These findings indicate that acute alcohol use (within 6 h of injury) is associated with hospital treatment for unintentional cutting or piercing injuries at home, among young and middle-aged adults. There is evidence of a dose-response relationship with the adjusted odds ratio for 4 or more drinks being considerably higher than that for 1 to 3 drinks, relative to no drinks (8.68 compared with 1.77, respectively). Smoking status modified the effect of alcohol on injury, so that the excess odds of alcohol exposure were much higher among smokers than non-smokers.
The strengths of this population-based study include the relatively high response rate of around 80%. The study base-Greater Auckland, Waikato and Otago regions--covers more than 60% of the total New Zealand population aged 20 to 64 years, and includes both rural and urban environments. The findings, however, need to be considered in light of several limitations. Although the study was designed to be population-based, the higher proportion of individuals of Māori and Pacific ethnicity among non-respondents has introduced a degree of selection bias. The study relied on self-reported data for capturing acute exposures and lifestyle factors and blood alcohol concentration (BAC) was only measured in 5.1% of cases. The accuracy of the information, provided by participants, limits the credibility of our reported effect measures. Actual intake may be underestimated, as has been found with other self report measures, due to reluctance to admit to consumption, or simply poor recall particularly for the period 1 week before the injury occurred [32, 33]. Increased missing information for the week before (over the day before) provides evidence for this effect (Table 2). In subjects with high levels of alcohol use (either before the event or who reported habitual high levels of intake), reported number of units of alcohol consumed during specific periods is unlikely to be accurate . Furthermore, as suggested by the wide confidence intervals around estimates for some of the effect modification analyses (e.g. AUDIT category ≥ 20, and smokers), the study was too small to allow precise subgroup analyses.
The prevalence of hazardous drinking as measured by the AUDIT score (≥ 8) was 32.0%, higher than the 21% of New Zealand adults who identified as having a hazardous drinking pattern in the most recent New Zealand Health Survey (2006/07) . Our findings were, however, similar to the proportion of 25 to 59 year olds who had a moderate to severe injury as a result of an unintentional fall at home (24.5%) . Twenty-nine percent of our participants identified as 'current smokers'. This proportion is higher than New Zealand national estimates which indicate that 22% of adults (15 to 64 years) are current smokers , but it is lower than a US study of moderate to severely injured adult (18 to 65 years) trauma patients (Injury Severity Score > 20), who were admitted to hospital of whom 47.7% were current smokers .
Given the study entry criteria, it is not possible to determine how generalisable the findings are to cutting and piercing injuries that are fatal, do not result in hospitalisation, or occur in settings outside the home, such as workplace or recreational environments.
In case-crossover studies, it is important to select control periods that are sufficiently distant in time from the case period to limit the correlation between the two periods . Our selection of 24 h before and 1 week before are consistent with other case-crossover studies which investigate the role of acute alcohol on injury risk [13, 40]. The 'hangover' or 'residual alcohol' effect in which fatigue may play a role, has been identified as a potential risk factor in previous injury studies [41–45]. The selection of the first control period (the same 6 h in the 24 h prior to the injury occurring) may have limited our ability to assess this phenomenon. However, the point estimates and confidence intervals for acute alcohol use and the odds of injury are concordant between the two control periods, which suggest that a 'hangover effect' is unlikely to bias our results. 'Hangover' effects generally start once BAC is close to zero [43–45] and this is less likely to influence results given we used a 6 h induction period.
Mis-reporting of alcohol use is another potential threat to the validity of this study . If participants had improved memory of alcohol intake immediately before the cutting or piercing injury, compared to their control periods, then the effect of acute alcohol consumption on injury risk may have been inflated. A study investigating the causes of Meniere's disease, explored this phenomenon by repeated questioning of cases during attacks and in different control periods . The authors concluded that outcome-dependent misclassification, was not a major threat to validity. No tendency to overestimate exposure close in time to attacks of the disease occurred, despite strong beliefs among patients of the likely causes of their acute symptoms. In addition, we did not ask if the person was at home using cutting tools in the control period which has been noted to be a potential bias in studies of motor vehicle injuries .
People who have higher socioeconomic status generally experience better health than those who are socially disadvantaged [48, 49]. As well as being considered as potential confounders or effect modifiers in the relationship between alcohol and risk of injury, they are important parameters independently linked to injuries and, as such, need to be incorporated into injury prevention strategies and policy targeting reduction in home injuries. This study was a case-crossover study which is designed to examine transient risk factors and the strength of this design is that participants are their own controls and so cases are self-matched on socioeconomic factors. However, as a result of the study design, we cannot examine the influence of time-invariant exposures such as social status. In a related case-control study  we were able to explore these factors in the subset of our cases that had a landline and we found that the proportion of cases with no individual level socioeconomic deprivation characteristics (55.9%) was similar to that estimated for New Zealand adults (50.7%) . The proportion of cases identifying as Māori or Pacific ethnicity (18.0% and 12.0% respectively) were higher than the expected proportions (9.7% and 9.0% respectively) based on 2006 Census figures for those in this age group and resident in the study regions .
Our findings contribute to the limited body of published evidence for risk factors associated with cutting or piercing injuries. The findings are consistent with previous research which has examined the association between acute alcohol use and unintentional injury [13, 52, 53]. A meta-analysis of acute alcohol use and different classes of injury reported a per-drink (10 g pure alcohol) pooled-effect estimate for unintentional injuries (other than falls or motor vehicle accidents) of OR 1.32 (95% CI: 1.27, 1.36) . This effect measure is similar to our adjusted odds ratio of consuming 1 to 3 units compared to none of 1.77 (95% CI: 0.84, 3.74).
Our study found that the effect of alcohol on injury was stronger among smokers compared to non-smokers. The interaction of alcohol and smoking on a number of outcomes including fire and traffic injury has been the subject of a systematic review by Taylor et al. . The authors concluded that this interaction may increase risk for traffic and fire injury, but suggested future research is required to confirm the relationship. Tobacco use, has previously been linked to some injuries , and impaired impulse control has been observed in smokers , and alcohol consumers . Other explanations offered to explain the increased risk of injury among tobacco smokers include: direct toxicity from nicotine or carbon monoxide; distraction associated with lighting or disposing of cigarettes; or associated medical conditions; such as cardiovascular disease, cataracts or cancer which may impair performance of tasks; during which, injuries may occur . Further analytical studies are required to confirm if the interaction between smoking and alcohol and injury risk exists.