Association of Clear Cell Likelihood Score on MRI and Growth Kinetics of Small Solid Renal Masses on Active Surveillance

Abstract

BACKGROUND.

The lack of validated imaging markers to characterize biologic aggressiveness of small renal masses (SRMs)—defined as those categorized as cT1a and 4 cm and smaller—hinders medical decision-making among available initial management strategies.

OBJECTIVE.

The purpose of this article was to explore the association of the clear cell likelihood score (ccLS) on MRI with growth rates and progression of SRMs.

METHODS.

This retrospective study included consecutive SRMs assigned a ccLS on clinical MRI examinations performed between June 2016 and November 2019 at an academic tertiary-care medical center or its affiliated safety net hospital system. The ccLS reports the likelihood that the SRM represents clear cell renal cell carcinoma (ccRCC) from 1 (very unlikely) to 5 (very likely). The ccLS was extracted from clinical reports. Tumor size measurements were extracted from available prior and follow-up cross-sectional imaging examinations, through June 2020. Serial tumor size measurements were fit to linear and exponential growth curves. Estimated growth rates were grouped by the assigned ccLS. Tumor progression was defined by development of large size (> 4 cm in at least two consecutive measurements) and/or rapid growth (doubling of volume within 1 year). Differences among ccLS groups were evaluated using Kruskal-Wallis tests. Correlations between ccLS and growth rate were evaluated by Spearman correlation (ρ).

RESULTS.

Growth rates of 386 SRMs (100 ccLS 1–2, 75 ccLS 3, and 211 ccLS 4–5) from 339 patients (median age, 65 years; 198 men, 141 women) were analyzed. Median follow-up was 1.2 years. The ccLS was correlated with growth rates by size (ρ = 0.19; p < .001; ccLS 4–5, 9%/year; ccLS 1–2, 5%/year; p < .001) and by volume (ρ = 0.14; p = .006; ccLS 4–5, 29%/year; ccLS 1–2, 16%/year; p < .001). Disease progression (observed in 49 SRMs) was not significantly associated with ccLS group (p = .61). Two patients (0.6%) developed metastases during active surveillance: one ccLS 1 was a type 2 papillary renal cell carcinoma and one ccLS 4 was ccRCC.

CONCLUSION.

Growth is associated with ccLS in SRMs, with higher ccLS correlating with faster growth.

CLINICAL IMPACT.

SRMs with lower ccLS may be considered for active surveillance, whereas SRMs with higher ccLS may warrant earlier intervention. The noninvasive ccLS derived from MRI correlates with growth rate of SRMs and may help guide personalized management.

Faster tumor growth has been associated with increased risk of metastases in small renal masses (SRMs)—defined as T category of cT1a or 4 cm and smaller—undergoing active surveillance [1]. Clear cell renal cell carcinoma (ccRCC), the most common histologic subtype of renal cell carcinoma (RCC), grows faster and carries a worse cancer-specific mortality than other histologic subtypes of RCC [2]. Currently, percutaneous renal mass biopsy (RMB) is the only method to diagnose ccRCC preoperatively. However, in a multicenter prospective trial of centers that are experienced in RMB and that support its clinical adoption, RMB was performed in only 159 of 239 (66.6%) patients with an SRM [2]. Further, when RMB is performed, its low NPV of 63% limits clinical utility [3].

Recent work on MRI of renal masses has focused on noninvasively predicting tumor histology [4]. The clear cell likelihood score (ccLS) on MRI was developed to grade the likelihood that the mass will represent ccRCC on final pathologic analysis. The ccLS has shown good diagnostic performance and moderate-to-good interreader variability (mean κ, 0.53) [5–7]. Given the faster growth and aggressive nature of ccRCCs [2], the current study was performed to assess the association of the ccLS on MRI with growth rate and progression of SRMs.

