Documente Academic
Documente Profesional
Documente Cultură
Am J Forensic Med Pathol • Volume 00, Number 00, Month 2018 www.amjforensicmedicine.com 1
Copyright © 2018 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Tse et al Am J Forensic Med Pathol • Volume 00, Number 00, Month 2018
To explore the differences between the 2 groups, nonpaired between age and LH (0.279, P < 0.05), and no significant corre-
Kruskal-Wallis test was used to compare age, and Fisher exact test lation between age with L and LB (P > 0.05).
was used to compare the difference in sex distribution between For dichotomous variables (Table 3), male sex had statisti-
each group. cally higher L (P < 0.05) but no statistical difference in LH and
Assuming monotonic and nonparametric relationships, LB (P > 0.05). Drowning deaths had statistically higher LH and
Spearman correlation coefficient (ρ) was determined between LB (P < 0.05), but no statistical difference in L (P = 0.15).
age and L, LH, and LB. The differences between L, LH, and LB The AUCs in the constructed ROC curves were 0.584, 0.787,
with dichotomous variables (sex and drowning) were analyzed and 0.678 for L, LH, and LB, respectively (Fig. 1, Table 4). The
using nonpaired Kruskal-Wallis test. Receiver operating charac- optimal cutoff points for L, LH, and LB were 1290 g (sensitivity,
teristic (ROC) curves were plotted for L, LH, and LB with drown- 0.62; specificity, 0.48), 2.99 g/g (sensitivity, 0.78; specificity, 0.48),
ing as the outcome. The optimal cutoff point was determined by and 17.26 g/kg (sensitivity, 0.6; specificity, 0.7), respectively.
minimizing the Euclidean distance to the perfect predictor (sensi-
tivity of 1 and false-positive rate of 0). The areas under the curve DISCUSSION
(AUC, indicator of diagnostic accuracy) for L, LH, and LB were Drowning is a potential cause of death in bodies retrieved
determined together with the 95% confidence intervals (CIs). from water and is defined as the process of experiencing respira-
Statistical analysis was performed using R (version 3.4.1; The tory impairment from submersion in liquid.13 The mechanism of
R Foundation for Statistical Computing, Auckland, New Zealand). death from drowning is from asphyxiation and hypoxia from inha-
A 2-tailed P < 0.05 was considered significant. lation of the immersed liquid, which can occur within minutes.3,14
Diagnosing drowning can be difficult and requires total circum-
stantial investigation correlating with autopsy findings.1 Not all
RESULTS bodies recovered from water have drowned. Bodies recovered
A summary for drowning and nonimmersion death groups from water may have died of natural disease before falling into
is presented in Table 1. In the nonimmersion death group, water or in water, died of injuries before entering or while in
there were 36 natural deaths (28 cardiovascular related deaths, water, died of effects of immersion other than drowning, or died
5 infective/inflammatory-related deaths, and 3 gastrointestinal- of drowning.3 However, in a practical level, it may be difficult
related deaths), 8 drug-related deaths (6 synthetic cannabis deaths, or impossible to provide a definitive cause of death in bodies im-
1 mixed prescription drug toxicity death, and 1 methamphetamine mersed in water, especially when unwitnessed and with a delay in
toxicity death), 2 asphyxia-related deaths, 2 hypothermic deaths, body recovery leading to decomposition.3 Furthermore, not only be-
and 2 traumatic deaths. The mean age in the drowning group ing a diagnosis of exclusion, drowning is commonly the final result
was significantly lower in nonimmersion deaths (P = 0.001). of different initiating causes of the victim's incapacitation in water.15
There was a trend toward statistical difference with higher male With respect to autopsy findings, although there is no single
sex in the drowning group, but this did not reach statistical “diagnostic test,” autopsy can provide clues to aid the diagnosis.1
significance (P = 0.086). A common change in drowning is in the lungs. Classic de-
For continuous variables (Table 2), Spearman correlation scription of the lungs in drowning deaths is waterlogged and
showed statistically significant positive correlation coefficient
Copyright © 2018 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Am J Forensic Med Pathol • Volume 00, Number 00, Month 2018 Lung-Heart and Lung-Body Ratios in Drowning Deaths
FIGURE 1. Receiver operating characteristic curves for L (in g, A), LH (in g/g, B), and LB (in g/kg, C).
