Skip to main content
. 2018 Jan 26;15(2):217. doi: 10.3390/ijerph15020217

Table 2.

Summary of included study characteristics and longitudinal findings.

Ref., Country, Exposure Aim Recruitment Process, Study Population, and Data Collection Time Points Study Design, Exposure Details and Comparison Outcome Measures Efficacy on Outcomes (only Significant Changes in Longitudinal Findings Reported)
Controlled studies
Green space
Quigg, R., et al. (2012) [28].
Dunedin, New Zealand
Exposure: Green space
To assess whether an upgrade of playgrounds in a neighborhood was associated with changes in local children’s physical activity levels. Recruitment: Local authority community boundaries used to identify the intervention neighborhood. Six intervention and four comparison community schools were invited to participate via information letter to students and parents (n = 4 intervention and n = 4 comparison schools accepted). All children aged 5–10 years in kindergarten to grade 4 residing in the community were given information and parent and child consent forms to take home. A sample size calculation was used (n = 100 in each group was required).
Incentive: Family swim vouchers worth $8 were provided as an incentive to wear accelerometers at each time point. A swimming bag worth $3.50 and goggles and a Frisbee worth $8 were given as incentives for completing the survey at T1 and T2.
Participants: T1 = n = 184. T2 = n = 156 (15% loss to follow-up: 10%, n = 9, from the comparison group and 20%, n = 19, from the intervention group). Participants with at least 1 day of accelerometer wear at T1 and T2 = n = 138 (n = 132 had 4 or more days). Survey data collection rate was 93% (128/138) at T1 and 96% (133/138) at T2.
Time points: T1 = October–December 2007. T2 (1-year post-T1 and 3-months post park upgrade) = October–December 2008 (spring).
Study design: Pre-post study with control.
Exposure: playground upgrades in 2/6 existing parks in the intervention community. One playground received 10 new components including: play equipment; seating; additional safety surfacing and waste facilities installed; and the removal of two existing components. The other playground received two new play pieces and modification to an existing piece of equipment.
Comparison: a similar matched community not undergoing park regeneration. Interactions between BMI-Z score and group on physical activity explored.
Total daily physical activity (accelerometer)
BMI Z-score (researcher measured)
Change in total daily physical activity was associated with an interaction between BMI and the participant’s community of residence (p = 0.006), with the intervention being associated with higher levels of PA for children with lower BMIs, but lower levels of PA for children with higher BMIs. Participants in the intervention group, compared to the comparison group, had increases in total PA for those with BMI z-scores below 0.4 and lower total PA for those with BMI z-scores above 0.4.
No other significant changes were reported.
Bike/walk trails
Dill, J. at al. (2014) [29].
Portland, Oregon, United States.
Data from the Family Activity Study (FAS)
Exposure: bike/walk trails
To evaluate changes in physical activity and active transportation associated with installation of new bicycle boulevards. Recruitment: Participants resided in street segments scheduled for bicycle boulevard installation (0.9 to 4.2 miles long). Residents living on and within 1000 ft of the selected street segments were recruited via door delivered flyers, of accessible housing units, and mailed to residents in inaccessible housing units (n = 54,381). The comparison consisted of 11 control street segments (1.0 to 5.7 miles long), similar in urban form and demographic characteristics (especially in terms of bicycle infrastructure). 3.1% of the estimated eligible population was recruited at T1.
Participants: T1 = baseline sample, N = 490; n = 307 adults in the exposure group and n = 183 in the comparison group.
Time points: T1 = 2010–2011 and T2 = 2012–2013. Follow-up varied between 2 and 12 months after exposure.
Study design: Pre-post with control.
Exposure: Bicycle boulevard installations across multiple areas.
Comparison: No bicycle boulevard installations introduced.
Physical activity; MVPA, number of bike/walk trips; number of minutes walking/cycling (accelerometer combined with GPS) Bicycle boulevard introduction was negatively correlated with bicycling (if >10 min, p = 0.00) and the number of bike trips (if >0, p = 0.06).
Supermarkets
Cummins, S., et al. (2008) [30].a
Cummins, S., et al. (2005) [31]. b
Glasgow, UK
Exposure: supermarket
To examine the impact of a new food retail development on diet and health and well-being.
To determine the effect of the introduction of a hypermarket in a deprived community on fruit and vegetable consumption and health including psychological health.