Methods

Patient Selection

This retrospective study was approved by the institutional review board. The requirement for informed consent was waived. The study was compliant with HIPAA. Patients were identified from an academic tertiary-care medical center (practice 1) and its affiliated safety net hospital system for Dallas County, TX (practice 2). Both practices incorporated the ccLS into the clinical reports of MRI examinations of renal masses beginning in June 2016. Electronic medical records were reviewed for patients who underwent MRI to evaluate a renal mass between June 2016 and November 2019 and for whom the clinical imaging report assigned a ccLS for the renal mass, identifying 942 consecutive masses in 841 patients. Renal masses were then excluded for the following reasons: MRI performed without IV contrast material (n = 7); histology of the renal mass known at the time of MRI because of prior biopsy (n = 27); the mass was ineligible for ccLS assignment according to the ccLS algorithm (e.g., presence of macroscopic fat, predominantly cystic mass) (n = 49); the mass had clinical T category greater than cT1a (i.e., size > 4 cm or extrarenal extension) on MRI (n = 282); known genetic predisposition to renal masses (e.g., von Hippel–Lindau syndrome or Birt-Hogg-Dubé syndrome) (n = 11); and no eligible prior or subsequent imaging examinations of the mass performed through June 2020 (n = 164). Prior or subsequent imaging was deemed eligible in the absence of any of the following characteristics: imaging was an unenhanced CT or ultrasound examination given these modalities’ limitations in evaluating tumor size [8]; imaging was performed after an intervention (partial or radical nephrectomy, ablation, systemic therapy, or radiation therapy); imaging showed marked change in size secondary to tumor hemorrhage (e.g., after biopsy); and imaging was performed within 90 days after the initial imaging of the mass. These exclusions resulted in 348 patients with a total of 402 renal masses with T category T1a that were clinically assigned a ccLS. A total of 16 SRMs in nine patients with known existing renal metastatic disease on initial MRI were removed from the primary analysis, leaving a final study sample of 386 renal masses in 339 patients (median age, 65 years; interquartile range [IQR], 55–74 years; 198 men, 141 women) (Fig. 1 and Table 1). A total of 121 patients with 122 masses [6] and 26 patients with 26 masses [7] were included in prior studies that evaluated the diagnostic performance of ccLS according to a pathologic reference standard.

Fig. 1—

Flow diagram of cohort of patients with small renal masses that were clinically assigned clear cell likelihood score (ccLS) on MRI and that were observed over multiple imaging time points.

TABLE 1:

Patient Demographics and Lesion Characteristics of the Cohort Included in the Primary Analysis

Variable	Practice 1	Practice 2	Total
Total
Renal masses	247 (64.0)	139 (36.0)	386 (100.0)
Patients	211 (62.2)	128 (37.8)	339 (100.0)
Renal masses
ccLS
1	27 (10.9)	25 (18.0)	52 (13.5)
2	32 (13.0)	16 (11.5)	48 (12.4)
3	54 (21.9)	21 (15.1)	75 (19.4)
4	85 (34.4)	44 (31.7)	129 (33.4)
5	49 (19.8)	33 (23.7)	82 (21.2)
Baseline volume (cm3), median (IQR)
Overall	1.8 (0.8–5.3)	2.4 (0.8–7.1)	2.0 (0.9–5.7)
ccLS 1–2	1.8 (0.7–5.5)	3.0 (0.6–9.2)	2.3 (0.7–6.5)
ccLS 3	1.7 (0.7–5.1)	2.0 (0.6–7.5)	2.0 (0.6–5.4)
ccLS 4–5	1.8 (1.0–5.5)	2.2 (1.0–5.8)	2.0 (1.0–5.6)
Baseline maximum diameter (cm), median (IQR)
Overall	1.7 (1.3–2.4)	1.8 (1.4–2.6)	1.7 (1.3–2.5)
ccLS 1–2	1.6 (1.3–2.7)	2.0 (1.4–2.8)	1.8 (1.3–2.7)
ccLS 3	1.7 (1.2–2.3)	1.8 (1.1–2.7)	1.7 (1.2–2.4)
ccLS 4–5	1.7 (1.3–2.4)	1.8 (1.4–2.4)	1.7 (1.3–2.4)
Laterality
Left	113 (45.7)	73 (52.5)	186 (48.2)
Right	134 (54.3)	66 (47.5)	200 (51.8)
Patients
Sex
Male	130 (61.6)	68 (53.1)	198 (58.4)
Female	81 (38.4)	60 (46.9)	141 (41.6)
Outcome
Continued AS	136 (55.1)	59 (42.4)	195 (50.5)
Total or radical Nx	3 (1.2)	2 (1.4)	5 (1.3)
Partial Nx	55 (22.3)	33 (23.7)	88 (22.8)
Ablation	19 (7.7)	15 (10.8)	34 (8.8)
ST and/or SBRT	2 (0.8)	2 (1.4)	4 (1.0)
Lost to follow-up	21 (8.5)	18 (12.9)	39 (10.1)
Death before interventiona	11 masses (4.5), 8 patients (3.8)	10 masses (7.2), 9 patients (7.0)	21 masses (5.4), 17 patients (5.0)
Deaths
Cancer related	0 (0.0)	0 (0.0)	0 (0.0)
Not cancer related	5 (2.4)	4 (3.1)	9 (2.7)
Unknown cause	4 (1.9)	5 (3.9)	9 (2.7)

