Study design
The two included studies were both cross-sectional analyses conducted by de Rome et al. [
16] in Australia and Giustini et al. [
17] in Italy. The Australian study conducted by de Rome et al. [
16] did not solely investigate the efficacy of back protectors but was aimed at quantifying the association between several types of protective clothing and injury in crashes. Regarding the use of back protectors, the study sought to determine whether foam inserts in motorcycle jackets or separate back armour affected the risk of back or spine injuries (all soft tissue injuries, open wound injuries, fractures and any injuries) compared to no back protection. The study characteristics are summarised in Table
1. Of the 212 riders or passengers involved in motorcycle crashes causing injury or vehicle damage, 126 (59.4 %) were identified through hospital records, 75 (35.4 %) from local crash repairers and 9 (4.2 %) were self-referred [
16].
Table 1
Characteristics of the included studies
Study design | Cross-sectional | Cross-sectional |
Study location and time | Australia, 2008 | Italy, 2011–2013 |
Sample size | 212 | 2,319 |
Participant identification | - Hospital records (60 %) - Local crash repairers (36 %) - Self-referred (4 %) | Collaboration of the Italian National Institute of health with the National Traffic Police |
Data collection | Baseline | Baseline + 30d after hospitalisation |
Intervention | 1. Motorcycle jacket 2. Motorcycle gloves 3. Motorcycle pants 4. Motorcycle boots 5. Helmet 6. Motorcycle back protector (Foam insert in jacket & back armour) | 1. Hard-shell back protectors 2. Jacket or vest with an air bag 3. Uncertified Protective clothing; |
Outcome | Back/spine injuries | Spine fracture and spinal cord injury |
The only significant finding reported by the study was that foam inserts increased the adjusted relative risk of any injury (RR 2.16,
p ≤ 0.05) [
16]. In contrast, separate back armour had a reduced unadjusted and adjusted relative risk of any injuries and all soft tissue injuries, however, these findings were non-significant [
16]. The findings of this study are summarised in Table
2. As also declared by the authors, the lack of effect of back protectors may reflect the insufficient sample size of their study. Indeed, data on the effect of back protectors to the adjusted relative risk of open wounds and fractures was not presented due to small numbers and convergence issues [
16].
Table 2
Summary of findings of de Rome et al. [
16] on the effects of back protection adjusted relative risk of back injuries
No | 75 | 10.7 % | Reference |
Foam insert in the back of jackets | 97 | 21.6 % | 2.16 (p ≤ 0.05) |
Back armour | 40 | 7.5 % | 0.77 (Not Significant) |
The Italian study by Giustini et al. [
17] sought to assess the effectiveness of back protectors in reducing the number and severity of spinal injuries. The 2319 riders or passengers involved in motorcycle crashes were recruited via a register of traffic police interventions.
In contrast to de Rome et al. [
16], the Giustini et al. study contained a much larger sample size and they found statistically different rates in spinal injury based on the level of protection (Table
3). In this study, hard-shell back protectors included only those compliant with the EN1621-2/12 standard while protective clothing designates those that have not reached the standards of this certification [
17]. However, since this study analysed hard-shell back protectors and jackets/vests with safety airbags together, it cannot be determined whether a hard-shell protector is more effective at reducing spinal injuries than airbags. These airbags are incorporated in motorcycle jackets and are deployed when a rider falls off their motorcycle. Nevertheless, there was an increased Mantel-Haenszel odds ratio for spinal cord fracture and injury when uncertified or no protection was compared to certified hard-shell back protectors or airbags (OR = 2.72,
P = 0.049, 95 % CI: 1.00–7.74) [
17].
Table 3
Summary of findings of Giustini et al. [
17] on the effect of back protection to risk of spinal injury
Spinal injury | 52 (59 %) | 24 (27 %) | 12 (14 %) |
No spinal injury | 293 (67 %) | 67 (15 %) | 80 (18 %) |
Giustini et al. [
17] incorporated analysis of motorcycle and moped crashes together. This was despite noted differences in mechanisms of injury between the two groups with moped crashes more likely to be involving other vehicles and less likely to be due to a loss of control. This reduces the generalisability of the study as a measure of the effectiveness for back protectors for motorcycle accidents only.
Although De Rome et al. [
16] sampled patients from local crash repairers and self-referred cases as well, the majority of their sample (60 %) was obtained from hospital records which is likely to bias the sample towards more severe crashes and injuries. This may involve more cases where back protectors had failed to work while excluding those where back protectors proved effective.
Furthermore, selecting for motorcycles undergoing crash repairs may bias the sample towards higher impact motorcycle crashes and therefore more severe injuries. Indeed, studies of car and motorcycle crashes have shown a correlation between crash severity (defined by damage to property) and injury severity in the acute phase [
16,
17]. In addition, lower impact crashes may be less likely to cause damage to a motorcycle, however, most motorcycle crashes don’t involve high speed [
18]. Hence, it is possible that obtaining participants through crash repairers is excluding those involved in slower speed crashes. This is pertinent as protective clothing is perceived to be more effective for reducing injuries for low impact crashes [
19].
The involvement of repair services in recruitment, where they received a recruitment fee to obtain consent and contact details from customers involved in crashes, also increases the susceptibility to selection bias. These non-research affiliated repairers may have been less stringent with their recruitment methods.
De Rome et al. [
16] list the combination of both injury and non-injury motorcycle crashes as a strength of the study, as previous studies utilising injury and/or police reported crashes were biased towards more severe crashes. However, combining injury and non-injury crashes does not necessarily achieve a proportion of injury to non-injury that accurately represents road motorcycle crashes. Indeed, the authors themselves note that the number of injury crashes in the sample is greater and non-injury crashes substantially less than those recorded by police. The study is therefore likely to underestimate any benefits of back protectors.
Baseline data were collected from face to face interviews of patients 2 weeks following their crash. However, for subjects retrieved through crash repairers, a limitation is that there was no corroboration of their clinical history. Given the absence of independent investigation to corroborate the participants’ reports, there is an increased susceptibility to response bias.
The Giustini et al. review is also susceptible to sampling bias as analysis as spinal injuries were determined in hospitalised or deceased victims who had a diagnosis of spinal injuries according to the large groups of ICD-9-CM diagnoses (diagnosis codes: 805, 806, 839) [
17]. Hence, the study only analyses the effectiveness of back protectors in reducing the most severe spinal cord injuries. Subjects were also identified by a register of traffic police records, hence, sampling for more severe crashes which required police attendance, and underreporting of crashes could have affected their sample.