overdistended,3,6,14,16 and an increase in L is well recognized.3 was proposed, suggesting it might be a better measurement in di-
During drowning, the lungs can have a substantial increase in agnosing drowning.12 This was followed by using LB, which
weight by immersed liquid aspiration, pulmonary alveolar dam- showed some success.4 The most recent use of L was in the form
age, cardiac failure, or the combination of the three.3–5 Although of a drowning index, which combined L, pleural effusion, and
“dry downing” had been described, it has been reported to have an spleen weight, but was subsequently disproved.7–9 A previous
incidence of less than 2%.17 Previous studies have described in- study using combined L, LH, and LB showed significantly
creasing L from 370 to 540 g to 600 to 700 g, with combined L higher L, LH, and LB in drowning deaths and subsequently pro-
of 1411 g in drowning compared with 994 g in controls.3,10,18,19 posed an optimal cutoff of 17.5 g/kg for LB, and when opti-
Variations using L were subsequently utilized, such as combin- mized for specificity, this would be 19.5 g/kg.4 The overall
ing lung and pleural effusion weight, L corrected for heart or success using L and its variations appeared to have limited
body weights, and factoring in spleen weights.4,5,7–9,12 An LH utility, and the diagnostic accuracies of these parameters are
not established.4,5,7–10,12,18–20
The presented study was carried out to investigate the diag-
TABLE 4. Area Under the Curve and Corresponding Optimal nostic accuracy in L, LH, and LB in drowning. Establishing the di-
Cutoff Points on the Plotted ROC Curves for L, LH, and LB agnostic accuracies would enable the pathologist to have a better
With Drowning as Outcome understanding of using these parameters in diagnosing drowning.
AUC 95% CI Cutoff Point Sensitivity Specificity
L, g 0.58 0.47–0.70 1290 0.62 0.48
Robustness of L, LH, and LB With Respect to Sex
LH, g/g 0.79 0.70–0.88 2.99 0.78 0.72
and Age
LB, g/kg 0.68 0.57–0.78 17.26 0.60 0.70 Organ and body weights are known to vary between age and
sex,3,21–24 limiting the utility of L on its own as a useful measurement
Copyright © 2018 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Tse et al Am J Forensic Med Pathol • Volume 00, Number 00, Month 2018
in diagnosing drowning. Previous studies used different methods to Although this study confirmed previous studies and indi-
adjust these factors to account for the variation from age and cated that LB should be assessed in drowning death rather than
sex, in which LH and LB seemed to be most accepted.4,5,7–9,12 L and LH,4 its utility as a diagnostic test is limited. The diagnostic
In the presented study, the overall age was lower in drowning accuracy of LB, having an AUC of less than 0.8, to diagnose
deaths than in nonimmersion deaths with no difference in sex drowning was ranked as fair and may just be better than by chance
distribution. The effects on the differences between age and sex alone. This study supports the idea that there is some use in
on the L, LH, and LB were explored using Spearman coefficient assessing LB, rather than L or LH, in suspected drowning deaths,
(ρ) for continuous variables and nonpaired Kruskal-Wallis for di- but its ability to diagnose drowning is at best fair.4 Lung-body ratio
chotomous variable. Overall, L was higher in male sex and did not is, however, in our opinion, would not be accurate enough as a
correlate with age. Lung-heart ratio did not show any difference standalone marker for drowning.
between sexes but had a positive correlation with age, albeit weak
(ρ < 0.3). Lung-body ratio did not show any significant difference
between sexes and had no correlation with age, making it the LIMITATIONS
most robust measurement with age and sex between the 3 pa-
rameters (L, LH, LB). This finding was not inconsistent with
Study Design
previous studies.4,5
The study examined the diagnostic accuracy of L and its var-
iations in diagnosing drowning deaths from nonimmersion deaths.
Diagnostic Accuracy This study did not separate fresh and salt water drownings and did
Previous studies have shown that L, LH, and LB are statisti- not segregate sex and stratify different age groups. Nonimmersion
cally higher in drowning deaths compared with nondrowning deaths were used as control and not further subcategorized to
deaths but were reserved in its utility as reliable diagnostic test.4,5 different causes of death.