Note—Unless otherwise indicated, values represent number (percentage). Practice 1 was an academic tertiary-care medical center and practice 2 was its affiliated safety net hospital system. ccLS = clear cell likelihood score, IQR = interquartile range, AS = active surveillance, Nx = nephrectomy, ST = systemic therapy, SBRT = stereotactic body radiation therapy.

^aDeaths before intervention are totaled as the number (percentage) of representative lesions and number (percentage) of patients.

For included patients, electronic medical records were reviewed to obtain patient demographics (including age at the time of initial ccLS assignment), tumor characteristics, disease interventions, and development of metastases. In patients who underwent more than one MRI examination showing the SRM, the initial examination was used for recording ccLS and defining growth rate. All eligible prior and subsequent imaging examinations relative to the selected MRI were used for determining growth rate.

Image Acquisition and Analysis

The MRI protocol performed at the two practices has been previously described [5] and represents a standard clinical MRI protocol performed without and with IV contrast material. The MRI interpretation was performed independently by one of 15 fellowship-trained abdominal radiologists assigned to the clinical service who had been previously trained in assigning ccLS. The interpreting radiologists had from 2.6 to 26.8 years of postresidency clinical experience at the time of MRI interpretation. The diagnostic algorithm for assigning ccLS has been previously reported [5, 9], and the schematic of the ccLS algorithm is shown in supplemental Figure S1, which can be viewed in the AJR electronic supplement to this article available at www.ajronline.org. The ccLS is a Likert scale ranging from 1 to 5, from least likely to most likely to represent ccRCC [10]. Although the diagnostic algorithm has not changed since its inception, additional rules guiding the ccLS assignment have been implemented over time [4, 9].

Tumor Measurements

All reported tumor dimensions were extracted from the imaging reports of all eligible imaging examinations for determining growth rate. The standardized reporting template for renal MRI at the two practices includes three tumor dimensions. Although no specific instructions were provided to interpreting radiologists, it is standard practice to measure tumors on sequences in which the tumor is best visualized. Radiologists do not receive prompts to document three tumor dimensions using the standardized template, and the interpreting radiologist may edit the standardized reporting templates for both renal MRI examinations and follow-up CT examinations (whether performed specifically for active surveillance or for other indications, providing opportunistic evaluation of the mass). Therefore, some imaging examinations do not include all three tumor dimensions.

Tumor size for each time-eligible imaging examination was recorded as the largest cross-sectional diameter documented in the corresponding imaging report. Tumor volume was calculated using the ellipsoid formula as

$$\frac{1}{6}\pi abc,$$ (1)

where a, b, and c represent three cross-sectional diameters. When only two diameters were available, volume was calculated as

$$\frac{1}{6}\pi ab\cdot \text{min }\left(a,b\right).$$ (2)

When only one diameter was available, volume was calculated as

$$\frac{1}{6}\pi a^3.$$ (3)

To test the reproducibility of tumor measurements, two radiology residents (R.G.R. and R.C.S., 2nd- and 4th-year residents, respectively) remeasured on the initial MRI the maximum cross-sectional diameters for a subset of 80 SRMs from the analyzed cohort. The clinically assigned ccLSs were not included in the information available to the residents.

Growth Rate Estimation

From the recorded tumor sizes and volumes, fixed absolute and percent growth models were applied respectively to each lesion. To calculate fixed absolute growth, linear models on the tumor diameter and volume were fit using the following equations:

$$D_t=D_0+ATD\cdot T$$ (4)

and

$$V_t=V_0+ATV\cdot T,$$ (5)

where D_t and V_t represent tumor diameter and volume at time t; D₀ and V₀ represent intercepts for tumor diameter and volume; and T is the duration of follow-up in years. ATD and ATV represent annual absolute tumor growth measured in centimeters and cubic centimeters, respectively.

