Trends in participant race and sex reporting in lung cancer phase III clinical trials

Abstract Background Clinical trials are an essential part of advancing care for cancer patients. Historically, however, racial minorities and females have been underrepresented in these trials. Efforts like the National Institute of Health Revitalization Act attempted to mitigate these disparities, but despite these efforts, they continue to exist. These disparities can subsequently lead to minorities and females receiving suboptimal care. Aims The purpose of our study was to understand the changing trends in reporting of participant race and sex as a demographic variable in phase III lung cancer clinical trials published over the last 35 years given these consequences of poor representation. Methods and results A total of 426 articles reporting the results of phase III lung cancer clinical trials published from 1984 to 2019 were identified in PubMed. From these articles, data on participant sex and race were collected from the demographic tables to construct the database for this study. This database was subsequently used to determine the rate of reporting of demographic factors like race and sex and the participation trends over the time of minority and female participation in lung cancer phase III clinical trials. The SciPy Stats package for Python was used to calculate descriptive statistics, 95% confidence intervals, two sample t‐test, one‐way analysis of variance test, and Pearson's correlation coefficients. The Matplotlib package for Python was used for figure generation. Only 137 (32.2%) of the 426 studies analyzed reported the race of participants. Among those studies, we found that the mean participation rate of White participants was significantly higher (82.65%; p < .001). We found a decrease in African American participants and an increase in Asian participants over time. When looking at sex, we found that although the rate of male participation (69.02%) was significantly higher than that of female participation (30.98%), female participation has improved with time at a rate of 0.65% per year. Conclusion We found that the reporting and participation of minority races continue to lag that of other demographic factors like sex in phase III clinical trials in lung cancer. Based on our analysis, we note a decline in participation of African Americans in lung cancer phase III clinical trials despite the rising incidence of lung cancer.


| BACKGROUND
Clinical trials are an essential means of improving the outcomes of cancer patients. Racial minorities and females have been underrepresented in clinical trials over time. Historically, such disparities have been a problem of multiracial societies, such as the United States, but with globalization and human migration, health disparities are becoming a global concern. In February 2020, the European Parliament in its report on health inequalities in the European Union identified addressing growing health inequalities as a priority. 1 The National Institute of Health (NIH) of the United States was tasked to address these growing challenges through the NIH Revitalization Act of 1993. An important mandate of the NIH Revitalization Act was adequate reporting on minority and vulnerable populations in clinical trials.
Despite these important efforts, various studies over the ensuing two decades continue to report persistent disparities in clinical trials. [2][3][4][5] These disparities in the representation of minorities and females in clinical trials can have significant consequences. Given these disparities, it becomes challenging to apply the findings of clinical trials to minority groups as they are not adequately represented in the trials themselves. 6 Thus, minorities and females may not be receiving the full benefits of scientific advancements as their White and male counterparts. 6 Another consequence of poor representation is that biological variants seen in minority groups are not adequately studied, creating biases in the definitions of what is "normal" and "abnormal." 7 The introduction of these biases is also dangerous as they are propagated into future studies and clinical practice, which can lead to suboptimal care for minority patients. 7 Thus, it is ever more important to assure clinical trials adequately represent the entire patient population.
As we enter the third decade of the 21st century, with increasing awareness and advancements in communications and global connectivity, an expectation would be that these disparities would be resolved. Given the subsequent consequences associated with such disparities, we set out to evaluate the landscape of race and sex reporting in phase III clinical trials in patients with lung cancer. We selected lung cancer for our study because, as the leading cause of cancer death in the United States, it has represented a significant proportion of phase III cancer studies (14%), and is seen across the entire spectrum of human races and sex. [8][9][10][11][12][13] The purpose of our study was to determine the historical trends and current landscape of reporting and representation of females and minorities in lung cancer phase III clinical trials. We chose to study lung cancer phase III clinical trials because lung cancer is the most common cause of cancer-related death and manifests differently depending on one's background; thus, adequate reporting and representation are essential. [8][9][10][11] Specifically, we looked at participation rates of different races and sex over a 35 years time period . In total, demographic data from 426 lung cancer phase III clinical trials were thoroughly reviewed, and information regarding rates of reporting and participation were obtained and analyzed from these trials. This study is significant because it identifies ongoing disparities in lung cancer phase III clinical trials and draws attention to the poor rates of reporting and representation of minority groups like African Americans despite high incidence rates of lung cancer.  For race, the number of White, Asian, African American, and Hispanic participants was recorded. The category "Other" was also used to include any participants who were listed as Other in the original study or did not belong to the White, Asian, African American, or Hispanic groups. The final category of "Unknown" was used to include any participants who were listed as Unknown in the publication. The number of participants in each category was then converted into the percent of the total participant pool each category represented and was then used to compare participation rates across studies. The studies that did not report demographic data on racial background were annotated as such. For sex, the total number of males and females was recorded. These values were subsequently used to calculate the percent of the participants' pool that was male or female, and the studies that did not report demographic data regarding sex were annotated as such. The data collected were then used to determine the percentage of studies that reported participant race or sex, differences in the rate of participation among racial groups and between sex, and how participation rates have changed over time.

