Study Sample

Using ClinicalTrials.gov, we constructed a database of US nAMD trials initiating enrollment in 2017 or later. ClinicalTrials.gov is the largest public web-based international registry of industry and non-industry-sponsored clinical trials. In accordance with the legal mandate of the Food and Drug Administration (FDA) Amendments Act of 2007, all clinical trials evaluating FDA-regulated therapeutics must be registered on ClinicalTrials.gov. Subsequent policy by the National Institutes of Health (NIH) has since expanded the registration mandate to include all NIH-sponsored investigations. We limited our analysis to studies beginning in 2017 or later to ensure a more universal sampling of trials, given recent tightening of registration enforcement stringency beginning in 2017.

We searched ClinicalTrials.gov to identify registrations of clinical trials studying “eye diseases” and then further limited our analysis to trials listed as evaluating “neovascular AMD,” “wet AMD,” or “exudative AMD.” We excluded registrations of observational studies, in addition to trials that were not yet recruiting, suspended, terminated, or withdrawn. International clinical trials without US locations were additionally excluded.

Data Collection

For each study, we collected data on enrollment status, intervention, phase, sponsorship, and number and location of trial sites. For each trial site, we determined the exact geographic coordinates of latitude and longitude.

Using the 2014-2018 US Census Bureau’s American Community Survey, we identified all census tracts (N = 72,450) in the contiguous 48 United States. Census tracts are unique geographically defined regions, established by the US Census Bureau, comprising approximately 6,000-11,000 residents; they are commonly employed in population-based statistical analyses because their population characteristics and composition remain relatively static over time. Census tracts in Alaska and Hawaii were excluded because air or water transport is often required to access trial sites in these states, which is beyond the scope of our road travel analysis.

For each census tract, we collected and categorized data on age (<18 years, 18-64 years, and ≥65 years), race/ethnicity (white, black/African American, Hispanic, American Indian/Alaskan Native, Asian, Native Hawaiian/Pacific Islander, or other), education level (less than high school, high school completion, some college, associate’s degree, or bachelor’s degree or higher), health insurance status, and income level (less than 200% of the federal poverty level by quartile). Census tracts are assigned Rural-Urban Commuting area (RUCA) codes by the US Department of Agriculture on a scale of 1-10 based on population density and commuting patterns, with “1” designating a “metropolitan core” and “10” a “rural area.” We categorized RUCA codes for each census tract as follows: urban (RUCA of 1-3) or rural (RUCA of 4-10). We further grouped census tracts by region as follows: Northeast, Midwest, South, and West.

Main Outcome Measures

The primary outcomes of our study were road travel distance (in miles) and travel time (in minutes) to the nearest nAMD clinical trial site. Using ArcGIS Pro 2.6 (Esri, Redlands, California, USA; 2020), we mapped the population-weighted geographic centroids of each census tract. The origin-destination cost-matrix analysis tool in ArcGIS finds the quickest route using street-level data between origins and destinations based on standard traffic patterns and using a personal vehicle. Using this software, we determined travel distance and travel time from each census tract centroid to the single nearest clinical trial site. Travel distance was categorized as follows: <30 miles, ≥30 but ≤60 miles, >60 miles but ≤120 miles, and >120 miles. Travel time was categorized as follows: <60 minutes or ≥60 minutes.

We examined the distribution of demographic and socioeconomic characteristics of US census tracts based on travel distance and travel time to identify census tract characteristics associated with greater travel burden.

Statistical Analysis

We used descriptive statistics to characterize census tracts and clinical studies. Univariate and multivariate logistic regression analyses were performed to examine associations between variables of interest and the outcomes of traveling more than 60 miles and traveling longer than 60 minutes to the nearest clinical trial center. Variance inflation factors were used to determine multicollinearity among covariates. All covariates that were significant in univariate analyses at the P = .25 level were included in the multivariate model. Variables either significant at the P = .05 level in the multivariable model or whose removal caused greater than a 20% change in the coefficient of at least 1 other variable were allowed to remain in the multivariate model. All modeling was performed in SAS 9.4 using PROC GENMOD; all models were weighted by each census tract’s population and clustered by state.