For fixed percent growth, linear models on the log transformed tumor diameter and volume were fit using the following equations:

$$\text{log}\left(D_t\right)=\text{log}\left(D_0\right)+T\cdot \text{log}\left(1+PTD\right)$$ (6)

and

$$\text{log}\left(V_t\right)=\text{log}\left(V_0\right)+T\cdot \text{log}\left(1+PTV\right),$$ (7)

where PTD and PTV represent annual percent tumor growth in diameter and volume, respectively.

Masses were deemed to exhibit tumor progression because of either development of large size (diameter > 4 cm in at least two consecutive measurements) during active surveillance and/or rapid growth (doubling of volume within 1 year). These criteria for local progression represent currently recommended triggers for treatment consideration [2, 11].

Statistical Analysis

Kruskal-Wallis tests were used to test for differences among ccLS groups in initial tumor size and volume. Time of follow-up between lesions measuring less than 2 cm and lesions measuring 2–4 cm was compared using Wilcoxon rank-sum test.

Lesions were grouped into three categories (ccLS 1–2, unlikely ccRCC; ccLS 3, intermediate likelihood of ccRCC; and ccLS 4–5, likely ccRCC), consistent with previously reported clinically useful thresholds for stratifying ccLS [7]. Estimated median annual growth rate and IQR for each ccLS group were reported. Correlations of growth rates with ccLS categories were calculated using Spearman rank correlation (ρ). Kruskal-Wallis tests were used to test the difference among ccLS groups in annual absolute tumor growth in diameter and volume and annual percent tumor growth in diameter and volume. Dwass-Steel-Critchlow-Fligner multiple comparisons post hoc procedure was applied if overall Kruskal-Wallis tests were significant. Distributions of lesion inclusions and exclusions and lesion progression by ccLS groups were compared using chi-square tests.

Linear mixed models (LMMs) were performed, applying all available follow-up information with the appropriate transformation. Lesion-level random intercepts were applied. The growth rates by log transformed diameter and volume for each ccLS group were estimated via a fixed interaction term between ccLS and follow-up time T. Estimated annual growth rate and 95% CIs for each ccLS group were reported. Ad-hoc pairwise comparisons in growth rates among groups were performed with Sidak correction on p values.

For the subset of masses for which the maximum cross-sectional diameters were remeasured, these measurements were compared with the original measurements from the clinical MRI reports. Intraclass correlation coefficient and Bland-Altman analyses were performed to assess the interreader measurement error for lesion size.

Times to reach diameter thresholds of greater than 3 cm (corresponding with the size at which an SRM is unamenable to radiofrequency ablation [12]) and greater than 4 cm (corresponding with a T category of T1b) from the initial median size among all lesions were estimated using the predictInterval function of the merTool R package (version 0.5.2, J. E. Knowles and C. Frederick) [13] based on the fit growth rate LMM. An online calculator was constructed based on the merTool R package and predictInterval function results to provide a time estimate to reach these thresholds for individual tumors [14]. Growth rates were estimated by absolute tumor diameter for the renal masses that were removed from the primary analysis because of known metastatic disease on the initial MRI.

A p value of less than .05 was considered statistically significant. All analyses were performed in SAS 9.4 (SAS Institute).

Results

Growth rates of 386 SRMs (100 ccLS 1–2, 75 ccLS 3, and 211 ccLS 4–5) from 339 patients were analyzed. A total of 1158 imaging time points were included in the primary analysis (median, 3.0 time points per SRM; IQR, 2.0–3.8 time points). Median SRM diameter at time of initial imaging was 1.7 cm (IQR, 1.3–2.5 cm). Mean (± SD) and median time of follow-up between the initial and most recent eligible time points were 1.7 ± 1.5 years and 1.2 years (IQR, 0.6–2.2 years), respectively. Patient demographics and lesion characteristics are shown in Table 1. These characteristics for SRMs without known renal metastatic disease on initial MRI that were excluded because of fewer than two eligible time points are shown in Table S1, which can be viewed in the AJR electronic supplement to this article available at www.ajronline.org. The distribution of assigned ccLS groups was not significantly different (p = .83) between this excluded cohort and the studied cohort. Initial tumor size (p = .55) and initial volume (p = .46) showed no significant difference by ccLS group. Larger tumors had shorter follow-up periods than smaller tumors: 0.9 years (IQR, 0.5–1.6 years) for masses measuring 2–4 cm compared with 1.5 years (IQR, 0.8–2.7 years) for masses measuring less than 2 cm (p < .001).

