Reporting of Participants' Sex, Race, Ethnicity, and Socioeconomic Status in Pituitary Surgery Literature

Abstract

Introduction  Social determinants of health (SDOH) are associated with differential outcomes after pituitary tumor treatment. However, the specific impact of SDOH is not well characterized. One reason may be the lack of collection and reporting of sociodemographic variables in the literature. This study aims to evaluate the frequency of reporting and distribution of participants' sex, race, ethnicity, income, and education level within pituitary surgery literature. We will compare the reported clinical research population demographics to the 2020 U.S. census.

Methods  A systematic review was performed by searching PubMed, Cochrane, and Embase databases for pituitary surgery clinical research published between July 1, 2021 to June 30, 2022. We excluded studies that lacked a comparison group, were not original research (i.e., systematic reviews, meta-analysis), or included national databases and registry data.

Results  The final analysis included 92 studies. A total of 99% of studies collected data on subject sex. On average 49% (range: 14–100%) of study populations were male. Only 4% ( n  = 4) studies included racial demographic data. Two studies included information on participants' ethnicity and two included education background. No studies included income or insurance data. Four U.S. studies included demographic distribution, and the reported race and ethnicity percentages are similar to the U.S. 2020 census distribution.

Conclusion  Most clinical pituitary research collects and reports data on participant sex. However, very few studies collect and report data on other sociodemographic variables that can play a role in outcomes. The lack of sociodemographic information in clinical research literature makes it difficult to determine the role of SDOH on pituitary surgery outcomes.

Introduction

Social determinants of health (SDOH) have repeatedly been shown to impact disease burden, effectiveness of medical interventions, and patient outcomes. Research has shown that sociodemographic factors, such sex, race, ethnicity, income, and education level, influence health outcomes and can lead to health disparities. Both the collection and reporting of study population demographics is critical, as this can impact the generalizability of research findings to diverse patient populations. Thus, organizations such as the National Institute of Health, ¹ U.S. Food and Drug Administration, ² and Agency for Healthcare Research and Quality ³ put into place guidelines to promote inclusion and diversity of recruitment in clinical research studies.

Previous work looked at the rate of reporting of SDOH variables within different disciplines and within various study design types. Orkin et al looked at 10% of all randomized clinical trials (RCTs) published between 2014 and 2020 in five major medical journals (237 studies were randomly selected from 2,351 RCTs) and found that 98.7% reported sex or gender and 48.5% reported race/ethnicity; however, only 14.3% reported education level or literacy, and income was rarely reported. ⁴ Stadeli et al looked more closely at surgical research published in four surgical journals in 2016 and found that 98% reported sex, 50% reported race, 27% reported ethnicity, and 24% reported education level. ⁵ Studies in ophthalmology and orthopedics reviewed their field's literature from top journals over a 1-year period and also found that sociodemographics were routinely underreported. ^{6 7}

Studies evaluating reporting of sociodemographics has not been performed in the pituitary surgery literature. Research has demonstrated disparities in complication rates in pituitary surgery based on race, ethnicity, socioeconomic status (SES), insurance status, and hospital factors. ⁸ However, there remains a paucity of robust data. The goal of this research is to assess the current rate reporting SDOH variables in clinical pituitary surgery research. We aim to accomplish this goal through a 1-year cross-sectional study to evaluate routine research reporting practices for sociodemographic variables in the recent pituitary surgery literature. Specifically, we will calculate the reporting rate of participants' sex, race, ethnicity, and SES and will identify any gaps in published clinical research reporting practices. Additionally, we will compare the reported research population demographics to the 2020 U.S. census and will map out the literature to evaluate any differences that may exist between different study designs or between international and U.S. studies.

Methods

This systematic review was conducted with assistance from a senior medical librarian using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews.

Literature Search

The protocol was designed to identify clinical research studies focused on the comparison of clinical outcomes related to pituitary surgery. In this review, literature searches were conducted using the following databases: Embase (Elsevier), PubMed (U.S. National Library of Medicine, National Institute of Health), and Cochrane. Below is the search strategy for the Embase database. The other two search strategies can be found in Supplementary Fig. S1 , available in the online version.