Although showing a trend in L, our study showed statistically Although both types of drowning deaths (salt and fresh water)
significant difference only in LH and LB, which was in keeping are documented to have similar increase in L, extrapolation of our
with previous studies.3–5,10,18,19 Furthermore, L, LH, and LB with results to fresh water drowning deaths must be done with caution.
their respective ranges in our study were also comparable with Our study did not include pediatric populations especially below
previous studies.4,5 the age of 10 years, and extrapolation beyond these parameters
Although a medical test can demonstrate a significant differ- cannot be supported in our study. The interactions between age
ence between case and control, it may not be diagnostically useful. and sex on L, LB, and LH were examined by correlation studies
Receiver operating characteristic curve analysis, particularly for continuous variables and Kruskal-Wallis tests for dichotomous
estimating the AUC, is an accepted way of determining and variables to examine the robustness of these lung measure-
comparing the diagnostic accuracies for medical tests.25 A test ments. It concluded that LB was the most robust measurement
with an AUC of 0.5 means that the diagnostic accuracy is no better and accurate measurement for diagnosing drowning, although
than chance, and an AUC of 1 would be the perfect test. AUC of it was still poor as a marker on its own. Further studies separat-
0.90 or greater is ranked as excellent, between 0.80 and 0.90 ing sexes and diving into different age group may further evalu-
being good, between 0.80 and 0.70 being fair, and below 0.7 being ate the role of sex and age might play on L in the context
poor.26 Another feature of using ROC curve is that a cutoff point of drowning.
can then be subsequently determined by maximizing both sensi- In terms of control groups, death not due to drowning but re-
tivity and specificity or either of them depending on the condition covered from water would be the ideal group. These types of
in interest.25 deaths are rare, and most would be excluded in the study. There
In the presented study, the diagnostic accuracy, assessed using were approximately 10 cases over approximately 10-year period
AUCs on the plotted ROC curves, showed that LH (0.79; CI, from our previous experience.27,28 These cases were commonly
0.70–0.88) had the highest diagnostic accuracy, followed by LB trauma related with disruption of the body, in which accurate
(0.68; CI, 0.57–0.78) and L (0.58; CI, 0.47–0.70). The 95% CIs organ and body weights cannot be established. Nonimmersion
for the AUCs for LH and LB were above 0.5 but were overlapping, deaths would be a reasonable alternative control group for the
and they were unable to be statistically determined which of LH or purpose of this study. However, in using nonimmersion death
LB was better than the other. The AUC determined with L had a as a control group, our study assumes that L, heart weight, and
95% CI falling between 0.47 and 0.70, making it unable to dis- body weight are unaffected by immersion, which, in the authors'
criminate drowning better than chance. opinion, is reasonable. It is under this assumption that the result
Our study showed a difference in age between the 2 groups of our study could be applied in differentiating drowning deaths
and a higher proportion of male in both groups, which may possi- from death not due to drowning but recovered from water. With
bly be confounding factors particularly with L and LH. The effects the overall low number of cases, future studies collaborating with
of age and sex on LB, LH, and LB were discussed in previous other centers on collecting death not due to drowning but recov-
section. Lung weight and LH were dependent on sex and age, ered from water are recommended.
respectively, and were considered confounding factors. Although Subcategorization of nonimmersion deaths was not per-
being most robust and showing no correlation and difference with formed as this study was not designed to discriminate drowning
age and sex, respectively, it cannot be entirely ruled out that age from other specific causes of death. However, the non–immersion-
and sex were not confounding factors for LB. related groups had values for L, LH, and LB that were not dissimilar
Based on the robustness of the L, LH, and LB with respect to to those found in other similar studies, and it is the authors' opinion
age and sex, and the diagnostic accuracy assessed by AUC, the that the nonimmersion group was robust as a comparison group
most accurate and robust diagnostic parameter would be LB. This for drowning for the purpose of this study.4,5,10,12
finding was in keeping with previous observations supporting the
use of LB over LH and L.4 The cutoff point determined in this Confounding Factors
study for LB (17.3 g/kg) was also in keeping with previous The underlying natural disease and possible role of intoxica-
study (17.5–19.5 g/kg).4 tion in the drowning deaths used in the study were not recorded
Copyright © 2018 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Am J Forensic Med Pathol • Volume 00, Number 00, Month 2018 Lung-Heart and Lung-Body Ratios in Drowning Deaths
and accounted for. It is possible that drowning deaths were 2. Davis JH. Bodies in water. Solving the puzzle. J Fla Med Assoc.
