The Hilar Height Ratio in Normal Chest Radiographs of Adults Seen at Selected Tertiary Hospitals in Kampala, Uganda; A Forgotten Parameter in Pulmonary Medicine

Background: The hilar height ratio (HHR) is a numerical expression of the hila position. Displacement of the pulmonary hilum is the most reliable indirect sign, in conditions that result in pulmonary volume changes. Despite the high utility of the chest radiograph, it has a relatively low diagnostic accuracy, high interobserver disagreement and numerous errors following interpretation. Routine use of the HHR with knowledge of the normal ranges would, therefore, help improve the overall chest radiograph sensitivity, reduce on errors and the interobserver disagreement in the interpretation of pulmonary volume changes. This study aims to determine the HHR in our study population, compare it with previous studies, and to relate it with sex, stratied age groups, height and body mass index (BMI). Methods: A consecutive cross-sectional study with purposive sampling were used to lter out a total of 384 normal chest radiographs of adults seen from three tertiary hospitals in Kampala, Uganda. The right and left HHRs were evaluated for each chest radiograph, along with the age, sex weight, height and BMI. Results: The median right HHR for the participants was 1.42 (IQR = 1.31- 1.57), and the median left HHR was 0.92 (IQR = 0.86- 0.98). There was a signicant mean difference in the right HHR between the sex groups (p = 0.017) and age groups (p = 0.001). The mean difference in the left HHR was not affected by sex (p = 0.178) or age (p = 0.198). The right and left HHRs showed a very weak correlation to height (r = -0.10 and r = 0.08, respectively). The right and left HHRs were not signicantly different among the BMI groups (p = 0.254 and 0.20 respectively). Conclusion: The median left HHR was not affected by sex, age, weight and height while the right HHR varied with sex and age in our study population. Weight and height do not affect HHRs. The females in our study population had a higher positioned right hilum, whereas the position of the right hilum progressively lowered with increasing age.


Background
In Uganda, chest radiography utility remains high despite the increasing availability of chest computed tomography and accounts for 30 to 40% of all x-ray examinations done at all levels of health-care delivery (1). Despite this, the diagnostic accuracy of chest radiography is relatively low with an overall sensitivity of 49% (95% CI, 40-58%) and speci city of 92% (86-95%) (2). In addition, interobserver disagreement in radiograph interpretation ranges from 0.3% to 58 % (3,4) with an estimated interpretation error of up to 49% (5). The chest radiograph also contains hidden areas that pose challenges in their evaluation due to the presence of overlying structures such as the lung apices, retrocardiac and inferior lung bases (6). These hidden areas contribute to the most commonly missed ndings and result in frequent false-negative or false-positive interpretations (7).
Displacement of the pulmonary hilum is the most reliable indirect sign found in many chest pathologies that result in alterations of pulmonary volume (8) as was originally described by Robbins and Hale (9) and later by Lubert and Krause (10). The hilar height ratio (HHR) is a numerical expression of the hila position (11) that helps to provide a standardised way to quantify pulmonary volume changes. It allows recognition of a hilar positional abnormality without the necessity for comparison of both sides (12).
When an abnormal HHR is present in the absence of pulmonary disease, then infrapulmonary and subphrenic disease should be suspected. An abnormal HHR is not always an absolute sign of disease, but it should cause the radiologist to scrutinise the radiograph carefully for underlying pathology (12).
The routine use of the HHR with knowledge of the normal ranges would, therefore, help improve the overall chest radiograph sensitivity, reduce on errors and the interobserver disagreement in the interpretation of pulmonary volume changes (3,8

Study population
Normal chest radiographs (Posterior anterior view, erect) of asymptomatic adults who came for routine screening examination as interpreted by the principal investigator and con rmed by an independent radiologist were used in this study. Purposive sampling was used to lter out a total of 384 normal chest radiographs of adults collected consecutively from all three sites in proportions based on the average normal chest radiographs seen per facility over three months. An adult was taken to be a person 15 years of age and above to allow for comparison with previous studies previous studies done in Kenya (11), United states of America (USA) (12) and south Korea (13) which used the same cut-off. A normal chest radiograph was de ned as one that ful lled all the quality requirements and had clear lung elds with no pulmonary or extrapulmonary abnormalities as interpreted by the principal investigator and con rmed by an independent radiologist. All chest radiographs included in the study were of non-symptomatic adults undergoing routine medical screening.