Eighteen (5.3%) patients died during the study period. Seventeen patients, representing 21 lesions, died during the study period before any treatment or intervention. None of these deaths resulted from renal cancer: eight were from other causes and nine from unknown causes, without a diagnosis of metastatic disease. One patient (0.3%) with significant comorbidities died in the perioperative period after partial nephrectomy of a lesion assigned ccLS 5 for which ccRCC was diagnosed at pathology.

In the 80 patients for whom the maximum cross-sectional diameter was remeasured, the intraclass correlation coefficient with the original clinical measurements was 0.97 (95% CI, 0.95–0.98). Corresponding Bland-Altman plots of the maximum cross-sectional diameter from each of the two reviewers and original dataset are shown in Figure S2, which can be viewed in the AJR electronic supplement to this article available at www.ajronline.org. The overall median absolute growth rates by diameter and volume were 0.08 cm/year and 0.32 cm³/year, respectively (IQR, 0.00–0.30 cm/year and 0.00–1.57 cm³/year, respectively) and median percent growth rates by diameter and volume were 5% per year and 19% per year, respectively (IQR, 0–18% per year and 0–62% per year, respectively). The median growth rates by absolute and percent tumor growth by diameter and volume for the ccLS groups are shown in Table 2. Higher growth rates were associated with higher ccLS groups (Table 2). The difference in growth rates was statistically significant (p < .05) among the groups for all measurements of tumor growth (Table 3). The ccLS was correlated with growth rates by size (ρ = 0.19; p < .001; ccLS 4–5, 9%/year; ccLS 1–2, 5%/year; p < .001) and by volume (ρ = 0.14; p = .006; ccLS 4–5, 29%/year; ccLS 1–2, 16%/year; p < .001). Pairwise comparison showed that lesions with ccLS 4–5 grew significantly faster (p < .05) than lesions with ccLS 1–2 by all measurements of growth rate and significantly faster than lesions with ccLS 3 by absolute tumor diameter and volume growth and percent tumor diameter growth. Figure 2 displays individual lesion growth curves in diameter and volume by ccLS group with the corresponding median percent tumor diameter and volume growth fits. The full growth curves, including all analyzed time points, are shown in Figure S3, which can be viewed in the AJR electronic supplement to this article available at www.ajronline.org. Figure 3 displays individual absolute tumor diameter growth rates sorted by ccLS group. Examples of initial MRI sequences of a ccLS 1 and a ccLS 5 renal mass with corresponding follow-up imaging are shown in Figure 4.

TABLE 2:

Yearly Tumor Growth Rates by Absolute Growth, Percent Growth, and Linear Mixed Model Estimates as Stratified by Clear Cell Likelihood Score (ccLS) Ranges

Variable	ccLS 1–2	ccLS 3	ccLS 4–5
Absolute tumor growth, median (IQR)
Diameter (cm)	0.00 (−0.02 to 0.24)	0.00 (−0.03 to 0.21)	0.15 (0.00–0.35)
Volume (cm3)	0.12 (−0.02 to 1.05)	0.17 (−0.07 to 0.85)	0.59 (0.02–1.92)
Percent tumor growth (%), median (IQR)
Diameter	0 (−2 to 18)	0 (−3 to 12)	9 (0–19)
Volume	12 (−3 to 44)	13 (−5 to 58)	25 (1–75)
Linear mixed model growth rate (%), mean (95% CI)
Diameter	5 (3–7)	4 (2–6)	9 (8–10)
Volume	16 (12–22)	15 (10–21)	29 (25–33)

Note—IQR = interquartile range.