overdiagnosed in these cases and were in fact natural causes 1992;79(9):630–632.
and/or from intoxication. Not taking account of these conditions 3. Saukko P, Knight B. Knight's Forensic Pathology. 4th ed. Boca Raton, FL:
in the drowning groups would result in a more robust result and CRC Press; 2016.
underestimating the diagnostic ability of L. Another possible
4. Azparren JE, Cubero C, Perucha E, et al. Comparison between lung
confounding factor was the presence of resuscitation, which weight and blood strontium in bodies found in seawater. Forensic Sci Int.
was not recorded in this study. Resuscitation may alter the 2007;168(2–3):128–132.
intravascular volume and possibly L, which may need to be
accounted for. Further study in controlling for these conditions 5. Zhu BL, Quan L, Ishida K, et al. Lung-heart weight ratio as a possible
is recommended. index of cardiopulmonary pathophysiology in drowning. Leg Med (Tokyo).
2003;5(suppl 1):S295–S297.
6. Piette MH, De Letter EA. Drowning: still a difficult autopsy diagnosis.
Optimal Cutoff Points Forensic Sci Int. 2006;163(1–2):1–9.
This study was not designed to determine an optimal cutoff 7. Nishitani Y, Fujii K, Okazaki S, et al. Weight ratio of the lungs and
point for L, LH, and LB in drowning death. An optimal cutoff pleural effusion to the spleen in the diagnosis of drowning. Leg Med
point can be determined by optimizing both sensitivity and spec- (Tokyo). 2006;8(1):22–27.
ificity or both with different mathematical techniques, which
may yield different optimal cutoff points with the same data set. 8. Sugimura T, Kashiwagi M, Matsusue A, et al. Application of the
drowning index to actual drowning cases. Leg Med (Tokyo). 2010;12(2):
Our study optimized both sensitivity and specificity by minimiz-
68–72.
ing the Euclidean distance to the perfect predictor and established
a cutoff point that was similar to previous proposed cutoff point 9. Wardak KS, Buchsbaum RM, Walyzada F. The drowning index:
for LB.4 The cutoff points determined in the presented study can implementation in drowning, mechanical asphyxia, and
be used as a reference only but still have poor specificity to diag- acute myocardial infarct cases. J Forensic Sci. 2014;59(2):
nose drowning on their own. 399–403.
10. Hadley JA, Fowler DR. Organ weight effects of drowning and asphyxiation
on the lungs, liver, brain, heart, kidneys, and spleen. Forensic Sci Int. 2003;
RECOMMENDATION 137(2–3):239–246.
This study suggests that although LB is the most accurate 11. Maeda H, Zhu BL, Ishikawa T, et al. Analysis of postmortem biochemical
and robust parameter in diagnosing drowning among L, LH, and findings with regard to the lung weight in drowning. Leg Med (Tokyo).
LB, it was still a poor marker. The accuracy of LB, compared with 2009;11(suppl 1):S269–S272.
other medical tests, was low, which limits its usefulness. It is rec-
ommended that LB may provide only supplemental information 12. Zhu BL, Quan L, Li DR, et al. Postmortem lung weight in drownings: a
comparison with acute asphyxiation and cardiac death. Leg Med (Tokyo).
for pathologists in making a diagnosis of drowning. Lung-body
2003;5(1):20–26.
ratio should be interpreted in conjunction with the circumstances
and other signs of drowning at autopsy. When there is any evidence 13. van Beeck EF, Branche CM, Szpilman D, et al. A new definition of
of decomposition or absence of any signs of drowning, LB should drowning: towards documentation and prevention of a global
be interpreted with caution. Further studies in combining LB with public health problem. Bull World Health Organ. 2005;83(11):
other signs or ancillary tests of drowning, such as serum or vitreous 853–856.
electrolytes, may provide an additive effect in increasing the diag- 14. Shkrum M, Ramsay D. Forensic Pathology of Trauma Common Problems
nostic accuracy of drowning. for the Pathologist. Totowa, NJ: Humana Press Inc; 2007.
