Long-term outcomes of image-guided ablation and laparoscopic partial nephrectomy for T1 renal cell carcinoma

This is a retrospective analysis of a prospectively maintained registry from 2003 to 2016. Following institutional health and research authority approval, consecutive adult patients who underwent image-guided CRYO, RFA, or laparosocpic LPN for cT1N0M0 histology-confirmed RCC were included for the study. The patient selection process at our institution was previously described [14]. cT1 renal masses were defined as a maximum tumour diameter of ≤ 7 cm limited to the kidney on radiographic imaging according to the American Joint Committee on Cancer staging manual [15]; with cT1 further divided to cT1a (≤ 4 cm) and cT1b (> 4 cm and ≤ 7 cm). Patients presenting with multiple renal tumours, recurrence, inherited RCC syndromes, or a solitary kidney were excluded from the analysis [16]. Patients with a history of LPN, CRYO, or RFA of the same kidney were also excluded from analysis. Primary outcome of the study was to evaluate and compare the long-term local recurrence-free survival (LRFS) between CRYO, RFA, and LPN. Secondary outcomes include overall survival (OS), cancer-specific survival (CSS), metastasis-free survival (MFS), rate and severity of complications, and change in renal function peri-operatively. The detailed methods of the performance of IGA and LPN are outlined in the supplementary appendix.

The follow-up protocol for IGA was previously described in detail [14]. All patients were followed at 1, 3, and 6 months after the procedure and annually onwards for a period of 10 years using MRI or CT. Local recurrence was defined as new area(s) of enhancement in the zone of ablation after at least one imaging study had shown complete lack of enhancement in the treated area. Metastatic disease was defined as extra-renal disease on imaging confirmed or suspicioned to have originated from the kidney. Cancer-specific death was defined as any deaths from RCC.

Patient clinical features such as age, sex, treatment date, follow-up details, histopathological details, R.E.N.A.L. nephrometry score [17], co-morbidities (according to the Charlson Comorbidity Index [CCI][18]), procedure details, complications (according to the Clavien Dindo Classification [19]), and estimated glomerular filtration rate (eGFR; CKD-EPI [20]) were extracted from the prospectively maintained database. Utilising the National Health Service (NHS) patient records, the patients were followed for their living status and cause of death until 25th January 2021.

Differences in baseline characteristics were evaluated using the chi-square test and the Kruskal-Wallis test. CSS, OS, LRFS, and MFS were evaluated from the time of treatment to the time of event using the Kaplan-Meier method. Ten-year survival rates and corresponding 95% confidence intervals (95% CI) were reported. The Cox proportional hazard regression model was utilised to evaluate survival in CRYO, RFA, and LPN patients, reporting as hazard ratios (HRs), 95% CI, and p-values. To allow evaluation of HRs when no events were observed in an arm, an event was artificially created at the latest follow-up for that arm. Complication rates and severity were evaluated using the chi-squared test and logistic regression. Changes in peri-operative renal function were evaluated using the Kruskal-Wallis test and the Wilcoxon matched pairs signed rank sum test. Propensity score matching [21, 22] was intended to be used to compare RFA and CRYO with LPN. However, the groups were so different for a number of the key matching variables that this approach became impractical, as detailed in the “Results” section, and in the supplementary appendix. In order to facilitate comparison of the two groups, we therefore used Cox’s proportional hazards model [23], including all the variables we had intended to use in the matching analysis (age, sex, laterality, CCI, R.E.N.A.L nephrometry score, lesion size, RCC type, grade, and t-stage), to adjust for imbalances in these variables between the various treatment groups. This is, of course, not a substitute for a randomised trial, and the results have to be interpreted with a degree of caution. Furthermore, there are relatively few events, creating sensitivity issues with the model results. However, given the difficulties in undertaking such a randomised trial, and the time that such a trial would take to complete, we decided that this approach provides an appropriate means of performing and interpreting inter-group comparisons. Amongst 10 patients with missing CCI or R.E.N.A.L. nephrometry score, median CCI or nephrometry score was imputed. Sensitivity analyses have shown identical results and hence all patients were included in the final analyses. All analyses are two-tailed at a significance level of 0.05. All statistical analyses were performed on STATA/MP 16.0 (StataCorp).