TABLE 3:

Statistical Significance of Correlations and Differences in Absolute and Percent Growth Rates Among Clear Cell Likelihood Score (ccLS) Groups

Tumor Growth	Spearman Correlation Coefficient (95% CI)	Spearman p	Kruskal-Wallis p	Comparison of Groups, Pairwise p
				ccLS 1–2 vsccLS3	ccLS 1–2 vs ccLS 4–5	ccLS 3 vs ccLS 4–5
Absolute
Diameter	0.19 (0.09–0.28)	< .001	< .001	.77	.007	< .001
Volume	0.16 (0.06–0.25)	.002	.003	.90	.03	.007
Percent
Diameter	0.19 (0.09–0.28)	< .001	< .001	.87	.007	.002
Volume	0.14 (0.04–0.24)	.006	.02	> .99	.04	.08

Fig. 2—

Growth curves stratified by clear cell likelihood score (ccLS). All x-axes are truncated at 5 years of follow-up to improve depiction of detail during this interval.

A and B, Growth curves for maximum tumor diameter (A) and volume (B) of individual masses (gray lines) and median growth curves (red lines) by percent tumor diameter (A) and volume (B) are shown for each ccLS group.

Fig. 3—

Estimated absolute tumor diameter growth rates of individual lesions in descending order as clustered by clear cell likelihood score (ccLS) group. Lesions representing existing renal metastatic disease on initial ccLS assignment (Existing mets) and those that developed presumed renal metastatic disease during active surveillance (New mets) are highlighted in orange and black, respectively.

Fig. 4—

Example imaging sequences of renal masses.

A–E, Renal mass (arrow) with clear cell likelihood score (ccLS) of 1 in 52-year-old man is shown on coronal T2-weighted SSFSE (A), coronal T1-weighted fat-saturated spoiled gradient-echo acquired during corticomedullary (CM) phase (B), and axial T1-weighted gradient-echo opposed-phase (OP) (C) and in-phase (IP) (D) images from initial MRI. Follow-up coronal T2-weighted image (E) was obtained at 2.3 years. Renal mass with ccLS 1 is hypointense compared with renal cortex on T2-weighted image and CM enhancement is mild compared with unenhanced T1-weighted image (not shown), mass has no microscopic fat on OP imaging (i.e., no drop in signal intensity) compared with IP imaging, and mass shows no significant change in size on follow-up (0.001%/year volume growth).

F–J, Renal mass (arrow) with ccLS of 5 in 64-year-old man is shown on coronal T2-weighted SSFSE (F), coronal T1-weighted fat-saturated spoiled gradient-echo acquired during CM phase (G), and axial T1-weighted gradient-echo OP (H) and IP (I) images from initial MRI. Follow-up coronal T2-weighted image (J) was obtained at 2.8 years. Renal mass with ccLS of 5 is heterogeneously hyperintense on T2-weighted image with intense CM enhancement (red outline, G), has microscopic fat on OP compared with IP image (i.e., drop in signal intensity), and shows significant interval growth (50.4%/year volume growth).

The LMMs of growth rate by diameter and volume showed rapid growth of the ccLS 4–5 SRMs (9% and 29%, respectively) (Table 2). The growth rates of ccLS 4–5 SRMs were statistically significantly faster (p < .05) than ccLS 1–2 and ccLS 3 SRMs by diameter and volume (Table 4).

TABLE 4:

Statistical Significance of Differences in Linear Mixed Model Growth Rates Among Clear Cell Likelihood Score (ccLS) Groups

Comparison of Groups	Growth Rate in Size		Growth Rate in Volume
	Raw p	Sidak p	Raw p	Sidak p
ccLS 4–5 vs ccLS 3	< .001	< .001	< .001	< .001
ccLS 4–5 vs ccLS 1–2	< .001	< .001	< .001	< .001
ccLS 3 vs ccLS 1–2	.39	.77	.72	.98

Forty-nine SRMs had progressed according to large size and/or rapid growth by the last follow-up. Of these, 44 (89.8%) progressed by rapid growth (volume doubling), 4 (8.2%) progressed based on an increase in size to greater than 4 cm, and 1 (2.0%) progressed by both criteria. Disease progression was not significantly different (p = .61) by ccLS group (progression in 13/100 [13.0%] ccLS 1–2, 7/75 [9.3%] ccLS 3, and 29/211 [13.7%] ccLS 4–5).

Table 5 shows the times to reach larger than 3 cm and larger than 4 cm for each ccLS level, according to the median overall initial diameter of 1.7 cm and the fit LMM diameter growth rates.

