8. Are deployers more likely to complain of respiratory problems than non-deployers?
Question 8
1 PECOT QUESTION
PECOT: Among US military personnel, are personnel previously deployed to the SW Asia Theater of Operations more likely to report respiratory symptoms than military personnel who did not deploy?
Why is this question important?
Patients don’t experience diseases, they experience symptoms. A greater proportion of deployed military complaining of experiencing respiratory symptoms enhances our confidence that they are also experiencing some underlying disease at a greater rate.
2 CONCLUSION
A series of meta-analyses indicate that US military deployed to the SWATO are more likely than non-deployers (or US military deployed elsewhere) to report a range of respiratory symptoms (self-reported cough, wheezing and shortness of breath). Data were drawn from eight studies including (n= 34,964 individuals).
Evidence Level: Taking into consideration evidence for the primary hypotheses, combined with uncertainty and evidence for the alternative hypotheses, the evidence level for this conclusion is Moderate/High.
3 KNOWN BIASES
There are known biases in the research on military exposures. We summarize the direction and magnitude of these biases in the table below.
In summary, because of our ability to detect and measure most sources of bias this question is subject to, it is primarily exposure misclassification underestimate (bias toward the null) that is of concern. Thus, for the most intensely exposed soldiers, the estimate of risk could be between 26%-54% higher.
4 SUMMARY
Eight cohort studies were included in this analysis (see Overview Table). Six of these studied US military,5-9 while the remaining two examined Swedish2 and Australian1 military. All but two compared deployed military to non-deployed military, the exceptions comparing military to civilian2 and military deployed to the SWATO to military deployed elsewhere (Germany).5
All studies recorded recent symptoms (cough, wheezing, or dyspnea [shortness of breath]) via self report. It is worth noting that Karlinsky et al 2004 was the only study to gather self-reported symptoms within the context of a medical interview. Perhaps as a result, the increased risk of cough was lower in this study than the others reporting on this outcome (though this was not the case for wheezing). As noted above, only two studies adjusted for smoking status (Saers et al 2017 and Kelsall et al 2004). Unfortunately, neither study was of US military, so any differences in effects of these two studies cannot be attributed only to smoking adjustment. Differences in the distribution of military responsibilities between countries could account for some of the difference in effect.
Differences between military deployed to SWATO compared to non-deployed (or civilian or deployed elsewhere) are presented below.
4.1 COUGH
Combining data from seven studies, we see an average increase in odds in deployed military of 80% (OR = 1.8, p=0.0028). There was substantial variation among studies, however (I2 = 85.6%, p<0.0001), likely due to differences in sampling approaches, methods of measurement and statistical analyses.
Figure 1. Meta-analysis of Self-Reported Cough in Military Deployed to SWATO Versus Persons Not Deployed to the SWATO
Note that Karlinsky 2004 uses a subset of the larger sample used in Kang 2000 (both from the National Health Survey of Gulf War Era Veterans and Their Families survey). So, we carried out a leave one out analysis to see how the effect would change if including only one study from this survey.
- Omitting Kang, the risk drops to a 77% increased odds (95%CI: 1.23, 2.57)
- Omitting Karlinsky, the risk increases to 95% (95% CI: 1.48, 2.57) increased odds.
In either case, the lower confidence interval is well above OR=1, indicating that the increase, while modest, is unlikely to be due to chance.
4.2 WHEEZING
Four studies reported on wheezing. The result was an average 64% increase in the odds of deployers having this symptom compared to others (OR = 1.64, p=0.0326). As with coughing, there was substantial variation between studies (I2 = 92.6%, p<0.0001).
Figure 2. Meta-analysis of Self-Reported Wheezing in Military Deployed to SWATO Versus Persons Not Deployed to the SWATO
For wheezing,
- Omitting Kang, the risk drops to 41% increased odds (95% CI: 1.09, 1.18) in deployed military
- Omitting Karlinsky, the risk remains stable at 63% (95% CI: 0.75, 3.53) increased odds.
Note that the confidence interval when omitting Karlinsky 2004 includes OR=1 indicating that the difference for wheezing could be due to chance.
4.3 DYSPNEA (SHORTNESS OF BREATH)
Six studies reported on dyspnea. For this comparison, military deployed to SWATO had 137% higher odds of reporting dyspnea than those who were not deployed (OR = 2.37, p=0.0164). As with the other comparisons, there was substantial variation between studies (I2 =82.4%, p<0.001)
4.4 EXAMINING POTENTIAL BIASES
As noted above, there were two potential sources of bias that we can examine in this analysis: bias due to lack of adjustment for smoking and healthy warrior bias.
