Page 120 - JSOM Summer 2018

TABLE 1 Guidelines for Determining a Causal Relationship 2 ratios indicated that smoking increased injury risk in prospec-
Number Criterion Guideline tive and retrospective study designs. Risk was also elevated
1 Temporal Exposure to smoking occurs before among smokers in case-controlled studies, but there was high
relationship injuries variability in the risk ratios as indicated by the I statistic.
2
2 Strength of The stronger the association, the Nonetheless, the collective data indicated a moderate to strong
association more likely that smoking directly relationship between smoking and injuries.
causes injuries
3 Dose-response Injury risk increases with greater Dose-Response Relationship
relationship amount of smoking The third criterion is that there should be a dose-response rela-
4 Replication of Relationship between smoking and tionship, although the absence of this does not rule out a causal
findings injuries is seen in several studies relationship because a threshold effect may be present (i.e.,
5 Biological Relationship between smoking an increase in injury risk after a certain critical level). Many
plausibility and injuries is consistent with the military studies have generally found a dose-response relation-
current biological knowledge
6 Alternative Alternative explanations that might ship or a trend toward this relationship, regardless of statisti-
5–7,10,14,16,18,38–40
explanations account for the relationship are cal significance. The studies that have included
taken into account dose-response data used a variety of measures to define the
7 Cessation of Smoking cessation decreases injury dose. These have included cigarettes smoked per day in last 30
exposure risk days, 5,6,10,14,16,38–40 number of days of smoking in last 30 days, 10,14,38
8 Specificity of Smoking is related to specific and cigarettes smoked per day in last year. 7,18 A meta-analysis of
association injuries nine military studies that had provided dose-response data clas-
15
9 Consistency with Relationship between smoking and sified each smoking level as low, medium, or high. As shown
other knowledge injuries is consistent with other in Table 3, the risk of injury increased as the level of smoking
knowledge increased, indicating a dose-response. Civilian studies that have
examined this have also noted a dose-response. 30,36
ranged from 0.74 (95% confidence interval [CI], 0.47–1.18 ;
23
i.e., smokers have 26% lower risk) to 3.1 (95% CI, 1.60–5.90 ; TABLE 3 Risk of Injury at Various Levels of Smoking 15
7
i.e., smokers had more than three times the risk of injury). A re- Relative Risk 95% Confidence
cent meta-analysis that combined data from 18 military inves- Level of Smoking (Smokers/Nonsmokers) Intervals
15
tigations reported that smokers were 1.31 (95% CI, 1.26–1.36) Low 1.27 1.16–1.39
times more likely to suffer an injury compared with nonsmokers. Medium 1.37 1.26–1.49
High 1.71 1.47–1.99
Using PubMed, we conducted a review to identify studies that
examined differences in occupational injuries between smok- Replication of Findings
ers and nonsmokers. Among the 14 studies found, the risk The fourth criterion is replication of findings, meaning that the
of injuries among smokers ranged from 1.2 to 5.9 times relationship should be seen in different studies and in different
24
25
higher than nonsmokers. 24–37 These studies had a variety of populations. Our review has focused on military studies and
study designs, including prospective cohort, respective co- there is substantial evidence that smoking is associated with in-
hort, and case-control. Where risk ratios and 95% CIs were juries in the Armed Forces, 5–22 but studies have been conducted
provided or could be calculated in these studies (n = 10), me- in other subgroups. The higher injury risk among smokers has
ta-analysis (random model) was performed separately for each been found in industrial petrochemical and chemical work-
26
study design category. As shown in Table 2, the pooled risk ers, postal workers, manual material handlers, 25,29 factory
27
28
TABLE 2 Meta-Analyses of Civilian Studies Examining Relationship Between Smoking and Injuries
Individual Studies Combined Studies (Pooled Data)
Risk Ratio: Risk Ratio:
Smokers/Nonsmokers Smokers/
2
Study Injury Definition (95% CI) Study Design Nonsmokers 95% CI I (p value) a
Ryan et al. 28 Any occupational injury 1.40 (1.13–1.73)
Craig et al. 29 Any occupational injury 1.59 (1.16–2.02) Prospective 1.54 1.35–1.73 0 (.39)
cohort
Wen et al. 36 Injury-related mortality 1.69 (1.39–2.05)
Bhattacherjee Any occupational injury 1.51 (1.04–2.20)
et al. 35 Retrospective
Dong et al. Any occupational injury 1.51 (1.10–2.08) cohort 1.73 1.30–2.17 63 (.07)
34
Kim et al. 25 Any occupational injury 2.18 (1.74–2.75)
Chau et al. 33 Any occupational injury 2.19 (1.92–2.50)
with sick leave
Engkvist et al. 24 Any occupational back 1.24 (1.01–1.52)
injury Case-control 1.51 1.04–1.99 93 (<.01)
Chau et al. 32 Any occupational injury 1.17 (1.06–1.29)
with sick leave
Kunar et al. 37 Any occupational injury 1.47 (0.95–2.25)
I is a measure of how similar the risk ratios are among studies. Smaller values indicate the risk ratios among studies are more similar and larger
a 2
95
values indicate risk ratios among studies are less similar. In calculating I , negative values are displayed as zero, which indicates very similar
2
risk ratios among studies. 96
118 | JSOM Volume 18, Edition 2/Summer 2018