement error, and this was most evident when we attempted to measure TD. We also acknowledge that there is very little published data relating the accuracy of PM imaging measurements to real life. Nevertheless, the inter-observer variation was small and our measurements were repeatable. Our TT insertion depths were also made to the nearest millimetre, but clinicians may not be able to achieve an accuracy of more than to the nearest 0.5 cm; we recommend rounding up or down appropriately. We look forward to testing our formula in clinical practice and potentially improving it with additional PM imaging and clinical data. 3D MRI Application for Predicting TT Depth Neonatology 2017;111:376–382 DOI: 10.1159/000455036 Conclusion PM imaging data provides reproducible anatomical measures of TL in order to predict ideal TT insertion depth. We have provided an easy-to use-internet application which may be used at the bedside to improve TT placement. This tool still has to be validated in clinical practice. Acknowledgement O.J.A. was funded by an NIHR Clinician Scientist fellowship award (NIHR-CS-012–002) and the Great Ormond Street Hospital Children’s Charity. This work was undertaken at GOSH/ICH, UCLH/UCL who received a proportion of funding from the UK Department of Health’s NIHR Biomedical Research Centre funding scheme. This article presents independent research funded by the NIHR. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR, or the Department of Health. Disclosure Statement 381 Downloaded by: University of Missouri-Columbia 18.104.22.168 - 10/26/2017 9:53:51 PM The authors have no conflicts of interest or financial relationships relevant to this article to disclose. Author Contributions Dr. Tupprasoot helped to co-design the study, performed the literature search, carried out the data analysis, and drafted the initial manuscript. Dr. Langan performed the statistical analysis for the study, and critically reviewed the manuscript. Dr. Hutchinson carried out the data analysis and critically reviewed the manuscript. Dr. Barrett carried out the data analysis and critically reviewed the manuscript. Dr. Sury co-designed the study, performed the literature search, and critically reviewed the manuscript. Dr. Arthurs co-designed the study, supervised the data collection, and critically reviewed the manuscript. References 382 8 Kempley ST, Moreiras JW, Petrone FL: Endotracheal tube length for neonatal intubation. Resuscitation 2008;77:369–372. 9 Embleton ND, Deshpande SA, Scott D, Wright C, Milligan DWA: Foot length, an accurate predictor of nasotracheal tube length in neonates. Arch Dis Child Fetal Neonatal Ed 2001;85:F60–F64. 10 Loew A, Thibeault DW: A new and safe method to control the depth of endotracheal intubation in neonates. Pediatrics 1974; 54: 506– 508. 11 Wyllie J, Bruinenbergb J, Roehrd GC, Rüdigerf M, Trevisanutoc D, Urlesbergerga B: European Resuscitation Council Guidelines for Resuscitation 2015 Section 7. Resuscitation and support of transition of babies at birth. Resuscitation 2015;95:249–263. 12 Flinn AM, Travers CP, Laffan EE, O’Donnell CP: Estimating the endotracheal tube insertion depth in newborns using weight or gestation: a randomised trial. Neonatology 2015; 107:167–172. 13 Peterson J, Johnson N, Deakins K, WilsonCostello D, Jelovsek JE, Chatburn R: Accuracy of the 7-8-9 Rule for endotracheal tube placement in the neonate. J Perinatol 2006;26:333– 336. Neonatology 2017;111:376–382 DOI: 10.1159/000455036 14 Amarilyo G, Mimouni FB, Oren A, Tsyrkin S, Mandel D: Orotracheal tube insertion in extremely low birth weight infants. J Pediatr 2009;154:764–765. 15 Arthurs OJ, Bevan C, Sebire NJ: Less invasive investigation of perinatal death. BMJ 2015; 351:h3598. 16 Norman W, Jawad N, Jones R, Taylor AM, Arthurs OJ: Perinatal and paediatric postmortem magnetic resonance imaging (PMMR): sequences and technique. Br J Radiol 2016;89:20151028. 17 Bland JM, Altman DG: Measuring agreement in method comparison studies. Stat Methods Med Res 1999;8:135–160. 18 Donn SM, Blane CE: Endotracheal tube movement in the preterm neonate: oral versus nasal intubation. Ann Otol Rhinol Laryngol 1985;94:18–20. 19 Todres ID, deBros F, Kramer SS, Moylan FM, Shannon DC: Endotracheal tube displacement in the newborn infant. J Pediatr 1976;89: 126–127. 20 Rost JR, Frush DP, Auten RL: Effect of neck position on endotracheal tube location in low birth weight infants. Pediatr Pulmonol 1999; 27:199–202. 21 Fearon B, Whalen JS: Tracheal dimensions in the living infant (preliminary report). Ann Otol Rhinol Laryngol 1967;76:965–974. Tupprasoot/Langan/Hutchinson/Barrett/ Sury/Arthurs Downloaded by: University of Missouri-Columbia 22.214.171.124 - 10/26/2017 9:53:51 PM 1 Blayney MP, Logan DR: First thoracic vertebral body as reference for endotracheal tube placement. Arch Dis Child Fetal Neonatal Ed 1994;71:F32–F35. 2 Mainie P, Carmichael A, McCullough S, Kempley ST: Endotracheal tube position in neonates requiring emergency interhospital transfer. Am J Perinatol 2006;23:121–124. 3 Kuhns LR, Poznanski AK: Endotracheal tube position in the infant. J Pediatr 1971;78:991– 996. 4 Rotschild A, Chitayat D, Puterman ML, Phang MS, Ling E, Baldwin V: Optimal positioning of endotracheal tubes for ventilation of preterm infants. Am J Dis Child 1991;145: 1007–1012. 5 Thayyil S, Nagakumar P, Gowers H, Sinha A: Optimal endotracheal tube tip position in extremely premature infants. Am J Perinatol 2008;25:13–16. 6 Kemper M, Dullenkopf A, Schmidt AR, Gerber A, Weiss M: Nasotracheal intubation depth in paediatric patients. Br J Anaesth 2014;113:840–846. 7 Tochen ML: Orotracheal intubation in the newborn infant: a method for determining depth of tube insertion. J Pediatr 1979; 95: 1050–1051.