Recruitment: Study site boundaries were identified using postcode districts of areas with the main shopping facilities. Random sample of households surrounding two sites identified from a postcode address file (within 1 km). Postal questionnaires sent to homes (n = 3975). Postal reminders sent after 2 weeks and a 2nd reminder after another 2 weeks (including the survey again), to those that did not reply. Control and intervention response rates were 15.50% and 14.84%, respectively, at T1 and 71.29% and 65.18% at T2.
Incentive: At follow-up £10 shopping vouchers, not for the exposure store, were given to survey responders.
Participants: Surveys were completed by 603 participants at T1 (15.16% response rate) and 412 participants at T2 (68.40%). At T1 293 participants were in the intervention group and 310 in the comparison group.
Study design: pre-post with control.
Exposure: Building of a new hypermarket.
Comparison: a deprived comparison area not undergoing significant infrastructure change. For sub-analyses: those that switched to using the new supermarket compared to non-switchers.
General and psychological health (Self-rated health and well-being survey)
Diet—fruit and vegetable consumption (self-report survey)
An improvement in poor psychological health was found (−12.13%, p = 0.017) in the intervention group from T1 to T2. Vegetable (p = 0.01), and fruit and vegetable combined (p = 0.003) consumption improved in the comparison group from T1 to T2.
Following adjustment for baseline psychological health, the odds of poor psychological health was reduced (OR 0.42, 95% CI 0.19 to 0.92) in switchers compared to non-switchers. Further adjustment for other confounders further reduced the odds of poor psychological health in switchers compared to non-switchers (OR 0.24, 95% CI 0.09 to 0.66). Adjusted odds of having poor health increased in the intervention group compared to the comparison group (OR 1.52, 95% CI 0.77 to 2.99).
At T2 191 participants were in the intervention group and 221 in the comparison group. A sub-analysis was conducted comparing data of those that switched their food purchasing to the new hypermarket (switchers in the 2005 paper: n = 66; switchers in the 2008 paper (n = 61; n = 58 from the intervention group and n = 3 from the comparison group) compared to those that did not.
Time points: Hypermarket opening = November 2001. Survey T1 = October 2001. Survey T2 = 1 year post-baseline and 10 months post supermarket opening.
Unadjusted odds of poor health improved in switchers (OR 0.62, 95% CI 0.34 to 1.11).
No other significant changes were reported.
Cummins, S., et al. (2014) [32].
Philadelphia, PA, USA
Exposure: supermarket
To determine the effects of the opening of a new supermarket, in a community considered a food desert, on BMI, daily fruit and vegetable intake and perceptions of food accessibility. Recruitment: Adult residents living within 1.5 miles of the supermarkets from two neighborhoods randomly selected from a directory list and using random digit dialing.
Incentive: Respondents were given $20 for participation.
Participants: Overall response rate was 47.2% at T1 = n = 1440 (response rate of 47.4% in the intervention group, n = 723 and 47.0% in the comparison group, n = 717). The response rate was 45.5% at T2 = n = 656 (response rate of 43.7% in the intervention group, n = 311 and 43.7% in the comparison group, n = 48.9%).
Time points: The supermarket opened in December 2009. T1 (June–September 2006). T2 (June–November 2010, at least 6-months post-intervention).
Study design: Pre-post with control.
Exposure: Opening of a new supermarket in a food desert neighborhood.
Comparison: Neighborhood without change in existing supermarket facilities (three miles from the intervention neighborhood). Sub-group analyses on those adopting the store as their main store for grocery shopping compared to those that did not adopt the store (they did not use the store at all). Those that used the store as their secondary source of shopping were also compared to non-adopters.
Sites were matched for race/ethnicity, income, demographics and size (3 miles2).
BMI (self-reported height and weight)
Fruit and vegetable intake (self-report survey)
No significant changes were reported.
Time line changes: There was a three-year delay in the construction of the supermarket.
Studies without a control group
Rail stops
Brown, B.B., and Werner, C.M. (2007) [33].
Salt Lake City, UT, USA
Exposure: Rail stop
To test whether a new light-rail stop increases the number of light-rail riders and if light-rail ridership relates to moderate physical activity bouts. Recruitment: Study notification letters delivered to addresses within ½ mile of the rail stop, followed by door-to-door recruitment.
Incentive: $20 given for completing each phase.
Participants: N = 529 (potential sample) living within half a mile of the new rail stop. Deemed ineligible = n = 33. Successfully contacted & invited, n = 215 (n = 102 agreed; n = 113 refused). T1 n = 102 (survey) and T2 n = 51 (survey) and n = 47 (accelerometer).
Age (longitudinal sample): 41 ± 13.82 years.
Time points: The rail-stop was added in autumn 2005. T1 = before summer 2005. T2 = after summer 2006 (1 year post T1).