TABLE 5:

Estimated Times to Reach Tumor Diameter Thresholds From an Initial Size of 1.7 cm by Linear Mixed Model Fits for Each Clear Cell Likelihood Score (ccLS)

ccLS	Time (y) to Growth
	> 3 cm	> 4 cm
1	11.6 (7.3–21.0)	17.5 (11.9–30.9)
2	13.7 (8.2–30.3)	20.7 (13.2 to > 40)
3	14.7 (9.2–27.6)	22.3 (15.0 to > 40)
4	6.9 (4.9–9.6)	10.4 (8.1–13.4)
5	6.3 (4.3–9.1)	9.4 (7.2–12.9)

Note—Values are reported in years with 95% CIs in parentheses. Median baseline tumor size in entire cohort was 1.7 cm.

Two patients (0.6%) developed metastases presumably related to the ccLS-assessed lesions during active surveillance. One patient with a 1.8-cm lesion assigned ccLS 1 developed retroperitoneal lymphadenopathy during active surveillance. This lesion was a type 2 papillary RCC on pathology; subsequent genetic testing was consistent with hereditary leiomyomatosis and RCC. The second patient, with a history of partial nephrectomy for a 3.6-cm ccRCC, developed an ipsilateral 1.2-cm SRM 3.7 years later that was ccLS 4, representing the lesion included in this study. This second lesion grew to 2.6 cm during 2.5 years of active surveillance. After confirmation of ccRCC after partial nephrectomy for this second lesion, the patient developed pancreatic lesions that were consistent by imaging with metastases. Although endoscopic biopsy of these lesions failed to confirm this diagnosis, they were classified as metastases for the purpose of this analysis [15]. Beyond these two SRMs with new metastatic disease during active surveillance, no other tumor showed extrarenal extent of disease (e.g., renal vein extension, T3a disease) on eligible follow-up imaging. A total of 18 masses exhibited renal metastatic disease that was new (the two patients in the primary analysis who developed metastases) or existing (the 16 T1a masses that were otherwise eligible but excluded from the primary analysis because of the existing metastases). The median absolute tumor diameter growth of these 18 lesions was 0.54 cm/year (IQR, 0.44–0.83 cm/year). The individual growth rates are highlighted in Figure 3.

One patient with papillary RCC developed retroperitoneal metastases 12 years after radical nephrectomy during active surveillance of new incidental contralateral SRMs. These contralateral masses that were assigned ccLS 3 were later characterized as oncocytoma on RMB.

Discussion

This study is a retrospective analysis of growth rates of SRMs stratified by clinically assigned ccLS on MRI examinations performed at either an academic tertiary-care medical center or its affiliated regional hospital network. To our knowledge, this study is the first to report a correlation between ccLS and growth in SRMs. This report has several main findings that support the use of ccLS in clinical management of SRMs.

First, SRMs assigned ccLS 4–5 grew at a faster rate than those assigned ccLS 1–2 or ccLS 3. This is consistent with the faster growth rates reported for ccRCCs [2] and high rate (84%) of ccRCC in ccLS 4–5 SRMs [6, 7]. These faster growth rates were significant for both absolute and percent tumor diameter and volume growth. The differences in the growth rates among the ccLS groups were more pronounced in the LMMs, which applied all available follow-up information. The overall growth rates of the SRMs were similar to the rates observed in studies with similar short-term follow-up periods of 0.09 cm/year (median) [16] and 0.12 cm/year (mean) [17]. However, the faster growth rate observed in the SRMs assigned ccLS 4–5 of 0.15 cm/year is similar to the rate previously reported for renal masses over longer-term follow-up periods [2, 18].

In our cohort, SRMs assigned ccLS 1–2 or ccLS 3 exhibited negligible median growth rates at 0.00 cm/year. A prior study reported significantly slower growth for masses showing homogeneity on T2-weighted images among 47 renal masses undergoing active surveillance [19]. This is consistent with our results in that masses showing homogeneity on T2-weighted images typically receive lower ccLS. These findings support a management algorithm for SRMs in which eligible patients with SRMs assigned ccLS 1–2 are generally recommended for active surveillance, those with SRMs rated ccLS 4–5 are recommended for disease intervention if the mass measures greater than 2 cm, and those with SRMs assigned ccLS 3 are further evaluated with RMB. Such strategy could help obviate pathologic confirmation by RMB in many patients before recommending active surveillance or other disease intervention.

