Natural history of untreated kidney cancer

CT-derived radiomics predict the growth rate of renal tumours in von Hippel-Lindau syndrome

AIM

To predict renal tumour growth patterns in von Hippel-Lindau syndrome by utilising radiomic features to assist in developing personalised surveillance plans leading to better patient outcomes.

MATERIALS AND METHODS

The study evaluated 78 renal tumours in 55 patients with histopathologically-confirmed clear cell renal cell carcinomas (ccRCCs), which were segmented and radiomics were extracted. Volumetric doubling time (VDT) classified the tumours into fast-growing (VDT <365 days) or slow-growing (VDT ≥365 days). Volumetric and diametric growth analyses were compared between the groups. Multiple logistic regression and random forest classifiers were used to select the best features and models based on their correlation and predictability of VDT.

RESULTS

Fifty-five patients (mean age 42.2 ± 12.2 years, 27 men) with a mean time difference of 3.8 ± 2 years between the baseline and preoperative scans were studied. Twenty-five tumours were fast-growing (low VDT, i.e., <365 days), and 53 tumours were slow-growing (high VDT, i.e., ≥365 days). The median volumetric and diametric growth rates were 1.71 cm3/year and 0.31 cm/year. The best feature using univariate analysis was wavelet-HLL_glcm_ldmn (area under the receiver operating characteristic [ROC] curve [AUC] of 0.80, p<0.0001), and with the random forest classifier, it was log-sigma-0-5-mm-3D_glszm_ZonePercentage (AUC: 79). The AUC of the ROC curves using multiple logistic regression was 0.74, and with the random forest classifier was 0.73.

CONCLUSION

Radiomic features correlated with VDT and were able to predict the growth pattern of renal tumours in patients with VHL syndrome.

A Predictive Model for Kidney Failure After Nephrectomy for Localized Kidney Cancer: The Kidney Cancer Risk Equation

RATIONALE & OBJECTIVE

Nephrectomy is the mainstay of treatment for individuals with localized kidney cancer. However, surgery can potentially result in the loss of kidney function or in kidney failure requiring dialysis/kidney transplantation. There are currently no clinical tools available to preoperatively identify which patients are at risk of kidney failure over the long term. Our study developed and validated a prediction equation for kidney failure after nephrectomy for localized kidney cancer.

STUDY DESIGN

Population-level cohort study.

SETTING & PARTICIPANTS

Adults (n=1,026) from Manitoba, Canada, with non-metastatic kidney cancer diagnosed between January 1, 2004, and December 31, 2016, who were treated with either a partial or radical nephrectomy and had at least 1 estimated glomerular filtration rate (eGFR) measurement before and after nephrectomy. A validation cohort included individuals in Ontario (n=12,043) with a diagnosis of localized kidney cancer between October 1, 2008, and September 30, 2018, who received a partial or radical nephrectomy and had at least 1 eGFR measurement before and after surgery.

NEW PREDICTORS & ESTABLISHED PREDICTORS

Age, sex, eGFR, urinary albumin-creatinine ratio, history of diabetes mellitus, and nephrectomy type (partial/radical).

OUTCOME

The primary outcome was a composite of dialysis, transplantation, or an eGFR<15mL/min/1.73m2 during the follow-up period.

ANALYTICAL APPROACH

Cox proportional hazards regression models evaluated for accuracy using area under the receiver operating characteristic curve (AUC), Brier scores, calibration plots, and continuous net reclassification improvement. We also implemented decision curve analysis. Models developed in the Manitoba cohort were validated in the Ontario cohort.

RESULTS

In the development cohort, 10.3% reached kidney failure after nephrectomy. The final model resulted in a 5-year area under the curve of 0.85 (95% CI, 0.78-0.92) in the development cohort and 0.86 (95% CI, 0.84-0.88) in the validation cohort.

LIMITATIONS

Further external validation needed in diverse cohorts.

CONCLUSIONS

Our externally validated model can be easily applied in clinical practice to inform preoperative discussions about kidney failure risk in patients facing surgical options for localized kidney cancer.

PLAIN-LANGUAGE SUMMARY

Patients with localized kidney cancer often experience a lot of worry about whether their kidney function will remain stable or will decline if they choose to undergo surgery for treatment. To help patients make an informed treatment decision, we developed a simple equation that incorporates 6 easily accessible pieces of patient information to predict the risk of reaching kidney failure 5 years after kidney cancer surgery. We expect that this tool has the potential to inform patient-centered discussions tailored around individualized risk, helping ensure that patients receive the most appropriate risk-based care.