Study design: pre-post WO control
Exposure: building and opening of a new light-rail stop (between two existing stops) in the center of the surveyed neighborhood
Comparison: No control group. Changes over time explored and associations between use of the light rail and physical activity.
Transit use—previous 2 weeks (self-report survey)
MVPA bouts of ≥8 min over 7 days (accelerometer). MVPA discussed with participant to identify if it related to walking to/from the rail stop.
Rail use increased from 50% to 68.75% from T1 to T2 (p = 0.011).
T1 MVPA was related to MVPA bouts at T2 (unstandardized beta coef = 0.38, SE = 0.12, p < 0.01). At T2 rail rides in the past 14 days (unstandardized beta coef = 0.03, SE = 0.01, p = 0.01) and bigger household sizes (unstandardized beta coef = 0.01, SE = 0.00, p = 0.01), account for variance beyond the effects of prior activity levels.
No other significant changes were reported.
Hong, A., Boarnet, M.G. and Houston, D. (2016) [34]. To determine the impact of a new light rail transit line on active travel behavior Recruitment: Invitation letters were sent to all households in the study area (n = 27,275). Incentive: $30 for T1 and $75 for T2 completion.
Participants: The total sample at T1, was n = 279 (1% response rate, 74%F, aged 52 ± 14 years, 49% African-American) and at T2 was n = 204. Accelerometer and GPS data collected in n = 143 (66%F, aged 50 ± 14 years, 55% African-American) and analyzed for n = 73 participants.
Time points: T1 = 5–7 months prior to the line opening. T2 = 2–6 months after the opening of the line.
Study design: pre-post WO control.
Exposure: building a new light-rail line (with several stops).
Comparison: No control group. Sub-group Changes over time explored in those residing <½ mile and >½ mile from the stations on the new line.
Transit usage and frequency of bus and train trips, frequency of walking and cycling (self-reported diary)
Physical activity (accelerometer)
There was a negative association between total walk trips at T2 based on the interaction of distance to rail stop group and baseline walking trips (beta coef = −0.02, p = 0.008).
Bike/walk trails
Burbidge, S.K., and Goulias, K.G. (2009) [35].
UT, USA
Exposure: bike/walk trail
To determine the impact of introducing a neighborhood trail on active travel and total physical activity of residents. Recruitment: NS. Participants: Activity diary component; n = 196 households (n = 175 individuals from n = 80 households at T1), n = 144 individuals from n = 56 households at T2 and n = 107 individuals from n = 41 household at T3). Questionnaire component; n = 290 households with 796 individuals living within 1 mile of the trail, plus a further 32 new resident households.
Time points: The trail opened in September 2007. Activity diary; T1 = February 2007 (prior to trail construction), T2 = 1-month post trail opening (October 2007), T3 = 5-month post trail opening (February 2008). Questionnaire = October 2007.
Study design: longitudinal survey study WO control
Exposure: building of a new trail along a canal route
Comparison: No control group. Changes over time explored. Proximity to the trail on physical activity was also explored.
Single day activity data: activity type, begin and end time, activity duration, interpersonal interactions, travel related or not, distance travelled if any and mode used if travelled (Self-report activity diaries). Data on residentially stable participants only reported here (data on new residents not reported).
t-test: Total physical activity episodes (p = 0.036) and total walking trips (p = 0.008) decreased from T1 to T3.
Regression: Total physical activity episodes (coef = −0.245, p = 0.036) and total walking trips (coef = −0.265, p = 0.008) decreased from T1 to T3. Regression after controlling for confounders: Total physical activity episodes increased from T1 to T3 in adults aged 18–64 years (B = 0.56, p = 0.024). No other significant changes were reported.
Evenson, K.R., et al., (2005) [36].
NC, USA
Exposure: bike/walk trail
To explore changes in physical activity in local residents that might be attributable to the construction of a multi-use trail. Recruitment: Approximately 28,304 people resided along the trail according to a census in 2000. A random list of 2125 households was generated from a telephone directory. Study postcards were mailed introducing the study followed by telephone surveys (<15 min) to residents living within two miles of the intervention site. The adult with the most recent birthday from each randomly selected household was invited to participate. Participants: N = 2125 adults from random households that had telephone numbers listed in the phone book were targeted from the 28,304 adults living in 11 census blocks that the trail traversed. N = 685 completed T1 surveys (47.2% response rate), n = 436 completed T2 surveys (63.7% retention; 4% refused a T2 survey). Final longitudinal sample: n = 366.