4.4.1 Bias Due to Smoking and Country Effect
When we carry out subanalyses by whether or not the study adjusted for smoking we see the following:
For all comparisons, the odds of having the symptom were still higher for deployers in the smoking adjusted analyses though, again, this was conflated with differences by country. For both the study of Swedish and Australian military, the effects on wheezing and cough are lower than any of the US estimates, but whether this is a function of lack of adjustment for smoking or because of differences in the roles during military operations cannot be determined. Additionally, because of the drop in power (i.e., only two studies adjusted for smoking), confidence intervals for the smoking adjusted effects are wider and so our confidence in the effect is less certain. This should not be interpreted as evidence of no effect (see Methodological Note), but rather that, without further studies that adjust for smoking status, especially in the US military, we cannot be confident of the exact increase in the odds of the reported symptoms.
Cough
Wheezing
Dyspnea
4.4.2 Bias Due to Healthy Warrior
Because of confounding with nation, we cannot be confident of the level of bias that may result due to the healthy warrior effect. Proctor 1998 (so, pre-911 conflict) is the only study to compare the relative prevalence of symptoms between two deployed groups and finds a much higher odds in military deployed to the SWATO (though, because of the small sample size, the confidence intervals are very wide and cannot rule out chance). Thus, within the US context, while we may have some confidence that the healthy warrior effect would reduce the difference between deployed and non-deployed, we cannot determine how much of an effect the healthy warrior bias may have had on our estimates.
4.5 SUMMARY
These meta-analyses indicate a reasonable likelihood that US military deployed to the SWATO are more likely than non-deployers (or US military deployed elsewhere) to report a range of respiratory symptoms. Our confidence is tempered somewhat by the fact that none of the smoking-adjusted meta-analyses excluded OR=1 from the confidence interval (due to both a drop in power and smaller effect sizes in the adjusted studies). However, because both studies that adjusted for smoking were non-US studies, we cannot determine how much of this effect is due to country and how much is due to the effects of smoking. Additionally, in only two of the eight studies (one non-US2 and one small US study5 ) did the effect estimate include OR=1. This, in addition to the and because of the documented exposure misclassification and healthy warrior biases, increases our confidence of increased risk.
5 UNCERTAINTY
Within the US context, there are two studies from the National Health Survey of Gulf War Era Veterans and Their Families. However, it is possible that participants from the other US surveys (six of the eight studies were of pre-911 military personnel) also took part in this survey. However, given that the Kang survey was a random sample of this pre-911 population, the overlap is unlikely to be serious.
Heterogeneity is substantial in nearly all analyses. However, heterogeneity was reduced when US (I2 =41%) versus non-US studies (I2 =60%) were split for the dyspnea sub-analysis—indicating that differences between countries as well as the combined adjustment for smoking explains much of that overall heterogeneity.
As noted above, the existence of co-occurring associations between analogous exposures and outcomes may indicate a similar underlying causal mechanism. 10 This, in turn, strengthens a causal argument regarding other analogous diseases or conditions. Disproportionate complaints of respiratory symptoms among military deployed to the SWATO compared to non-deployed military gives credence to an increased prevalence of underlying respiratory conditions giving rise to these symptoms. However, respiratory symptoms are gross indicators—being produced from many different respiratory diseases and conditions.
6 STRENGTH OF EVIDENCE
Below is the strength of evidence table.
7 ALTERNATIVE EXPLANATIONS
Smoking
While smoking is known to cause the symptoms examined here, it is unlikely that differences in rates of
smoking completely explains the increased odds of deployers experiencing these respiratory symptoms.
- All studies except Proctor 1998, report higher proportions of ever smokers in the deployed sample. However, the smoking rates between deployers and non-deployers (or not deployed to SWATO) are relatively balanced across studies, with the rate of never smokers being only 2.9% higher on average in the non-deployed arms. It is unlikely that a difference this small could completely account for the differences in odds reported above.
- In the two smoking adjusted studies, elevated odds of respiratory complaints remained significant in the deployers for two of the three symptoms examined (cough and wheezing). Though the estimates were statistically significant in both these studies, the conservative meta-analysis (using the Knapp-Hartung adjustment combined with the small number of studies) could not rule out chance.
Post-Deployment Occupational and Environmental Exposure
There is no evidence for differences in occupational or environmental exposures between deployed and
non-deployed military personnel.
8. OVERVIEW TABLE
9 RISK OF BIAS
The risk of bias for the included studies is provided below.