15. Bell MD. Drowning. In: Dolinak D, Matshes E, Lew E, eds.
Forensic Pathology Principles and Practice. San Diego, CA:
CONCLUSIONS
Elsevier; 2005.
Among L, LH, and LB, LB was the determined to be the
most accurate and robust parameter in diagnosing drowning. 16. Lunetta P, Modell JH. Forensic Pathology Reviews. Vol. 3. Tsokos M, ed.
However, the overall diagnostic accuracies of the parameters Totowa, NJ: Humana Press; 2005.
(L, LH, and LB) were low (AUC <0.8) compared with other ac- 17. Lunetta P, Modell JH, Sajantila A. What is the incidence and significance
cepted medical tests. Overall, LB (as well as L and LH) should not of “dry-lungs” in bodies found in water? Am J Forensic Med Pathol. 2004;
be interpreted as a discriminating test for drowning and should be 25(4):291–301.
used in conjunction with the circumstances and other signs of 18. Copeland AR. An assessment of lung weights in drowning cases. The
drowning at autopsy (Supplemental Digital Content, http://links. Metro Dade County experience from 1978 to 1982. Am J Forensic Med
lww.com/FMP/A10 and http://links.lww.com/FMP/A11). Pathol. 1985;6(4):301–304.
19. Kringsholm B, Filskov A, Kock K. Autopsied cases of drowning in
ACKNOWLEDGMENTS Denmark 1987–1989. Forensic Sci Int. 1991;52(1):85–92.
The authors acknowledge the technical support from both 20. Chen JH, Quan L, Ishikawa T, et al. Postmortem lung weight
the Departments of Biochemistry and Forensic Pathology at with regard to survival time. Legal Med (Tokyo). 2009;11(suppl 1):
LabPLUS, Auckland City Hospital. The authors thank the forensic S238–S240.
pathology technicians and operations manager, Shane French, 21. Molina DK, DiMaio VJ. Normal organ weights in men: part II—the brain,
Claire Barker, Julie Ruddell, Bonnie To, Jennifer Sucich, Alvin lungs, liver, spleen, and kidneys. Am J Forensic Med Pathol. 2012;33(4):
Fondling, Stacey Wilkinson-Hunt, Bonnie To, and Ross Anderson. 368–372.
22. Molina DK, DiMaio VJ. Normal organ weights in men: part I—the heart.
REFERENCES Am J Forensic Med Pathol. 2012;33(4):362–367.
1. Davis JH. Bodies found in the water. An investigative approach. Am J 23. Molina DK, DiMaio VJ. Normal organ weights in women: part I—the
Forensic Med Pathol. 1986;7(4):291–297. heart. Am J Forensic Med Pathol. 2015;36(3):176–181.
Copyright © 2018 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Tse et al Am J Forensic Med Pathol • Volume 00, Number 00, Month 2018
24. Molina DK, DiMaio VJ. Normal organ weights in women: part II—the 27. Cala AD, Vilain R, Tse R. Elevated postmortem vitreous sodium and
brain, lungs, liver, spleen, and kidneys. Am J Forensic Med Pathol. chloride levels distinguish saltwater drowning (SWD) deaths from
2015;36(3):182–187. immersion deaths not related to drowning but recovered from
25. Hajian-Tilaki K. Receiver operating characteristic (ROC) curve analysis saltwater (DNRD). Am J Forensic Med Pathol. 2013;34(2):
for medical diagnostic test evaluation. Caspian J Intern Med. 2013;4(2): 133–138.
627–635. 28. Garland J, Tse R, Oldmeadow C, et al. Elevation of post mortem
26. Muller MP, McGeer AJ, Hassan K, et al. Evaluation of pneumonia vitreous humour sodium and chloride levels can be used as a reliable test in
severity and acute physiology scores to predict ICU admission and mortality cases of suspected salt water drowning when the immersion times are less
in patients hospitalized for influenza. PLoS One. 2010;5(3):e9563. than one hour. Forensic Sci Int. 2016;266:338–342.
Copyright © 2018 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.