Search Strategy Embase

“pituitary” AND “surgery” AND [humans]/lim NOT (“meta analysis”/de OR “systematic review”/de OR “review”/de) AND [2021–2022]/py AND [01–07–2021]/sd NOT [01–07–2022]/sd

Selection Criteria

Studies were limited to those published in the English language as full-text articles. We restricted the search to a 1-year period between July 1, 2021 and June 30, 2022. The search was run in August 2022. The study population included patients aged 18 years and older undergoing pituitary surgery. Primary clinical research studies with a control or comparison group were included. Exclusion criteria were as follows: fewer than 10 subjects per cohort, national or international database studies, lack of clinical outcomes, systematic reviews, and case reports.

Selection of Studies and Data Extraction

At least two reviewers screened the abstracts and titles of articles retrieved by our literature searches using the inclusion and exclusion criteria (A.N. and J.A.P.). After screening, we obtained full-text copies of all articles deemed to satisfy screening criteria. Each full-text article was then assessed by two reviewers, one of them being a senior author (R.G., or A.M.R. and C.G.L.). Discrepancies in whether the article should be included was resolved by the senior author (C.G.L.). The Covidence Systematic Review software (Melbourne, Australia) was utilized for streamlining the process of article selection. Once full-text articles meeting inclusion criteria were identified, data were independently extracted (A.N. and J.A.P.). Extracted variables included data on total study population and specific group demographics (sex, race, ethnicity) and SES, which was measured by education level and self-reported household income. We extracted data on whether these SDOH variables were included in the study analysis. We also noted if justification was provided in the study text regarding the decision to make adjustments or include the SDOH variable in the study analysis. All studies were categorized as RCTs, cohort studies, or case–control studies. An additional round of confirmation of all extracted data was performed prior to analysis (A.N.).

Sociodemographic Variable Reporting Analysis

Data were collected and analyzed in Microsoft Excel. For each study, we calculated the presence of reporting (yes/no) for each sociodemographic category (sex, race, ethnicity, and SES). We used descriptive comparisons of the overall reporting frequency to compare U.S. with international studies and to evaluate for reporting differences between study designs. Assessment of the rate of reporting of SDOH variables were compared between all studies, in U.S. studies, and between study design types (RCTs, cohort studies, and case–control studies). Weighted mean frequencies of demographic data were calculated to account for the proportion of the total cross-sectional research population contributed by each individual study. Demographic sex, racial and ethnic weighted mean distributions of the study populations in U.S. studies were compared with 2020 U.S. census.

Comparison to Census Data

Census data were obtained from 2020 U.S. census. ^{9 10 11} Weighted mean frequencies of demographic strata were calculated to account for the proportion of the total cross-sectional research population contributed by each individual study. Weighted mean frequencies were compared with census data.

Results

The initial search after duplicates were removed retrieved 2,848 articles. After examining titles and abstracts for exclusion criteria, we reviewed 191 articles in full. We determined that 99 did not meet inclusion criteria. Ninety-two articles were included in the final analysis, 14 U.S. studies and 78 international studies. Fig. 1 shows the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flow diagram depicting the process of article selection. Each included study was categorized based on type of study design. There were in total 83 cohort studies (90%), 5 RCT (5%), and 4 case–control studies (4%). Of the U.S. studies, there were 11 cohort (79%), 1 RCT (7%), and 2 case–control (14%) studies. Fig. 2 shows the distribution of studies based on study design types.

Fig. 1.

PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flowchart of study inclusion. A total of 3,451 studies were identified in the original search, and 92 studies were ultimately included in the final analysis based on inclusion and exclusion criteria.

Fig. 2.

Distribution of studies based on study design types in all studies and U.S. studies. The majority of studies are cohort studies. RCT, randomized clinical trial.

Study Participants

The average number of study participants in the 92 studies was 184 patients with a total of 16,722 participants in all studies (median 101, range: 34–1118). For the 14 U.S. studies, the average number of study participants was 270 with a total of 3,778 participants in the studies (median: 178, range: 36–987). This is summarized and subclassified based on study design type in Table 1 . Cohort studies on average had over twice as many study participants compared with RCTs.

Table 1. Population distribution. Average number of participants and total number of participants in included studies based on study design types. (A) All studies ( n  = 92), (B) U.S. studies ( n  = 14) .