Time points: The first segment of the trail (3.2 miles) opened in June 2000. The second segment (under investigation here) was 2.8 miles plus a 2.0 miles spur, opened in September 2002. T1 = July 2000–April 2001. T2 (1 year 7 months—2 years 4 months post-T1) = November 2002.
Study design: Pre-post WO control.
Exposure: Building of a new walking and cycling trail. The new section of the trail passed by two schools, shopping areas, apartments, and neighborhood divisions and had several access points along the route.
Comparison: No control group. Changes over time explored only in those that used the trail compared to those that did not use the trail were performed.
Leisure PA (self-report survey)
Walking and cycling (self-report survey)
MVPA (self-report survey)
Transportation activity (self-report survey)
Trail use (self-report survey)
General health (self-report survey)
BMI (self-reported height and weight)
Time in moderate PA (p = 0.03), time in vigorous PA (p < 0.0001) and cycling for transport (p = 0.01), decreased in those that reported not having used the trail. Time in vigorous PA (p = 0.01) decreased in those that had ever used the trail.
Those that had used the trail were less likely to increase walking by >30 (OR = 0.46 (95% CI = 0.21–1.01)) or >45 min/week (OR = 0.43 (95% CI = 0.19–0.98)), and less likely to increase cycling by >30 (OR = 4.17 (95% CI = 1.70–10.20)), >15 (OR = 3.99 (95% CI = 1.81–8.79)) or >45 min (OR = 4.14 (95% CI = 1.33–12.90)) from baseline.
No other changes were reported.
Goodman, A., et al. (2013) [37].c
Goodman, A., et al. (2014) [38].d
Cardiff, Kenilworth & Southampton, UK
Exposure: bike/walk trail
To examine and compare patterns of use of high quality traffic free walking and cycling routes, including exploration of journey purpose for which routes were used and the modes by which it was used. Individual and household predictors of use are also determined.
To determine the effects of new cycling and walking routes on overall physical activity levels, walking and cycling.
Recruitment: The electoral register was used to identify 22,500 adults living within 5 km of one of the sites. Surveys were mailed.
Participants: Surveys were completed by n = 3516 adults at T1, n = 1885 adults at T2, and n = 1548 at T3. T2 comprised of n = 1849 (53% retention rate and 8% of the invited population) and T3 of n = 1510 (43% retention rate and 7% of the invited population) surveys. Physical activity data was collected in n = 1796 adults at T2 and n = 1465 adults at T3.
Compared to local and national data, the sample had fewer young adults, were slightly healthier, better educated and less likely to have children than the general population.
Time points: Most feeder routes were upgraded and the core projects had begun in Southampton and Cardiff in July 2010. By September 2011 the core Kenilworth project had begun and almost all feeder routes were complete. T1 = April 2010. T2 = 2011 (12-month follow up). T3 = 2012 (24-month follow-up). Baseline characteristics were measured in the 2010 questionnaire, and infrastructure use was measured in 2011.
Study design: pre-post WO control.
Exposure: Building of new walking and cycling routes in three municipalities. Traffic-free bridges were built in Cardiff and Kenilworth, and a riverside footpath developed into a boardwalk in Southampton.
Comparison: No control group. Changes over time explored. Changes based on distance to walk/cycle routes were included in the 2014 paper.
Use of new infrastructure, journey purpose and journey mode (self-report survey)
Walking and cycling for different journey purposes (7-day recall)
Recreational physical activity—total, moderate and vigorous intensity walking and cycling (IPAQ)
At T2 and T3, 32% and 38% of participants reported using the new infrastructure, respectively (change statistics over time NS and T1 values also NS). Walking for recreation was the most common use. Previous 7-day walking and cycling increased more from baseline in those living nearer to the exposures at T3 (adjusted effect = 15.3 min/week per km closer to the intervention; 95% CI = 6.5, 24.2; p < 0.001) in comparison to those living further from exposures. Proximity to exposure was strongly associated with total physical activity (12.5 min/week per km closer to the intervention; 95% CI = 1.9, 23.1). T3 effects of proximity were found for those reporting using routes (adjusted effect = 30.0 min/week; 95% CI 3.5, 55.5 in users) for total walking and cycling. Proximity to exposure was also associated with change in subdomains of physical activity at T3: cycling for recreation (adjusted effect = 2.5 min/week per km closer to the exposure; 95% CI = 0.1, 4.9); and walking for transport (adjusted effect = 8.8 min/week per km closer to the exposure; 95% CI = 2.8, 14.8).
