Demographic Factors Associated With Non-Guideline–Based Treatment of Kidney Cancer in the United States

This cohort study investigates the association of sex, race/ethnicity, and insurance status with undertreatment and overtreatment of kidney cancer.


Introduction
The past several decades have seen an increased incidence and downward stage migration of kidney cancer. These phenomena have been driven by incidental detection of small kidney masses on nonurologic cross-sectional imaging, in what one set of authors described as "nonsystematic partial screening of the population." 1 It has been recognized that small (ie, <4 cm) tumors in the kidney often have low metastatic potential and that patients with these tumors are often more likely to die of competing causes. 2,3 As a result, major organizations have issued guidelines favoring nephronsparing approaches (ie, partial nephrectomy or ablation) or active surveillance in patients with small kidney masses. [4][5][6] Although the overtreatment of clinically indolent lesions is a recognized and important phenomenon in urologic oncology, increasing attention has also been drawn to undertreatment of clinically aggressive cancers, particularly among underserved racial/ethnic groups and among patients who are uninsured or underinsured. [7][8][9] However, the association between patient demographic characteristics and clinical decision-making in kidney cancer remains poorly studied. In this study, we sought to assess the association between demographic factors (including sex, race/ ethnicity, and insurance status) and receipt of non-guideline-based kidney cancer treatment in a comparatively young and healthy cohort of patients with localized kidney cancer and low competing risk of mortality.

Methods
This was a retrospective cohort study using data from the National Cancer Database (NCDB). The NCDB is a joint project of the Commission on Cancer (CoC) of the American College of Surgeons and the American Cancer Society. It is a hospital-based registry containing data from more than 1500 CoC-accredited hospitals in the United States and includes approximately 70% of all new cancer diagnoses in the United States. 10 This study was exempted from review by the University of Texas Southwestern Medical Center (UTSW) institutional review board (IRB) owing to the use of deidentified patient data. Informed consent was waived by the UTSW IRB owing to the minimal risk to patients and the inability to identify specific patients in the deidentified data set. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline was followed.

Study Population
The NCDB contained 339 641 patients with a primary diagnosis of kidney cancer (using International Statistical Classification of Diseases and Related Health Problems, Tenth Revision [ICD-10] site code C64.9) from January 1, 2010, to December 31, 2017. These dates were chosen to coincide with the implementation of the 2009 American Urological Association (AUA) guideline on management of small kidney masses. 11 Of these patients, 214 763 patients were aged 30 to 70 years and had a Charlson-Deyo Comorbidity Index (CCI) score of 0 or 1. We excluded patients who had documented metastatic disease or were missing information on metastatic status; we then performed multiple imputation of missing values as described in the eAppendix in the Supplement. (Missing values are listed in eTable 1 in the Supplement.) After additional exclusions, the final study population consisted of 158 445 patients with cT1-2, N0, M0 kidney cancer (eFigure in the Supplement). Cancer staging was according to the seventh edition of the American Joint Committee on Cancer TNM system. 12 larger masses. Patients were assigned to 1 of 4 tumor classifications based on tumor size and clinical stage data. Treatments were categorized as observation or surveillance (including any patient with Ն6 months elapsed from diagnosis to treatment), ablation (eg, cryosurgery or thermal ablation), partial nephrectomy, and radical nephrectomy. Patients were then categorized as receiving guideline-based treatment, undertreatment, or overtreatment according to the scheme shown in eTable 2 in the Supplement. These designations were intended to allow broad categorization of treatment choices among patients without implying judgment of appropriateness for any individual patient.

Statistical Analysis
The primary outcome of interest was receipt of guideline-based or non-guideline-based treatment (ie, undertreatment or overtreatment). Rates of non-guideline-based treatment were assessed across demographic and institutional variables using the Pearson χ 2 test for categorical variables and Kruskal-Wallis 1-way analysis of variance on ranks for continuous variables. A multinomial logistic regression model was fitted for the outcome with all demographic and institutional factors as independent variables. Finally, subgroup analysis was performed by stratifying the study population by tumor classification (as used to define guideline-based treatment) and performing binary logistic regression. This allowed us to account for differential distribution of tumor size across populations.
Regression models included the following covariates: age, year of diagnosis, sex, race/ethnicity, CCI score, insurance status, facility type, facility location, facility annual patient volume (by quartile), urban vs rural location, distance traveled to facility (by quartile), and median income and educational attainment of patient's home zip (postal) code. Patient sex and race/ethnicity were abstracted from medical records by the participating institutions and were therefore dependent on documentation practices of individual treating institutions, which could not be determined with certainty. Race/ ethnicity was assessed because this information is required by the NCDB. Multicollinearity was tested

Patient Characteristics
Patient characteristics and receipt of guideline-based treatment are presented in Table 1. By intent, our population represented a comparatively young and healthy cohort among whom advanced age and comorbidity were unlikely to be associated with treatment decisions. The median (interquartile  Table 1).