A
All studies	All studies	Cohort	RCT	Case–control
Average # participants	184	194	67	150
Total # participants	16,722	15,720	401	601
B
U.S. studies	All studies	Cohort	RCT	Case–control
Average # participants	270	301	113	180
Total # participants	3,778	3,306	113	359

Abbreviation: RCT, randomized clinical trial.

Sociodemographic Data Collection

Nearly every study (91/92) specified sex distribution; however, 13 studies (14%) did not report sex distributions within each comparison group. Four studies (4%) included demographic information on race, although one paper did not include information on subgroup distribution. Study population information on ethnicity and education were included in two studies (2%) each. None of the papers included data on income. Fig. 3 shows the breakdown of studies that reported each of the five basic sociodemographic variables: sex, race, ethnicity, and SES (which was grouped as income and education level). The distribution of the reported participant demographic information was similar in the U.S. studies compared with all studies.

Fig. 3.

Sociodemographic distribution in all pituitary studies and U.S. pituitary studies. Percentage and number of all studies and U.S. studies that included sociodemographic variables (sex, race, ethnicity, and socioeconomic status [SES]). SES includes income and education level. Almost every study included sex data, whereas few studies included race, ethnicity, and SES data. Left: All studies ( n  = 92), Right: U.S. studies ( n  = 14).

Sex Data

Among the pooled population from all studies, the weighted average was 49% of the participants were male with a range of 14 to 100% and median of 50%. Only one study (Fang et al 2021 ¹² ) included 100% male participants. In the U.S. studies, on average 47% of the participants were male with a range of 14 to 60% and median of 49%. We calculated the weighted mean frequency to account for the proportion of subjects contributed by studies of different sizes. The weighted averaged for cohort, RCT, and case–control studies were 49, 46, and 47%, respectively, for all studies and 47, 44, and 46%, respectively, for U.S. studies. There is slightly higher percentage of male participants in international studies compared with the U.S. studies, reflecting the slight difference between U.S. and world averages based on the 2020 U.S. census. Fig. 4 depicts the sex distribution and median of male study participants in all studies and U.S. studies compared with the U.S. (49%) and world (50%) average.

Fig. 4.

Distribution of male population. Percentage of male study participants in all studies ( n  = 92, median = 50%, weighted average = 49%, range = 14–100%) compared with U.S. studies ( n  = 14, median = 49%, weighted average = 47%, range = 14–60%) as compared with 2020 U.S. census (49%) and world census data (50%, from the World Bank data; worldbank.org). ^{9 20} Dotted line at 50%, horizontal lines represent median values.

Race and Ethnicity Data

Five out of 92 studies (5%) reported race and ethnicity data of study participants. One paper by Guo et al, published in China, distinguished participant ethnicity as “Han nationality” versus “other nationality” for each subgroup ¹³ . The remaining four papers were U.S.-based studies. Two papers (Lee et al ¹⁴ and Snyder et al ¹⁵ ) only included distribution of White participants; furthermore, Lee et al only included the race breakdown in one of their subgroups. Ioachimescu et al reported the distribution of White versus Black participants in both subgroups ¹⁶ . Little et al was the only paper to include a complete racial and ethnic distribution of all 113 participants. ¹⁷

One additional paper (Arshad et al ¹⁸ ) noted that information on ethnicity was collected, but that data were not reported in the printed publication. Fig. 5 depicts the frequencies of the study participants in the four U.S. studies with reported demographic data and compared them with the demographic distribution in the U.S. based on the 2020 U.S. census. The distributions were very similar. Of note, most RCTs (⅘ studies) did not elaborate on patient demographics such as race and ethnicity.

Fig. 5.

Race and ethnic distributions were included in 4 of the 14 U.S. studies. The pooled population of study participants in these 4 U.S. studies is compared with the 2020 U.S. census data. Four studies included distribution of White participants, whereas only one study included distribution of Asian and Hispanic participants. “ n ” represents number of studies that reported demographic data .

Education Level Data

Two papers included education-level data. Both papers were by Cao et al, evaluating cognition and attention processing in patients with pituitary adenoma ¹⁹ . The authors collected and matched subgroups for the patient's highest education level. None of the RCTs included information on education or health literacy.