Change in walking and cycling was greater in those using the routes for ≥2 types of transport (adjusted effect = 46.4 min/week/km; 95% CI = 5.1, 87.7) compared to those using the route for <2 types of transport. Change in walking for recreation was greater in those reporting using the route for walking compared to those not reporting using the route for walking (adjusted effect = 33.3 min/week/km; 95% CI = 4.6, 62.0).
Effects were attenuated but still significant in sensitivity analyses. No other significant changes were reported.
MacDonald, J.M., et al. (2010) [39]. Charlotte, NC
Exposure: bike/walk trail
To determine the effect of using a light rail transit system on BMI, obesity and weekly physical activity. Recruitment: Telephone sampling from census tract addresses within 1 mile of the new train line was undertaken. The adult with the most recent birthday was invited to participate. Overall response rate at T2 was 87% and 3% were refusals (n = 20).
Participants: At T1 n = 839 (45% response rate) and at T2 n = 498 (60% of the T1 sample), adults participated. Only longitudinal sub-group analyses comparing users (n = 26) and non-users (n = 275) were reported; daily light-rail work commuters (n = 26 or 5.2%) compared with non-light rail users (n = 275).
Time point: T1 (18 months prior to the opening of the system) = July 2006–February 2007. T2 = March–July 2008.
Study design: pre-post WO control.
Exposure: Introduction of a new light rail transit system.
Comparison: No control group. Changes over time explored in users versus non-users.
BMI and obesity (self-reported height and weight)
Physical activity (self-report survey)
The exposure was associated with an average −1.18 (95% CI −2.22, −0.13) reduction in BMI (p < 0.05) and an 81% reduced odds (95% CI = 0.04, 0.92, p < 0.05) of becoming obese over time, in users compared to non-users.
No other changes were reported.
Pazin, J., et al. (2016) [40].
Brazil
Exposure: bike/walk trail
To examine the effects of a new cycling and walking route on physical activity in adults residing near the route. Recruitment: Systematic sampling of households from lists of landlines, were used to identify individuals from six neighborhoods (n = 55,700) within 1500 m from the route (n = 7630). The first adult aged >18 years to answer a telephone invite was invited to participate. A sample size calculation was used and was fully reported on, to determine changes over time in the total sample (n = 656 participants were required).
Participants: T1 = 745 (91% response rate from telephone invites and 10% of eligible individuals living in the neighborhoods). T2 = 519 (70% retention).
Time points: T1 = March-July 2009. T2 = March–December 2012 (30 months post baseline).
Study design: Pre-post study WO control.
Exposure: a new avenue, parking lots and a walking and cycling route, along a seashore.
Comparison: No control group. Sub-groups consisting of residents that lived 0–500 m, 501–1000 m and 1001–1500 m from the route were compared.
Total weekly leisure time physical activity using questionnaire (IPAQ through telephone interview). Leisure time walking increased, by 14 min/week (95% CI: 3–24) in residents. Leisure time walking increased by 32 min/week (95% CI: 15–51) in residents living up to 500 m from the new route, which was greater than in those living 501–1000 m away at follow-up (δ = 31 min/week; 95% CI: 11–51). Leisure time walking plus MVPA increased by 51 min/week (95% CI: 2–81) in those living up to 500 m from the new route.
The percentage of participants that initiated leisure time walking or MVPA after the new route was negatively associated with the distance to the route. In participants that did not use the route, (n = 280), a greater proportion of residents in the –500 m (52%) and 501–1000 m (60%) groups reported intention to use the route compared to those in the 1001–1500 m (33%) group (p = 0.006).
No other significant changes were reported.
Miller, H.J., et al. (2015) [41].e
Brown, B.B., et al. (2015) [42]. f
US, Salt Lake City
Exposure: bike/walk trail
To test if light rail transit (LRT) generated new PA in Salt Lake City, UT, USA.
To assess effects on physical activity (PA) and weight among participants in a complete street intervention that extended a light-rail line in Salt Lake City, UT, USA.
Recruitment: Participants were recruited via door-to-door canvassing. Participants resided within 2 km of the new light rail transit line (exposure).
Participants: N = 939 adults. A total of 614 participants completed 12-month follow-up, and 536 of these participants (51% female, 25% Hispanic) had valid Global Positioning System (GPS) data for analysis.
Time points: T1: March–December 2012 and T2: May and November 2013 (the line opened in April 2013).
Study design: pre-post WO control.
Exposure: building and opening of a new light-rail transit line. The transit line included the introduction of five additional residential stops a bike path and improved sidewalks in the area.