Analysis of the follow-up data supports the use of exponential (percentile) growth models. This result is in agreement with recent work that analyzed log-transformed lesion growth rates on active surveillance by RMB histology [2]. This finding may reconcile the differences in estimated growth rates in prior studies with differing follow-up periods. Because of the flatter slope in the early portion of an exponential curve, studies with shorter follow-up periods typically estimate lower growth rates compared with studies with longer follow-up periods that estimate growth inclusive of the steeper portions of the exponential curve. Although the estimated absolute tumor diameter growth rate of ccLS 4–5 lesions was similar to those estimated across all SRMs in recent studies, the shorter follow-up period of our study supports the conclusion of the more aggressive faster-growing nature of ccLS 4–5 SRMs. This again favors our recommendation of disease intervention in larger ccLS 4–5 T1a masses, in consideration of the patients’ clinical context and preferences.

Nearly all SRMs associated with new or known existing metastatic disease displayed rapid growth. These results highlight the importance of growth rates with respect to metastatic potential, whereby high growth rates are suggestive of greater metastatic potential, and lack of growth or local spread nearly rules out metastases. This remains consistent with an initial management strategy of active surveillance for SRMs assigned ccLS 1–2, which typically grow slowly compared with the more aggressive SRMs assigned ccLS 4–5.

Although numerous important associations were observed between ccLS and growth, management of renal masses is complex and multiple factors are to be considered, including patient preferences, anxiety, life expectancy, and medical comorbidities. Moreover, as is the case for growth rates among tumors with a given histologic subtype of RCC [2], growth rates were variable among tumors in the same ccLS group, as evidenced by estimated growth rates’ wide IQRs and 95% CIs. Thus, the results support the use of the ccLS as an informational tool to assist with clinical decision-making. To that end, we developed an online calculator that enables the patient and treating physician to consider the likelihood of growth within the context of the patient’s life expectancy and comorbidities [14].

This study has some limitations. First, the study was conducted retrospectively using clinical data. Thus, ccLS assignment of a given SRM was based on a single radiologist interpreting an MRI examination as part of the clinical workflow. Interobserver agreement for ccLS was not assessed. Nonetheless, prior work has shown moderate-to-good interreader agreement (mean κ, 0.53) [5]. Although current clinical practice at the included medical practices generally follows consensus guidelines [20] for the timing of imaging in active surveillance, patients were not prospectively placed in active surveillance cohorts with strict imaging time points or strict criteria for recommending disease intervention. As such, the ccLS may be obtained at a middle time point in the continuum of imaging performed during SRM follow-up, and follow-up time points are variable because of a multitude of factors. Second, our median follow-up period (1.2 years) is relatively short, which likely limited analysis of differences in progression stratified by ccLS. Moreover, the clinically assigned ccLS itself may have been a confounding factor that influenced the timeline to disease intervention. At the two practices, patients with SRMs assigned ccLS 4–5 are likely to be recommended for disease intervention earlier, particularly for larger T1a masses. Such earlier intervention of ccLS 4–5 SRMs is expected to underestimate linear growth rates relative to other SRMs that remain on active surveillance given the exponential nature of tumor growth. Third, given the retrospective nature of the analysis, this study is unable to calculate the percentage of all incidental SRMs that were evaluated on MRI at the two practices that met study inclusion criteria. Lastly, variability in imaging time points and active surveillance periods may have contributed to the wide variability in growth rates.

In conclusion, the ccLS of SRMs is significantly associated with growth; SRMs assigned ccLS 4–5 grow more rapidly than SRMs assigned ccLS 1–2 and ccLS 3. The information provided by ccLS may be beneficial in counseling patients and planning personalized management strategies.

HIGHLIGHTS.

Key Finding

• Among 386 SRMs in 339 patients assigned a ccLS on MRI, those with ccLS 4–5 grew significantly faster (9% diameter, 29% volume yearly) than those with ccLS 1–2 (5% diameter; 16% volume; both p < .001) or ccLS 3 (4% diameter; 15% volume; both p < .001).

Importance

• The standardized noninvasive ccLS, derived from MRI, correlates with growth rate of SRMs and may help guide personalized management.