| Statistical analysis
Statistical analysis was done using the SciPy stats package, and figure generation was done using the Matplotlib package for Python. 14,15 To determine the rate of reporting of race, the percentage of studies that reported race as a demographic and those that did not report race were calculated. Out of the studies that did report race, the percentage of those studies that reported on each racial group was also determined. The same method was also used to determine the rate of reporting of sex.
To determine whether disparities in clinical trials' participation exist, the mean participation rate of each race and sex across all the studies that reported race or sex was calculated, and the 95% confidence interval (CI) for the mean of each demographic was determined.
Then, to compare participation rates among different races, the percent of White participation from each study was compared with the percent participation of the remaining races (Asian, African American, Hispanic, Other, and Unknown) using a two-sample t-test, and a p-value <.01 was considered statistically significant. A one-way analysis of variance (ANOVA) test was also used to determine whether a statistically significant difference in participation among the Asian, African American, Hispanic, Other, and Unknown groups exists, and a p-value <.01 was considered statistically significant. To compare participation rates between sex, a two-sample t-test was used, and a pvalue <.01 was considered statistically significant.
Finally, to determine how the participation rate has changed over time, the participation rates of different races from each study were plotted on a scatter plot using the Matplotlib package. 15 Using those data points, the SciPy stats package was used to calculate Pearson's correlation coefficient and the p-value (with p < .05 considered statistically significant), and to draw a line of best fit. 14 The data points for White participation were plotted with each of the remaining races to serve as a comparison as to how participation rates have changed with time. F I G U R E 1 CONSORT diagram illustrating the study selection process.
demographic ( Figure 2A). The remaining 289 studies (67.8%) did not report race. Interestingly, we found that from all the studies we analyzed, no study published before 1992 reported demographic data on patient race ( Figure 2C). After 1992, more studies started reporting race ( Figure 2C). However, race reporting did not appear to improve over time since 1993 ( Figure 2C) Figure 3A).
When looking at how participation rates changed with time, we found that the rate of White participants decreased at a rate of 0.51% per year (r = À0.23, p < .01) ( Figure 3B). This decrease in participation was accompanied by an increase in the rate of Asian participants by a rate of 1.07% per year (r = 0.279, p < .01) ( Figure 3C). However, the rate of African American participants decreased at a rate of À0.30% per year (r = À0.34, p < .001) ( Figure 3D). The rates of participation among Hispanics (r = À0.01, p = .97), the other category (r = À0.04, p = .65), and the unknown category (r = 0.14, p = .55) did not change with time ( Figure 3E-G), although we did not have enough studies

| Rates of sex reporting and participation trends
From the 426 studies that were analyzed, 420 studies (98.8%) reported sex as a patient demographic ( Figure 4A). Among those 420 studies, the mean rate of participation for males was 69.02% (95% CI 67.66%-70.38%), and the mean for females was 30.98% (95% CI 29.62%-32.34%), yielding a statistically significant difference in participation rates between males and females ( p < .001) ( Table 2; Figure 4B).
When looking at the change of participation rates over time, we found that the disparity between male and female participation has significantly improved. The rate of male participation has decreased by a rate of 0.65% a year since 1985 (r = À0.31, p < .001), and the rate of female participation has increased by a rate of 0.65% a year since 1985 (r = 0.31, p < .001) ( Figure 4C). Moreover, we found that the mean rate of participation from 2015 to 2019 for males was 65.21%, and the mean rate of participation for females was 34.79%, which is markedly improved from the mean rate of participation from 1985 to 1995 when the mean participation rate was 73.60% for males and 26.40% for females (Table 2).

| DISCUSSION
Our study is the first one to look at the reporting trends in race and sex in phase III lung cancer trials over the last 35 years.  There can be several explanations for this decline in African American participation. One reason can be attributed to historical influences relating to novel research. 6

| CONCLUSIONS
In conclusion, we found that disparities in reporting and participation of minority racial groups continue to exist. We believe that requiring reporting of patient demographic data in clinical trials, in compliance with the NIH Revitalization Act, will improve transparency and the rate of reporting. Furthermore, it will critically inform health policymakers and other stakeholders to address the root causes of the underrepresentation of minorities by addressing social barriers and hesitancy to participate in clinical trials. 23 AUTHOR CONTRIBUTIONS Faaiq N. Aslam: Conceptualization (equal); data curation (equal); formal analysis (equal); investigation (equal); methodology (equal); visualization (equal); writingoriginal draft (equal). Rami Manochakian: Conceptualization (equal); data curation (equal); formal analysis (equal); methodology (equal); writingreview and editing (equal).

ACKNOWLEDGMENTS
Not applicable.

CONFLICT OF INTEREST STATEMENT
The authors have stated explicitly that there are no conflicts of interest in connection with this article.

DATA AVAILABILITY STATEMENT
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

ETHICS STATEMENT
Not applicable.