METHODS

Inclusion criteria

We included randomised controlled trials and controlled before/after studies of area-wide traffic calming schemes. Eligible schemes included those that involved a number of specific changes to the road layout, road hierarchy or road environment, for example road narrowing, road closures, creation of one way streets, changes at junctions, mini-roundabouts, road surface treatment, or speed humps. We excluded studies describing the enforcement of legal interventions, financial incentives or disincentives, and interventions investigating alteration to road signage or traffic lights alone, or studies solely describing interventions to separate different road users (cycle lanes, bus lanes, pedestrian walkways). The outcomes of interest were all road user deaths, all road user injuries (fatal and non-fatal), and the number of traffic crashes.

Identification of studies

We searched the following electronic databases; Cochrane Injuries Group Specialized Register, Cochrane Central Register of Controlled Trials, Medline, EMBASE, Sociological Abstracts Science (and Social Science) Citation Index, National Technical Information Service, Psychlit, Transport Research Information Service, International Road Research Documentation, and TRANSDOC (the last three combined in the TRANSPORT database). One reviewer examined titles, abstracts, and keywords of citations, as given on electronic databases, for eligibility. Where possible the full text of all of potentially relevant citations was obtained. We also searched the web sites of road safety organisations, contacted experts, hand searched conference proceedings, and checked reference lists of relevant papers. There were no language restrictions. Further details of the search strategy can be seen in box 1.

Data extraction and analysis

One reviewer decided whether studies met the inclusion criteria, and this was checked by a second reviewer. Using a data collection form two reviewers independently extracted data on road user deaths, injuries (fatal and non-fatal), traffic crashes, characteristics of the intervention and control area, and types of measures implemented. To assess study quality we collected information on how the intervention and control areas were matched, duration of the before and after periods, and, because of the potential for contamination, we also noted the proximity of the intervention and control areas.

For each study we calculated a rate ratio: the ratio of event rates before and after intervention in the traffic calmed area divided by the corresponding ratio of event rates in the control area. This gives the reduction in the incident rate in the intervention area compared to that in the control area. For example, a rate ratio of 0.8 corresponds to a 20% reduction in events compared with that predicted from the rates in the control area. For the calculation of 95% confidence intervals, standard errors of the logarithms of the rate ratios were constructed assuming that the number of events in each area in each period followed a Poisson distribution,⁷ provided there was at least one event in each period. For studies with no events in one or more periods exact confidence intervals were calculated where the rate ratio was defined. Rate ratios were combined on a logarithmic scale using a random effects meta-analysis model. The assumption of random effects means that the effect estimates and confidence intervals allow for variation in study specific rate ratios over and above that due to variability within studies.⁸ In this meta-analysis such additional variability reflects both underlying heterogeneity in rate ratios across studies and any variability arising through overdispersion⁹ if the assumption that events follow Poisson distributions is violated.

For studies with no events in one or more periods 1/2 was added to all counts in the pooled analysis. In the analysis of road user deaths, where the majority of studies had no events in at least one period, no test of heterogeneity was carried out, and a pooled estimate of the rate ratio was obtained from the column totals. Analyses were carried out in Stata version 7.0 (Stata corporation, College Station, Texas 77845, USA).

RESULTS

The searches identified 12 986 published and unpublished reports which were screened for eligibility. We obtained the full text of 586 reports and of these 12 reports, describing 16 controlled before/after studies, met our inclusion criteria (see table 1).^10–21 We found no randomised controlled trials. Seven studies were done in Germany,^10–12,16 six in the UK,^13,17–21 two in Australia,¹⁴ and one in the Netherlands¹⁵; all were done in the 1970s and 1980s. In most studies attempts had been made to match the intervention and control sites. However, in three differences in the land use characteristics or type of district are reported,^14,15 and in one the control area was much larger than the intervention area.¹⁵ Outcome data was collected from police or local authority records in all studies.

Eight studies reported the number of road user deaths.^10,14 The pooled rate ratio was 0.63 (95% confidence interval (CI) 0.14 to 2.59). This result should be interpreted with caution since many of the studies include at least one period in which no road user deaths were observed. Sixteen studies reported the number of road traffic injuries (fatal and non-fatal).^10–21 The pooled rate ratio was 0.89 (95% CI 0.80 to 1.00) (fig 1), with statistically significant heterogeneity between the studies (p = 0.05). Nine studies reported the total number of road traffic crashes.^10,19,20 The pooled rate ratio was 0.95 (95% CI 0.81 to 1.11) (fig 2), again with statistically significant heterogeneity between the studies (p = 0.001). Thirteen trials reported the number of pedestrian crashes.^{10,14,17–21} The pooled rate ratio was 1.00 (95% CI 0.84 to 1.18) There was no significant heterogeneity (p = 0.21).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Number of road traffic injuries (fatal and non-fatal).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Number of road traffic crashes.

DISCUSSION

This systematic review of controlled before/after studies shows that area-wide traffic calming has the potential to prevent road traffic injuries. Although the effect of traffic calming on road user deaths is in the same direction as for injuries (fatal and non-fatal), because the number of road user deaths in the included studies is low the estimated rate ratio is imprecise. Indeed, the imprecision in the rate ratio may be understated by the confidence interval because the way that the confidence interval was calculated ignores the likely heterogeneity between studies. Although we found no reliable evidence that traffic calming reduces the number of road traffic crashes, because traffic calming may reduce vehicle speeds,²² this is not inconsistent with a reduction in the occurrence of injury. Our estimates of the effectiveness of traffic calming provide a basis for future cost effectiveness analyses that would be important in informing decisions about resource allocation.

Several methodological issues may have a bearing on the validity of these results. Publication and other selection biases are a potential threat to validity in all systematic reviews, but this is a particular problem in road safety where a large proportion of the available research is published in the grey literature. In this review only two of the included studies were published in journals. There are also problems identifying published controlled studies in the road safety databases.²³ Search strategies for identifying controlled studies in medical databases can achieve high sensitivity because terms describing the study methodology are included among the indexing (descriptor) terms. Road safety databases, however, have a very limited range of indexing terms describing the study methodology. Despite our considerable efforts to identity all eligible studies, published and unpublished, irrespective of language of publication, we cannot exclude the possibility that some studies were missed resulting in reduced precision and the potential for bias.

Although we found no randomised controlled trials of traffic calming schemes, the inclusion of studies with well matched intervention and control areas, with adequate before and after periods, may avoid the problem of confounding by changes in the background rate of injury. All but one¹⁵ of the included studies had attempted to match the intervention and control areas and all had collected at least two years before and two years after data, with a number collecting up to five years before or after data.

Because there was significant heterogeneity between the studies reporting the number of road traffic injuries and crashes, these results should be interpreted with caution. The observed heterogeneity may be due to differences in study design, in the types of traffic calming schemes involved, or in the way outcomes were defined and data collected.

The included studies were all conducted in the 1970s and 1980s, and, apart from two Australian studies, were all done in Europe. As a result it may make it more difficult to generalise from this systematic review and make inferences about the effectiveness of present day area-wide traffic calming schemes. In addition road traffic crashes are a major cause of death and injury in low and middle income countries where most of the casualties are pedestrians, cyclists, and riders of motorised two wheelers. Although traffic calming appears to be a promising intervention for preventing road traffic injuries because none of the included studies were conducted in low and middle income countries further rigorous evaluation is required in these settings.