Sample size
The ideal sample size for this study was estimated using the Kish Leslie formula given below: Where; n = sample size z = level of con dence according to the standard normal distribution (for a level of con dence of 95%, z = 1.96) p = estimated proportion of the population that presents the characteristic (when unknown we use p = 0.5) d = tolerated margin of error (for example we want to know the real proportion within 5%) To calculate a proportion with a 95% level of con dence and a margin of error of 5% we obtain n = (1.96)2 / 4(0.05)2 = 384

Study Procedure
A semi structured data collection tool was used to collect bio-demographics and relevant history (Appendix 1). The principal investigator collected data with the help of two quali ed radiographers trained as research assistants. The research assistants identi ed the adults sent for routine screening and ensured that the radiographs were well taken and digitally stored. Purposive sampling was used to lter out those with normal chest radiographs to be included in the study as interpreted by the principal investigator and con rmed by an independent radiologist.
The HHR was calculated by drawing a straight line parallel to the thoracic spine from the highest point of the pulmonary apex to the diaphragm and an intersecting line from the midpoint of the hilum perpendicular to the vertical line. The distance from the lung apex to the hilum was then divided by the distance from the hilum to the diaphragm as suggested by Homer (12) (Figure 1).

Data analysis
The HHR was evaluated on the right and left side for each chest radiograph, and this information was recorded along with the age, sex weight, height and BMI. The nal results were analysed using STATA version 16. The socio-demographic and clinical characteristics were summarised using frequencies and percentages. The right and left hilar heights were normally distributed and hence summarised using mean and standard deviation whereas the right and left hilar height ratios were skewed and hence summarised using median and interquartile ranges.
The Wilcoxon rank sum test was used to test the mean difference in the left and right hilar height ratios between the sex groups. The Kruskal Wallis test was used to test the mean difference between the left and right hilar height ratios and the different age groups and BMI groups. The Spearman correlation coe cient was used to determine the linear relationship between right and left hilar height ratios and height of the participant.

Discussion
This study sought determine the HHR in our study population, compare it with previous studies, and to relate it with sex, strati ed age groups, height and BMI. Our study had the highest values of central tendency for both the right and left hilar height ratio. The values of the right and left hilar height ratio for our study compared favourably with the original study done by Homer but with a lower level of variability (12).
The differences in hilar height ratio values in our study compared to previous studies done in Kenya (11), United states of America (USA) (12) and south Korea (13) ( Table 4) could be explained by the difference in demographic characteristics, difference in sample sizes and the skewed male to female ratios. Because mean was used as the measure of central tendency in the three previous studies, it is assumed that the data was normally distributed, and in normally distributed data, mean and median are equal (14); therefore, our data's median values can be compared to the means in the previous studies.
There was a signi cant difference (p=0.017) in the median values of the right HHR between males and females at 1.47 and 1.38 respectively whereas there was no signi cant difference (p=0.178) in the left HHR between males and females at 0.93 and 0.92 respectively. The mean right hilar height ratio between males and females in the study done by Homer (12) was 1.31 and 1.35, respectively whereas the left HHR between males and females was 0.89 and 0.82 respectively.
The mean right hilar height ratio between males and females in the Kenyan study done by Wanene (11) was 1.19 and 1.21 respectively whereas the left HHR between males and females was 0.86 and 0.84 respectively. In both the Kenyan and USA populations, females had a higher mean of the right HHR though not statistically signi cant. This is reversed in our study, in which females had a higher positioned right hilum compared to the males. The left hilum is, however, not affected by sex.
There was a signi cant mean difference (p = 0.001) in the right hilar height ratio among the age groups strati ed into 15-year intervals. In contrast, there was no difference (p = 0.198) in the left hilar height ratio. There were no statistically signi cant differences in the different strati ed age groups in the right and left hilar height ratios in the Kenyan study (11). However, the Kenyan study population was skewed to the left with only 3 (1.6%) of its population aged above 45 years, which could explain the differences. In our study population, the right hilar height ratios increased with age which implies the position of the right hilum lowered with increasing age while the left hilar position remained unchanged. This could be explained by the fact that the left hilum is anchored by its course above the left main and upper lobe bronchus. On the other hand, vessels that constitute the right hilum course anterior to the right main bronchus and its divisions and their positions are not xed. To the best of our knowledge, our study is the rst to document the lowering of the right hilum with increasing age.
The right and left hilar height ratios were generally not affected by changes in height and weight. It is, therefore, not necessary to measure height and weight when calculating hilar height ratios.

Conclusions
The study found that the median right hilar height ratio was Application for waiver of consent from SOMREC was made since the study focused on the consecutive collection of already acquired radiographs with no additional risk to participants included in the study. The waiver did not adversely affect the welfare or rights of participants.

Consent for publication
Not applicable Availability of data and materials The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.  Box and whisker plots of hilar height ratios against the gender of participants Figure 4 correlation between hilar height ratio and height of participants A; PA chest radiograph of a 23-year-old male 1.64 m in height, BMI of 24.84 kg/m2, a=13.61cm, b=9.5cm, c=11.0cm, d=12.86cm, RHHR=1.43 and LHHR=0.86. Figure 5B;