Sex
Multinomial logistic regression analysis of demographic and institutional factors associated with receipt of non-guideline-based treatment is shown in Table 2. In the overall study population, women were treated more aggressively than men, with statistically significantly lower adjusted odds

JAMA Network Open | Urology
Demographic Factors Associated With Non-Guideline-Based Treatment of Kidney Cancer in the United States  (Figure).

Insurance Status
Compared with having private insurance, there was a significant association between uninsured status and receiving less aggressive treatment. Uninsured status was associated with statistically significantly higher adjusted odds of undertreatment (OR, 2.63; 95% CI, 2.29-3.01; P < .001) and lower adjusted odds of overtreatment (OR, 0.72; 95% CI, 0.67-0.77; P < .001) ( Table 2). In subgroup analysis, these associations were significant for overtreatment of small cT1a masses (OR, 0.40; 95% CI, 0.33-0.49; P < .001) (Figure) and  Table 2). The higher adjusted odds of overtreatment appeared to be associated The following covariates were included in the regression models but are omitted here for brevity: age, year of diagnosis, Charlson-Deyo Comorbidity Index score, facility type, facility location, facility annual patient volume, urban vs rural location, distance traveled to facility, income, and educational attainment. Odds ratios (ORs) from the primary multinomial regression analysis of the overall population are included for comparison. with higher adjusted odds of overtreatment of large cT1a masses (OR, 1.20; 95% CI, 1.14-1.26; P < .001).

Sensitivity Analysis
As a sensitivity analysis, the primary multinomial logistic regression was repeated on a separate, nonimputed data set in which patients with incomplete data on tumor stage and size or treatment were excluded (eTable 4 in the Supplement). Subgroup analyses were repeated with the inclusion of tumor size as a continuous covariate (eTable 5 in the Supplement). Similar outcomes were observed as in the primary analysis.

Discussion
This cohort study found that female, Black, and Hispanic patients had higher adjusted odds of non-guideline-based treatment for localized kidney cancer and, notably, that Black and Hispanic patients had higher adjusted odds of both undertreatment and overtreatment. The diagnosis and management of kidney cancer has undergone dramatic shifts in the past several decades. Widespread use of cross-sectional imaging has led to an increase in incidental diagnosis of early stage kidney cancers, and patient outcomes have improved over this period in association with stage migration as well as new systemic treatments for advanced disease. 1,13,14 However, it is increasingly recognized that overtreatment of small, indolent kidney masses is associated with significant morbidity in itself and may be harmful to patients. 2,3,15 Robotic-assisted surgical technology has been widely adopted and is associated with decreased treatment morbidity and increased use of nephron-sparing approaches; however, the availability of robotic technology may itself be associated with overtreatment of small kidney masses. [16][17][18] Despite the increasing attention devoted to overtreatment of indolent malignant tumors in urologic oncology, many patients with aggressive disease are undertreated, particularly members of racial/ethnic minority groups. Numerous studies have found adverse outcomes among patients with kidney cancer who are Black or members of other racial/ethnic minority groups compared with White patients. 7,9,[19][20][21] Two 2020 studies 22,23 found that Black patients with kidney cancer were less likely to be referred to high-volume surgical centers than other patients and that hospitals serving patients from racial/ethnic minority groups were less likely to offer surgical treatment. Similar phenomena have been described in other genitourinary malignant tumors. For example, in 1 series, young and healthy Black and Hispanic patients with high-risk prostate cancer were less likely than White patients to be treated with curative intent. 24 In a series of older patients with prostate cancer, Black patients were more likely than White patients to have delayed treatment and to incur higher treatment costs and had higher odds of readmission. 25 In a state-level series, Black patients with bladder cancer were more likely to be treated at low-volume centers, although this disparity did decline over time with increasing centralization of care. 26 Disparities in cancer treatment have also been associated with uninsured or underinsured status; for example, patients with kidney cancer who were uninsured or had Medicaid were more likely to present with advanced disease, less likely to receive definitive treatment, and more likely to die of disease than those with private insurance. 8 Additionally, although disparities in cancer outcomes between the sexes are often attributed to underlying biological factors, there may be systematic inequalities between the sexes in access to care or in treatment decision-making. For example, inappropriate underevaluation of female patients with hematuria has been postulated as a cause of poorer outcomes in female patients with bladder cancer. 27 In this study, we performed a retrospective analysis of the NCDB to assess the interaction of demographic factors with decisions to treat patients who had localized kidney tumors more or less aggressively. To minimize confounding effects of age and comorbidity, we confined our analysis to a young and healthy subset of patients. We categorized patients as having received guideline-based treatment, overtreatment, or undertreatment based on NCCN guidelines.
Female sex was associated with significantly lower odds of undertreatment and higher odds of overtreatment. This observation is interesting in light of prior literature suggesting underreferral and undertreatment of women presenting with hematuria, as well as undertreatment of female patients with other malignant tumors, including head and neck cancer. 27,28 One might question whether there are underlying clinician-driven or patient-driven reasons for this disparity. For example, there may be a tendency of clinicians to perceive female patients as having greater potential longevity and therefore warranting more aggressive cancer treatment. Alternately or concurrently, there may be a systematic preference among female patients for more aggressive treatment.
With regard to race/ethnicity, Black patients had higher odds than White patients of receiving non-guideline-based treatment, consistent with previously reported racial/ethnic discrepancies in surgical management of kidney cancer. 22,23 Interestingly, Black race was associated with higher odds of both undertreatment and overtreatment, and this association was significant across all subgroups, with the exception of the group with the smallest cT1a tumors, in which Black patients had lower odds of overtreatment. Systematic undertreatment of Black patients has been described in other malignant tumors, including prostate cancer. 24 However, our observation that Black patients also had higher odds of overtreatment for kidney cancer is novel and highlights the complexity of the individual patient-level decision-making processes that underlie systemic inequities. For example, overtreatment may result from limited physical resources (eg, lack of access to a surgical robot, rendering partial nephrectomy less appealing), limited human resources (eg, lack of a surgeon who is familiar with partial nephrectomy), or a higher level of real or perceived patient comorbidity resulting in a preference for radical nephrectomy over partial nephrectomy. The higher overtreatment odds appeared to be associated primarily with the large (ie, 2-4 cm) cT1a subgroup, suggesting increased use of radical nephrectomy in small but clinically significant masses that might otherwise be appropriate for partial nephrectomy.
We noted a complex association between insurance status and receipt of non-guideline-based treatment. Uninsured status was associated with lower odds of overtreatment of small cT1a tumors and higher odds of undertreatment compared with private insurance, in a general trend toward less aggressive treatment. As mentioned above, this phenomenon may be associated, at least in part, with decision-making in resource-limited settings; for example, a patient with a large cT1a tumor in a hospital that lacks a surgical robot may be more likely to undergo a radical rather than a partial nephrectomy. One notable observation was that there were higher odds of overtreatment associated with Medicare insurance, associated in part with higher rates of overtreatment for large cT1a masses.
It should be noted that patients with Medicare in this cohort represented a select group. Our inclusion criteria enriched for patients with Medicare under age 65 years, and eligibility for Medicare in this population requires a qualifying disability (eg, end-stage kidney disease, amyotrophic lateral sclerosis, or Ն24 months of Social Security disability payments). Our eligibility criteria would be expected to exclude patients with major comorbidities; however, it is possible that the overtreatment outcome seen here was partially associated with decisions to perform radical nephrectomy in patients with end-stage kidney disease who had moderate size tumors in nonfunctional kidneys.