An MRI-based radiomics model to predict clear cell renal cell carcinoma growth rate classes in patients with von Hippel-Lindau syndrome

PURPOSE

Upfront knowledge of tumor growth rates of clear cell renal cell carcinoma in von Hippel-Lindau syndrome (VHL) patients can allow for a more personalized approach to either surveillance imaging frequency or surgical planning. In this study, we implement a machine learning algorithm utilizing radiomic features of renal tumors identified on baseline magnetic resonance imaging (MRI) in VHL patients to predict the volumetric growth rate category of these tumors.

MATERIALS AND METHODS

A total of 73 VHL patients with 173 pathologically confirmed Clear Cell Renal Cell Carcinoma (ccRCCs) underwent MRI at least at two different time points between 2015 and 2021. Each tumor was manually segmented in excretory phase contrast T1 weighed MRI and co-registered on pre-contrast, corticomedullary and nephrographic phases. Radiomic features and volumetric data from each tumor were extracted using the PyRadiomics library in Python (4544 total features). Tumor doubling time (DT) was calculated and patients were divided into two groups: DT <  = 1 year and DT > 1 year. Random forest classifier (RFC) was used to predict the DT category. To measure prediction performance, the cohort was randomly divided into 100 training and test sets (80% and 20%). Model performance was evaluated using area under curve of receiver operating characteristic curve (AUC-ROC), as well as accuracy, F1, precision and recall, reported as percentages with 95% confidence intervals (CIs).

RESULTS

The average age of patients was 47.2 ± 10.3 years. Mean interval between MRIs for each patient was 1.3 years. Tumors included in this study were categorized into 155 Grade 2; 16 Grade 3; and 2 Grade 4. Mean accuracy of RFC model was 79.0% [67.4-90.6] and mean AUC-ROC of 0.795 [0.608-0.988]. The accuracy for predicting DT classes was not different among the MRI sequences (P-value = 0.56).

CONCLUSION

Here we demonstrate the utility of machine learning in accurately predicting the renal tumor growth rate category of VHL patients based on radiomic features extracted from different T1-weighted pre- and post-contrast MRI sequences.

Kidney tumor diffusion-weighted magnetic resonance imaging derived ADC histogram parameters combined with patient characteristics and tumor volume to discriminate oncocytoma from renal cell carcinoma

PURPOSE

To assess the ability to discriminate oncocytoma from RCC based on a model using whole tumor ADC histogram parameters with additional use of tumor volume and patient characteristics.

METHOD

In this prospective study, 39 patients (mean age 65 years, range 28-79; 9/39 (23%) female) with 39 renal tumors (32/39 (82%) RCC and 7/39 (18%) oncocytoma) underwent multiparametric MRI between November 2014 and June 2018. Two regions of interest (ROIs) were drawn to cover both the entire tumor volume and a part of healthy renal cortex. ROI ADC maps were calculated using a mono-exponential model and ADC histogram distribution parameters were calculated. A logistic regression model was created using ADC histogram parameters, radiographic and patient characteristics that were significantly different between oncocytoma and RCC. A ROC curve of the model was constructed and the AUC, sensitivity and specificity were calculated. Furthermore, differences in intra-patient ADC histogram parameters between renal tumor and healthy cortex were calculated. A separate ROC curve was constructed to differentiate oncocytoma from RCC using statistically significant intra-patient parameter differences.

RESULTS

ADC standard deviation (p = 0.008), entropy (p = 0.010), tumor volume (p = 0.012), and patient sex (p = 0.018) were significantly different between RCC and oncocytoma. The regression model of these parameters combined had an ROC-AUC of 0.91 with a sensitivity of 86% and specificity of 84%. Intra-patient difference in ADC 25th percentile (p < 0.01) and entropy (p = 0.030) combined had a ROC-AUC of 0.86 with a sensitivity and specificity of 86%, and 81%, respectively.

CONCLUSION

A model combining ADC standard deviation and entropy with tumor volume and patient sex has the highest diagnostic value for discrimination of oncocytoma. Although less accurate, intra-patient difference in ADC 25th percentile and entropy between renal tumor and healthy cortex can also be used. Although the results of this preliminary study do not yet justify clinical use of the model, it does stimulate further research using whole tumor ADC histogram parameters.