A total of 58 T1b patients were included in this study. A summary of their clinical and pathological characteristics are outlined in Table 2. RCC histology, Fuhrman grade, age, tumour size, R.E.N.A.L nephrometry score, baseline eGFR, and CCI were found to be significantly different between the three groups. The median (IQR) follow-up duration is 72.5 (42.0–100.9) months, 59.5 (27.5–99.39) months, and 67.9 (50.8–91.3) months for CRYO, RFA, and LPN, respectively. CSS, OS, LRFS, and MFS are all comparable between patients undergoing CRYO, RFA, or LPN (Figs. 1 and 3). The details of the results are outlined in the supplementary appendix.

The rate and severity of post-operative complications for all three modalities were found to be similar in both cT1a (CRYO: 11.1%, RFA: 18.4%, LPN: 14.1%) and cT1b patients (CRYO: 19.4%, RFA: 15.4%, LPN: 7.7%). Both logistic regression and multinomial logistic regression did not show significant difference between the three groups’ rate and severity of complications (Supplementary Table 1 and 2). A summary of all complications occurring during the study period are reported in Supplementary Table 3.

The post-operative eGFR and change in eGFR peri-operatively of T1a and T1b patients undergoing CRYO, RFA, and LPN are shown in Table 3. Only small changes in eGFR were found in patients undergoing CRYO and RFA, as compared to substantial falls in eGFR in LPN patients (Wilcoxon matched pairs signed rank sum Z and p-values; CRYO: 3.0, 0.003, RFA: 2.4, 0.02, LPN: 6.0, < .0001). When comparing the change in renal function peri-operatively using the Wilcoxon 2-sample rank sum test, in both T1a (Z = 4.1, p < .0001) and T1b (Z = 2.5, p = .01) patients, those undergoing IGA had a significantly smaller median change in eGFR compared to LPN (Table 3).

Initially, it was intended to explore the propensity score matching approach, as described in the “Methods” section. However, this proved to be infeasible due to large differences in baseline factors between the treatment groups, most substantially in age (Supplementary Figure 4; Tables 1 and 2). Further details, results, and explanation are given in supplementary Figures 2 and 3. Therefore, as described in the “Methods” section, the Cox multivariate method was used to adjust for these imbalances and compare the treatment arms (Table 4). As events are relatively scarce in this study, sensitivity analyses were performed by replacing an event with censoring at that time (results not presented). Minimal differences to the results presented were observed for all of the outcomes, demonstrating that the results are relatively insensitive to such small changes, and are therefore relatively robust. Certainly, the overall findings would be unchanged as a result of a single patient having a different outcome.

In univariate Kaplan-Meier analyses, IGA and LPN were shown to have comparable LRFS. However, given that the CRYO and RFA groups consist of patients with considerably worse prognostic factors, after multivariate adjustment, CRYO and RFA appear to be superior to LPN for LRFS. The magnitude of the effect in the two ablative therapy groups is almost identical (see Supplementary Figure 5) so a combined group analysis, stratified by group, was performed, demonstrating ablative therapies to be superior to LPN for LRFS (HR 0.006, 95% CI 0.00–0.15, p = 0.002). Note that the RFA/LPN comparison reaches statistical significance on its own (Table 4; p = 0.003), and, although the CRYO/LPN result is not statistically significant (Table 4, p = 0.087), this is largely a result of paucity of patient and event numbers. Although effect sizes (HR) appear to be substantial for statistically significant outcomes (LRFS, MFS), suggesting extreme advantage to IGA patients, they are unlikely to reflect real effect sizes due to a combination of the extreme selection bias. Finally, the lower 90% confidence interval on the hazard ratio is less than 1 for CRYO (Supplementary Table 5), which demonstrates at least 90% confidence that CRYO is as good as LPN for LRFS. For clarity, characteristics of all patients with T1a tumours and subsequent local recurrences are shown in Table 5.