Limitations
Several limitations of our study should be noted. It should be acknowledged that clinical guidelines may not represent the ideal treatment for all patients, and the decision to treat a patient more or less aggressively than recommended by guidelines may be entirely appropriate. It should also be acknowledged that we retrospectively applied a contemporary guideline to patients treated in the past. However, the NCCN guideline used here broadly mirrors the recommendations of the 2009 AUA guideline statement, which was in effect for the entirety of the study period. 11 While there may be concerns about the application of our overtreatment and undertreatment classification to individual patients, we believe the classification represents a reasonable metric for determining the aggressiveness of cancer treatment decisions across large groups of patients.

JAMA Network Open | Urology
Demographic Factors Associated With Non-Guideline-Based Treatment of Kidney Cancer in the United States With regard to other limitations, the NCDB relies on registry data submitted by CoC-accredited cancer centers, and its underlying accuracy cannot be ascertained. Systematic inaccuracies may be present. In particular, race/ethnicity as reported in large databases does not correlate perfectly with patients' self-reports, and such discrepancies could be a source of systematic error in our findings. 29 The NCDB is not a population-based database, and while it captures more than 70% of new cancer diagnoses in the United States, it does represent a nonrandom subset of the overall population. Our study population may not represent a comprehensive sample of patients presenting for evaluation of kidney masses, given that individuals with smaller masses who undergo surveillance may not be captured by the database owing to nonreferral. We made efforts to minimize the effects of comorbidity and clinical factors by excluding patients who were older, had comorbidities, or had advanced-stage disease from our analysis; however, unaccounted confounding factors may remain.
Additionally, although we found that several demographic and institutional factors were associated with receipt of non-guideline-based treatment for kidney cancer, we are unable to determine the reasons underlying these disparities or to assess the associations of these disparities with clinical outcomes. Our study carries the limitations of a retrospective study, including the inability to attribute causality to the associations we found.

Conclusions
To our knowledge, this study represents the first attempt to assess the associations between demographic factors and receipt of non-guideline-based treatment among patients with kidney cancer. We found that female patients with kidney cancer had higher odds of receiving more aggressive treatment than men, which was associated with increased rates of overtreatment for small kidney masses and potentially increased risk for unjustified complications. Black race and Hispanic ethnicity were associated with higher odds of undertreatment and overtreatment, highlighting the bidirectional nature of inequities in treatment. Patients without insurance had markedly lower odds of overtreatment for very small kidney masses, potentially representing a rare example of a salutary association of reduced access to care. Clinicians should bear these disparities in mind when counseling individual patients, and health policy makers should take the existence of these disparities into account.