Comparison: No control group. Changes over time explored. Sub-group analyses were completed: ‘Never’ (N = 391, including participants who had never used transit; used transit but not within the defined neighborhood; or only biked/walked in the neighborhood) ‘Continued’ (N = 51), ‘Former’ (N = 42, including those who had used transit during the first time period, but not the follow-up)
‘New’ (N = 52, including those with complete transit trips in follow-up, but not T1).
Physical activity (total and transit measured by GPS combined with accelerometer)
BMI (researcher measured)
From T1 to T2, new riders increased transit physical activity by 3.46 min (95% CI: 2.20, 4.72, p < 0.0001). Former riders experienced a decrease of 2.34 min (95% CI: −3.56, −1.08, p = 0.0005) of transit physical activity. Accelerometer counts decreased in former riders from T1 to T2 (−49.35 ± 14.97 cpm; 95% CI: −78.75, −19.94), which was a greater change than in the never-riders, who slightly increased their accelerometer counts by 11.97 cpm, (t = −3.30; p = 0.001). New transit users accrued more accelerometer counts from T1 to T2 (37.40 ± 13.74 cpm; 95% CI: 10.41, 64.39) than never-riders (t = 2.72; p = 0.007). Former riders decreased MVPA minutes (−6.37 ± 2.01 min; 95% CI: −10.32, −2.43), which was different than the change in never riders from T1 to T2 (SE = 2.01; t = −3.17; p < 0.01). New riders increased MVPA by 4.16 ± 1.84 min; 95% CI: 0.54, 7.78), which was a bigger change than in the never riders (SE = 1.84; t = 2.26; p < 0.05). Sedentary behavior sig increased in the former riders by 16.38 ± 66.09 min; 95% CI: 4.41, 28.35, which was different than change in the never riders (SE = 6.09; t = 2.69; p < 0.01). In new riders, sedentary behavior decreased −12.83 ± 5.59 min; 95% CI: −23.82, −1.85, which was different to the change in the never riders (SE = 5.59; t = −2.30; p < 0.05). Former transit increased their BMI (0.64 ± 0.24 kg/m2 95% CI: 0.18, 1.11), (t = 2.72; p = 0.007), whilst new riders had a decrease in BMI (−0.50 ± 0.22 kg/m2 95% CI: −0.93, −0.08), (t = −2.32; p < 0.022). Both changes in former and new rider BMI were different than in never-riders, who had an increase in BMI of 0.19 kg/m2.
Sensitivity analysis: All effects were sustained when 2012 baseline variables were included in analyses as a dependent variable as a predictor. One 1 new effect emerged: former riders had 11.34 fewer minutes of light PA than never-riders (p = 0.03).
Green space
West, S.T. & Shores, K.A. (2011) [43]. Southeastern U.S.
Exposure: Green space
To determine if a new greenway increases physical activity levels in residents residing nearby. Recruitment: The city planning department provided a list of property owners within one mile of the greenway (owning single-family units values >$5000). Invitation was random and via mail (study information letters and surveys were sent). Reminders were mailed 1-week later and another full package sent after the reminder. A total of 1168 invites were sent out. At T1 368/1168 replied (31.5% response rate). At T2 166/368 replied (45.1% response rate from T1 sample and 14.2% response rate from total invites sent out).
Participants: Residents living within 0.5 miles = n = 597. Residents living within 0.5–1.0 miles = n = 571.
Time points: The greenway was completed in early 2008.
T1 = 2007.
T2 (11 months after the intervention was complete) = 2008.
Study design: pre-post WO control
Exposure: development of five miles of greenway (open space for recreation) alongside a river, which connects urban centers.
Comparison: No control group. Changes over time explored. Sub-analyses on looking at differences between residents living within 0.5 miles compared to those living 0.51–1.0 miles from the greenway.
Physical activity, (self-report survey) For the full sample, increases in days of walking for ≥30 min in the past week (2.9–3.3 days), participation in moderate PA (1.7–2.3 days) and participation in vigorous PA (1.3–1.8 days) increased (NS if changes were significant or not).
Comparing those living <0.51 miles to those living 0.51–1.0 miles from the greenway, days of walking for ≥30 min in the past week (Eta2 = 0.53, p = 0.003), moderate activity (Eta2 = 0.133, p < 0.001) and vigorous activity (Eta2 = 1.47, p < 0.001) increased from T1 to T2. No interactions between greenway development and residential proximity were found for any measures.
No other significant changes were reported.