Sociodemographic Data Adjustment

Lastly, we looked at whether studies adjusted for differences in sociodemographic variables between groups. Data adjustment evaluation means that a variable was used to adjust for differences in reference values, such as when endocrinological factors were used between male and female participants, and when sociodemographics was used as a variable in the analysis, such as in a logistic regression analysis. Overall, only 12 studies (13%) adjusted the analysis for one or more demographic measures. Almost every adjustment that was made was for endocrinological assessment of study participants using sex-adjusted insulin-like growth factor 1 or growth hormone levels. An additional 24 papers (26%) used sex as part of their analysis, including it as a variable in the multiple variable analysis. Five studies (5%) noted that they matched groups based on at least one sociodemographic variable. The distribution of the number of studies that matched groups for SDOH variables, adjusted for one or more SDOH variables, and incorporated sex as a variable in the analysis was similar in the U.S. studies. This is shown in Table 2 . Supplementary Table S1 (available in the online version) outlines outcome measures, patient population demographics, and adjustments investigated in each study in this review.

Table 2. Incorporation of sociodemographic variables in analysis and matching of (A) all studies and (B) U.S. studies.

A
All studies	RCT, 5 studies ( n  = 401)		Cohort, 83 studies ( n  = 15,720)		Case–control, 4 studies ( n  = 601)		All, 92 studies ( n  = 16,722)
	Studies	Participants	Studies	Participants	Studies	Participants	Studies	Participants
Adjusted for one or more identified differences in cohorts	1 (20%)	60	11 (13%)	3,188	0	0	12 (13%)	3,248
Used sex as variable in analysis	0	0	24 (29%)	4,856	0	0	24 (26%)	4,856
Matched groups	1 (20%)	113	4 (4.8%)	309	0	0	5 (5.4%)	422
B
U.S. studies	RCT, 1 study ( n  = 113)		Cohort, 11 studies ( n  = 3,306)		Case–control, 2 studies ( n  = 359)		All, 14 studies ( n  = 3,778)
	Studies	Participants	Studies	Participants	Studies	Participants	Studies	Participants
Adjusted for one or more identified differences in cohorts	0	0	2 (18%)	533	0	0	2 (14.3%)	533
Used sex as variable in analysis	0	0	4 (3.6%)	1,310	0	0	4 (29%)	1,310
Matched groups	1 (100%)	113	0	0	0	0	1 (7.1%)	113

Abbreviation: RCT, randomized clinical trial.

Discussion

The purpose of this scoping review was to investigate the frequency of reporting of sociodemographic variables within pituitary surgery literature and match the observed distribution with the 2020 U.S. census. Ninety-two studies were included in this review, 14 of which were U.S. studies and 78 were international studies. We found that the vast majority of studies collected and reported sex information, whereas only five studies included racial and/or ethnic demographic data. Noteworthy, most of these five studies had incomplete reporting, such as only reporting only one race or not including the racial breakdown in all comparison groups. Two studies reported education level, and no studies included income data.

Reporting sociodemographic data of study populations in research publications is increasingly emphasized. The International Committee of Medical Journal Editors (ICMJE) reporting guidelines serve to promote quality reporting of race in medical journals. Recently, Maduka et al ²¹ reviewed surgical literature adherence to ICMJE and found that frequency of reporting continues to be low among most medical journals and surgical literature and called for stricter and more standardized guidelines for journal publication to adhere with these guidelines.

Several studies have shown that SDOH, including sex, race, ethnicity, education level, and income, are independent predictors of surgical management and outcomes. ^{5 6 7 8} Many of the studies included in this scoping review used postoperative complications, such as incidence of cerebrospinal fluid leaks and diabetes insipidus, and hospitalization metrics, such as readmission rates, as the primary outcome measures ( Supplementary Table S1 , available in the online version). Goljo et al found that the likelihood of postoperative complications was higher in Black and Hispanic patients compared with White patients. ⁸ Tiwari et al found that Hispanic ethnicity and patients living in more disadvantaged neighborhoods had a higher incidence of postoperative diabetes insipidus. ¹⁴ McKee et al and Parasher et al both found that Black race and Hispanic ethnicity were predictors of prolonged hospitalizations. ^{23 24} McKee et al additionally found that Asian race and female sex were predictors of readmissions. Furthermore, Deb et al found that Caucasian patients and patients with higher SES had a higher likelihood of receiving surgery. ²⁵ Thus, it is likely that there is a selection bias in the patient population recruited for these pituitary surgery studies.