The most common biases (outside the known general biases listed Known Biases section) were lack of adjustment for important confounding variables (most importantly, in this context, smoking status) and bias from post-exposure treatment. Smoking status could bias results if arms were not balanced, and smoking status not adjusted for. Whether post-exposure treatment would be a serious concern would depend on the underlying condition that gave rise to the symptoms of interest. In the case of interstitial lung diseases, organizing pneumonia is one of the few that is responsive to current treatment. So, seeking treatment for any of these symptoms if the underlying cause is an interstitial lung disease would be unlikely to seriously bias the effects of these analyses. The Kang 2000 was a short report and so suffered from inadequate reporting of detail. Other reports from the same study do not give rise to these concerns.
10 METHODOLOGICAL NOTE
Following the guidance of the Cochrane Handbook12 we do not interpret the failure to reach statistical significance as evidence of no effect. Especially in the context of a small number of studies and events, precision is a challenge and so our level of confidence must be adjusted accordingly. If the confidence intervals (visible in the forest plots) include OR = 1, that means there is the possibility that these results could be due to chance (e.g., if different people were surveyed, the results could be different).
Because we anticipated between study heterogeneity (due to differences in samples, methods, countries, etc.), a random-effects model was used to pool effect sizes. The restricted maximum likelihood estimator13 was used to calculate the heterogeneity variance τ2 and Knapp-Hartung adjustments14 to calculate the confidence interval around the pooled effect.
11 REFERENCES
1. Kelsall HL, Sim MR, Forbes AB, et al. Respiratory health status of Australian veterans of the 1991 Gulf War and the effects of exposure to oil fire smoke and dust storms. Thorax. 2004;59(10):897-903.
2. Saers J, Ekerljung L, Forsberg B, Janson C. Respiratory symptoms among Swedish soldiers after military service abroad: association with time spent in a desert environment. European Clinical Respiratory Journal. 2017;4(1):1327761.
3. Rivera AC, Powell TM, Boyko EJ, et al. New-onset asthma and combat deployment: findings from the Millennium Cohort Study. American journal of epidemiology. 2018;187(10):2136-2144.
4. Haley RW. Point: bias from the "healthy-warrior effect" and unequal follow-up in three government studies of health effects of the Gulf War. Am J Epidemiol. Aug 15 1998;148(4):315-23. doi:10.1093/oxfordjournals.aje.a009645
5. Proctor S, Heeren T, White R, et al. Health status of Persian Gulf War veterans: self-reported symptoms, environmental exposures and the effect of stress. International Journal of Epidemiology.1998;27(6):1000-1010.
6. Doebbeling BN, Clarke WR, Watson D, et al. Is there a Persian Gulf War syndrome? Evidence from a large population-based survey of veterans and nondeployed controls. Am J Med. Jun 15 2000;108(9):695-704. doi:10.1016/s0002-9343(00)00405-8
7. Kang HK, Mahan CM, Lee KY, Magee CA, Murphy FM. Illnesses among United States veterans of the Gulf War: a population-based survey of 30,000 veterans. J Occup Environ Med. May 2000;42(5):491-501. doi:10.1097/00043764-200005000-00006
8. Karlinsky JB, Blanchard M, Alpern R, et al. Late prevalence of respiratory symptoms and pulmonary function abnormalities in Gulf War veterans. Archives of internal medicine.2004;164(22):2488-2491.
9. Skabelund AJ, Rawlins FA, 3rd, McCann ET, et al. Pulmonary Function and Respiratory Health of Military Personnel Before Southwest Asia Deployment. Respir Care. Sep 2017;62(9):1148-1155.doi:10.4187/respcare.05438
10. Shimonovich M, Pearce A, Thomson H, Keyes K, Katikireddi SV. Assessing causality in epidemiology: revisiting Bradford Hill to incorporate developments in causal thinking. European journal of epidemiology. 2020;36(9):873-887. doi:10.1007/s10654-020-00703-7
11. Li B, Mahan CM, Kang HK, Eisen SA, Engel CC. Longitudinal health study of US 1991 Gulf War veterans: changes in health status at 10-year follow-up. American journal of epidemiology.2011;174(7):761-768.
12. Schünemann H, Vist G, Higgins J, et al. Chapter 15: Interpreting results and drawing conclusions.
In: Higgins JPT TJ, Chandler J, Cumpston M, Li T, Page MJ, Welch VA, ed. Cochrane Handbook for Systematic Reviews of Interventions version 64 (updated August 2023). 2023.
13. Viechtbauer W. Bias and Efficiency of Meta-Analytic Variance Estimators in the Random-Effects Model. Journal of educational and behavioral statistics. 2005;30(3):261-293. doi:10.3102/10769986030003261
14. Knapp G, Hartung J. Improved tests for a random effects meta-regression with a single covariate. Stat Med. Sep 15 2003;22(17):2693-710. doi:10.1002/sim.1482