Food retail
Evans, A.E., et al. (2012) [44]. Austin, TX, USA
Exposure: farmers’ markets
To determine if introducing small farm stands without any other strategies in low-income communities increases fruit and vegetable intake in local residents. Recruitment: Data collectors made door-to-door survey visits to low-income households within 0.5 miles of the stands at different times of day (recruitment goal was n = 100 adults). Streets were randomly selected for recruitment (all on the same side of the highway as the farm stands) and only houses perceived relatively safe were targeted (e.g., WO unleashed dogs). Only one attempt was made at each house. A total of N = 312 households were approached; n = 133 answered the door (43%) of total approached homes; n = 36 were not eligible or did not wish to participate. T2 data collection was over the telephone or via mail (if participant was not reached after five telephone call attempts). Six mail packets (8%) were undeliverable at T2 and 24 packets (51%) were not returned. Study design: Pre-post WO control.
Exposure: Two new farm stands introduced to a community (outside community centers 1 day/week for 12 weeks for 2–3 h each). Vouchers to assist low-income families to purchase healthy food were accepted by the stands. No advertisement of the stands occurred. No foods other than fruits and vegetables were available.
Comparison: No control group. Changes over time explored.
Fruit and vegetable intake (self-report survey)
Use of farm markets/stands (self-report survey)
Consumption of fruit (p < 0.001), fruit juice (p < 0.001), green salad (p < 0.05), tomatoes (p < 0.01) and other vegetables (p = 0.001) increased. Awareness of the market increased (from 19.3% to 39.3%, p = 0.001), as did purchasing of fruit and vegetables at the market (from 4.8% to 23.0%, p = 0.004).
No other changes were reported.
Incentive: $10 gift cards were given to participants at T1 and T2. Participants: A total of n = 97 participated (n = 5 had missing data). Final T1 sample = n = 92. At T2 n = 47 or 51% of T1 participants completed the survey via telephone and n = 17 (36%) completed the survey via mail. Final longitudinal sample = n = 64. Time points: Intervention period was June-August 2010. T1 = May 2010. T2 = (2 months post-farm stand introduction) = July/August 2010.
Wrigley, N., et al. (2002) [45].g
Wrigley, N., et al. (2003) [46].h
Leeds, UK
Exposure: supermarket
To examine changes in food consumption and poverty after a sudden and significant change in food retail access.
To explore the impact of a significant change in food retail provision in a highly deprived area on food consumption patterns
Recruitment: from a local authority housing estate area (population 38,000 and ~15,000 households). From a low income, socially deprived, largely white ethnicity background area. The main household domestic food purchaser was recruited. Target sample size was n = 1000 at T1 and n = 600 residents at T2. A target of inviting 3000 households at T1 was set. Incentives: vouchers for non-food related outlets were provided to participants at T1 and T2. Participants: T1 respondents = n = 1009. T2 respondents = n = 615. Primarily female participants (81.9% at T1 and 84.1% at T2). The non-respondents at T2 had moved residence (9%), could not be contacted after four attempts (13%) refused further participation (13%) and returned data that was unsatisfactory for inclusion (4%). Subgroup analyses in: participants eating ≤2 portion of fruit and vegetables/day at T1 that did switch (n = 124) and did not switch (n = 115), participants eating >2 to <3 portions at T1 that did (n = 52) and did not switch (n = 82) and those eating ≥3 portions at T1 that did (n = 100) and did not switch (n = 142); participants that switched to using the new supermarket at T2 (n = 276) compared to those that did not (n = 339); participants that switched to using the supermarket from using limited-range/budget stores (n = 48), a specific major retailer store (n = 110), other major retailer stores (n = 99, of which n = 87 were the same chain as the new supermarket) and other stores (n = 19) at T1 were compared to each other. Study design: pre-post WO control.
Exposure: New food retail in neighborhood.
Comparison: Changes over time explored. (Sub-group exploration of: participants with poorest diets at baseline compared to others; those that switched to using the new facility compared to those that did not; those that stopped versus continued smoking; and those residing closer or further away from the facility).