Sex Collection and Reporting

Many pituitary tumors secrete hormones (i.e., functional pituitary tumors), and the distribution can be different between males and females; thus, it is imperative to collect sex data to minimize the risk of confounding. The sex epidemiology of pituitary tumors, including pituitary adenomas, pituitary carcinomas, and uncertain behavior pituitary adenomas, was evaluated in a SEER (Surveillance, Epidemiology, and End Results) database study by Chen et al. ²⁶ They used data from 2004 to 2016 and noted a higher incidence rate in female patients compared with males (5.3 cases per 100,000 person-years vs. 4.3 cases per 100,000 person-years). Females, additionally, had a bimodal age-related distribution with a first peak in 25 to 34 years and a second peak in 60 to 69 years, whereas males had a unimodal age-related peak in the sixth decade of life. Gittleman et al similarly found that the annual age-adjusted incidence rate was 2.94 in males and 3.40 in females in 2009. ²⁷ One theory is that this sex difference is that for pathologies such as prolactinomas cause hormonal shifts with more clear-cut clinical presentation in females, like amenorrhea and galactorrhea, resulting in earlier disease presentation for evaluation and diagnosis.

In this scoping review, we found that while almost every study reported sex data, a proportion of these studies (14%) did not report sex distribution for all subgroups. Without specific subgroup data, it is not possible to determine proper matching of cohorts or to determine if analysis of a given demographic variable confounded the results. Additionally, most studies did not include sex as a variable in analysis. Of note, one study (Fang et al) included 100% male participants with nonfunctional pituitary adenomas and noted in their discussion that this was done intentionally to prevent “errors [found] in the assessment of female hormones.”

Other Social Determinants of Health Variables Collection and Reporting

Our review highlights that some sociodemographic variables, such as race, ethnicity and education level and income, are much less frequently reported. When sociodemographics are not conveyed, the assumption is often that the dominant ethnicity and race make up for the majority of the population and gives the appearance of an overwhelmingly homogenous population. The weighted mean average population of the 4 U.S. studies that reported race and ethnicity demographics showed a similar distribution when compared with the 2020 U.S. census. However, some studies ( n  = 2) reported only the frequency of White participants. While efforts were made in several papers to include nonsex sociodemographic variables, it often lacked consistency and completeness.

We also noted differences in rate of reporting among different study designs: cohort, RCT, and case–control studies and found that there was not a significantly different distribution. This is noteworthy as RCTs require randomization of subgroups, and an important technique to show that randomization has been successful is to demonstrate equal distribution of demographic data between groups. Therefore, especially in RCTs, demographic differences may confound outcomes and should be reported and adjusted for during subgroup analysis.

Limitations

This study has limitations. It is possible that there was reporting bias, where study authors collected sociodemographic variable data but did not publish, possibly due to character constraints or feeling that the data were irrelevant. Furthermore, this review may have overestimated the number of studies that collected sociodemographics by including studies with any data on sex, race, ethnicity, and SES, even if the studies did not provide the frequencies for all strata or comparison groups.

Future Directions

Given the implementation of the electronic medical record, basic sociodemographic variables are collected and readily available in the chart of patients in the U.S. health care system and many systems internationally. Collecting this information should not require a researcher to significantly change their study design. Regulatory and editorial standards could help ensure that clinical researchers extract sociodemographic information and report it, transparently, within their work. This will not only contribute to a more robust analysis, but also this information can be used in future research to better understand the role of sociodemographic variables in pituitary surgery outcomes.

Conclusion

This systematic review found that most studies include sex data. However, few studies report data on other sociodemographic variables, such as race, ethnicity, income and education level, introducing issues with generalization and potential confounding. When sociodemographics were reported, they were often incomplete and not used as relevant variable. This gap in reporting key sociodemographic data may restrict our understanding about the potential influence of sociodemographics on pituitary surgery outcomes and can limit research validity and generalizability.