Discussion
Our findings from a retrospective, single-institution database study over 5 years confirmed previous studies that in patients with rib fractures caused by blunt trauma, smoking was associated with a decreased risk of mortality and AUD was associated with an increased risk of pneumonia. We also found that AUD was associated with a decreased risk of pneumothorax. These results were based on regression analyses controlling for demographic variables of age, sex, race, and smoking or AUD. Previous studies have shown that pulmonary complications and mortality are associated with ISS and number of ribs fractured.1 35 We did not control for these variables because we only controlled for covariates that were known confounders, not intermediates.
Grigorian et al performed a retrospective analysis of the TQIP database and found that smokers with rib fractures experienced nearly a 40% decreased risk of in-hospital mortality compared with non-smokers (p<0.05), despite having higher rates of pneumonia.20 Our findings reinforce that smoking is associated with a decreased risk of in-hospital mortality (p<0.05). Our results differed from Grigorian et al when examining pneumonia rates between these two groups; we found that patients who smoked were no more likely to get pneumonia than non-smoking patients. This result supports the conclusions of a single-center retrospective study by Marco et al, which determined that smoking was not associated with increased risk of pneumonia in patients with traumatic rib fractures.23
Previous research has demonstrated a ‘smoker’s paradox’: smokers with cardiovascular disease have improved outcomes compared with non-smokers with similar disease.36–38 Studies examining trauma patients who smoke demonstrated that smoking did not increase risk of pneumonia, ARDS, or mortality.21 22 39 Our study provides additional support that the smoker’s paradox may exist in patients hospitalized with rib fractures. Hypothesized pathophysiologic mechanisms to explain these findings have focused on the demonstrated effects of cigarette smoke or its individual components, including nicotine and carbon monoxide (CO). Nicotine acts as a vasoconstrictor in the peripheral vasculature by direct action and by promoting release of epinephrine and norepinephrine,40–42 which could be responsible for maintaining blood pressure and limiting blood loss.43–45 CO may also have contributory vasoactive effects.41 42 The hemostatic effects of cigarette smoking could also contribute to the smoker’s paradox in trauma patients.22 Cigarette smoke promotes coagulation through platelet activation,46–48 effects on blood cells and components,41 49 and endothelial changes.50–52 Finally, the anti-inflammatory effects of nicotine could contribute to the paradox.53 54 When considered together, these mechanisms provide a plausible explanation for the physiologic and biochemical mechanisms of the reduction of in-hospital mortality in this population.
Chronic alcohol exposure is associated with pneumonia and other adverse outcomes in trauma patients.27 Recent studies of trauma patients with rib fractures have shown an association between AUD and higher rates of pneumonia.23 24 Vartan et al also found an association with AUD and higher rates of ARDS and mortality.24 Our results support previous findings of an association between AUD and higher rates of pneumonia; however, we did not observe a higher rate of ARDS, potentially due to low incidence. Our results showed that patients with AUD had a trend toward increased risk of in-hospital mortality, but the association was not statistically significant. Marco et al found no association of pneumothorax in patients with AUD,23 but we found a lower rate of pneumothorax in patients with AUD. Possible explanations for the discrepancy are that their sample size was smaller (n=152) or that the two populations were different in terms of risk factors for pneumothorax that were not measured. Our results add to another recent study that found a surprising inverse relationship between pneumothorax and mortality in patients with rib fractures.2 These findings could be associated with statistical limitations of retrospective design.
There are established associations between AUD and pneumonia or pulmonary inflammation.55 Proposed pathophysiologic mechanisms for this association are multifaceted: increased risk of aspiration,56 decreased upper-airway mucociliary clearance of bacterial pathogens,57 and impaired pulmonary host defenses.58–60 Inflammation can be further exacerbated in this demographic because of a diminished ability to manage oxidative stress and reduced ability to tolerate the fluid accumulation associated with trauma or pneumonia.61 62
A limitation of our study is the possibility of uncontrolled covariates acting as confounders. We made a concerted effort to mitigate this limitation by using DAGs to identify relationships between variables (figures 1 and 2). We chose to control for confounders and not to control for mediators (ie, ISS). Controlling for covariates that are not confounders increases the risk of collider bias or switchover effects, which we hoped to avoid.63 64 The limitation caused by control variable selection has been discussed specifically in the context of the smoker’s paradox.65 As with any retrospective study, unmeasured covariates could have affected the results. These could include certain comorbidities (eg, chronic obstructive pulmonary disease), vital signs (eg, heart rate and blood pressure), and Glasgow Coma Scale. Another limitation associated with the retrospective design is that missing data or inaccurate documentation or coding may have affected the results. We identified patients as smokers or AUD based on patient charts, which are subject to missing or inaccurate data. We did not evaluate the quantity of smoking or alcohol use for each patient, which is an important consideration for future research. Additionally, analysis of some less common complications may have been limited by sample size. Finally, this study was conducted at a single institution in a rural state and results may not be generalizable to other settings.
The results of the primary objective of this study fit within existing literature. We isolated a unique patient population because the study site is the only ACS-verified level 1 trauma center in a rural state. This population has distinct characteristics and is relatively under-represented in national databases, but our results were similar to previous studies. This replication establishes a strong basis for the external validity of our analysis and the generalizability of these findings to other similarly situated trauma centers. To maintain internal validity, multivariate analysis controlling for confounders identified by DAGs was used.
Future research should continue to isolate demographics that are under-represented in national trauma databases to further improve the generalizability of current rib fracture analyses. There is also a need to examine more patient characteristics and injury factors to establish reliable predictors for outcomes. Finally, basic science research is needed to evaluate the mechanisms of unexpected findings in recent retrospective studies, including the associations between smoking and mortality.