Food consumption (self-report, 7-day diary)
Interviewer administered surveys
Distance travelled to the main food store in those that switched to using the supermarket decreased from 2.25 to 0.98 km (statistics NS) from T1 to T2. In those that had shifted to using the supermarket, walking to the store as a mode of transportation increased from 12.3% to 30.8% and walking from the store increased from 6.5% to 22.8% (reported as significant in text, p = NS) from T1 to T2. Those that switched to using the supermarket increased fruit and vegetable consumption by 0.23 portions per day from T1 to T2 (p = 0.034). Participants eating ≤2 portions of fruit and vegetables/day at T1 that switched to using the new supermarket, increased fruit and vegetable consumption from T1 to T2 (from 1.25 to 1.72 portions/day, p < 0.001), but so did those that did not switch that were eating the same amount of fruit and vegetables at T1 (from 1.37 to 1.78 portions/day, p < 0.001). Those eating ≥3 portions of fruit and vegetables at T1 that did not switch to the supermarket had a decrease in fruit and vegetable intake (4.78–4.20 portions/day, p = 0.005). An area (area name = LS14 1) effect was found with participants living in one postcode area on fruit and vegetable intake at T2 (NS if this negative), but appears so from the table). All 2SLS and parameter estimates and OLS estimates had the same signs, with greater significance in relationships between fruit and vegetable intake and switching to using the supermarket, proximity to the supermarket and switching to using the supermarket from a limited-range/budget store at T1.
Participants living ≤750 m (n = 176) were compared to those that lived >750–≤1000 m (n = 113) and those that lived >1000 m (n = 326) from the supermarket and those living ≤500 m (n = 65) to those living >500–≤1000 m (n = 224) were also compared; participants that stopped smoking (n = 20) compared to those that did not; and participants living in different area codes. Multivariate analyses were conducted in n = 598 participants as n = 17 participants had missing information.
Time points: Supermarket opening = November 2000. T1 (5 months before opening of the supermarket) = June–July 2000. T2 (7–8 months post-opening) = June–July 2001. The survey was piloted in February 2000. A repeatability survey was conducted during T2 of the main survey to examine random and systematic error extent (n = 140 households).
The effect of pre-intervention fruit and vegetable intake was less significant using 2SLS. Model 1 OLS parameter estimates and (SEs) for change in fruit and vegetable intake were: T1 fruit and vegetable consumption = −0.281 (0.034), p = 0.01); distance to supermarket ≤500 m = 0.440 (0.227), p = 0.05; switched to using supermarket from limited range/budget store in T1 = 0.386 (0.188), p = 0.05; and household within LS14 1 = −0.426 (0.160), p = 0.01. Model 2 OLS parameter estimates and (SEs) for change in fruit and vegetable intake were: T1 fruit and vegetable consumption = −0.282 (0.034), p = 0.01); and household within LS14 1 = −0.429 (0.159), p = 0.01. 2002 results: In those that had ‘poor’ diets at T1, fruit and vegetable intake increased by 60% (from 1.31 to 1.75 portions/day, with fruit/fruit-juice intake increasing by two-thirds nearly) and in those that had the ‘worst’ diets at T1, fruit and vegetable intake increased from 0.59 to 1.41 portions/day, with fruit/fruit-juice intake increasing five-fold (statistics NS). In those that were eating <1 portion/day at baseline, fruit and vegetable intake increased from 4.13 to 9.83 portions/week and fruit and fruit juice consumption increased from 0.77 to 3.92 portions/week, between T1 and T2. In those that were eating ≤2 portion/day of fruit and vegetables at baseline, fruit and vegetable intake rose from 9.17 to 12.25 portions/week and fruit and fruit juice intake increased from 2.82 to 4.59 portions/week (changes are stated as significant in text but significance values NS). In those completing T1 and T2 surveys, 45% switched to using the new supermarket as their main food retail source and 35% used the supermarket as their main fruit and vegetable source. In participants eating ≤2 portion/day of fruit and vegetables at baseline, 42% switched to using the new supermarket for fruit and vegetable purchasing. In participants eating <1 portion/day of fruit and vegetables at baseline, 70% switched to using the new supermarket for fruit and vegetable purchasing (significance of changes NS). No other significant changes were reported.

NS = not stated; IPAQ = International Physical Activity Questionnaire; MVPA = moderate to vigorous physical activity; WO = without; T1 = time point 1 (baseline); T2 = follow-up 1 (post-baseline); T3 = follow-up 2 (post-follow-up 2); SD = standard deviation; SE = standard error; ‘±’ = SD; ~ = approximately; cpm = accelerometer counts per minute; OR = odds ratio; CI = confidence interval. All outcomes collected are listed in the outcome measures column. Only significant changes are reported in the efficacy column, where significance is set at p < 0.05 (e.g., if an outcome measure is not included in this column then there was no significant interaction effect (for RCTs) or change over time (for pre-post studies) for that outcome. Superscripts a–h indicate papers relating to the same study. Aims of each individual paper are provided directly in line with the respective reference details, whilst other data extracted common to both papers are not aligned (e.g., recruitment strategies and sample characteristics are the same for these papers).