Abstract
Background:
Sequential intravesical gemcitabine and docetaxel (Gem/Doce) are increasingly used for non–muscle-invasive bladder cancer (NMIBC), especially in the setting of bacillus Calmette-Guérin (BCG) shortage or failure. However, the contribution of docetaxel to gemcitabine remains unquantified. We compared the effectiveness of gemcitabine alone vs Gem/Doce in high-risk NMIBC.
Methods:
We conducted a retrospective study of 296 patients treated with gemcitabine (n=173) or Gem/Doce (n=123) at 3 Mayo Clinic sites between 2018 and 2023. The primary outcome was high-grade recurrence-free survival (HGRFS). Secondary outcomes included recurrence-free, progression-free, cystectomy-free, cancer-specific, and overall survival and adverse events. Analyses included Kaplan-Meier and Cox regression with 2-stage residual inclusion to adjust for confounding.
Results:
Median HGRFS was similar between groups (gemcitabine: 22.6 months; Gem/Doce: 19.5 months [P=.25]). Among patients with BCG-unresponsive carcinoma in situ (n=49), median HGRFS was comparable (gemcitabine: 12.2 months; Gem/Doce: 7.6 months [P=.12]). No significant differences were observed for secondary outcomes. In the 2-stage residual inclusion analysis, Gem/Doce was not associated with improved HGRFS (hazard ratio [HR], 1.09; 95% CI, 0.65–1.82; P=.80), including among patients with BCG-unresponsive disease (HR, 2.24; 95% CI, 0.96–5.24 [P=.06]). Low-grade adverse events were common (gemcitabine: 63%; Gem/Doce: 55% [P=.19]), but grade ≥3 adverse events were rare (2.3% vs 2.4%). Limitations included imbalances in maintenance utilization (gemcitabine: 19%; Gem/Doce: 40%), baseline characteristics, gemcitabine dosing, and follow-up duration between cohorts.
Conclusions:
In this multicenter study, Gem/Doce was not associated with improved outcomes compared with gemcitabine alone. Prospective trials are needed to quantify the specific contribution of docetaxel in NMIBC.
Keywords: comparative effectiveness research, docetaxel, gemcitabine, intravesical therapy, non–muscle-invasive bladder neoplasms
1. Introduction
The incidence of bladder cancer is projected to increase by 73% by 2040.1 Of patients with bladder cancer, approximately 75% present with non–muscle-invasive cancer (NMIBC), and over half of these cancers have high-risk features.2, 3 Standard treatment for high-risk NMIBC includes complete transurethral resection of bladder tumor (TURBT), followed by adjuvant intravesical bacille Calmette-Guérin (BCG) therapy. Despite adequate treatment, over 40% of patients experience recurrence, and BCG-unresponsiveness disease develops in approximately 20%, a challenge compounded by ongoing global BCG shortages.4, 5
These limitations have led to increased use of alternative intravesical agents, including gemcitabine alone or combined with docetaxel. Sequential use of gemcitabine and docetaxel (Gem/Doce) has been associated with promising outcomes in some observational studies, particularly in BCG-unresponsive disease.6 However, the individual contribution of each agent remains unclear, and prior studies have used inconsistent end points and variable definitions for BCG failure. Sequential Gem/Doce also poses logistical barriers, including extended chair time, sequential instillation, staffing demands, and reimbursement issues, which may limit its adoption, especially in community settings.7
To address this gap in our understanding, we conducted a retrospective, multisite comparative effectiveness study of gemcitabine monotherapy vs sequential Gem/Doce therapy in patients with high-risk NMIBC, including a subgroup with BCG-unresponsive disease as defined by strict US Food and Drug Administration (FDA) critieria.8 Our aim was to characterize the comparative effectiveness and tolerability of these regimens using standardized outcomes, thereby informing future clinical trial designs.
2. Materials and Methods
2.1. Patients
The study was reviewed and approved by the Mayo Clinic Institutional Review Board (Protocol No. 21–011911).
We retrospectively reviewed health records of patients with high-risk NMIBC as defined by American Urological Association/Society of Urologic Oncology criteria.2 Patient confidentiality was maintained through standard deidentification processes. To be eligible for inclusion, patients must have received an induction course of either gemcitabine or sequential Gem/Doce from January 1, 2018, through December 18, 2023, at 1 of 3 Mayo Clinic sites. Follow-up data were collected through March 14, 2025. For a subgroup analysis within this cohort, BCG-unresponsive disease was defined as specified by the FDA draft guidance,8 and BCG-exposed disease was defined per the International Bladder Cancer Group consensus statement.9
2.2. Intravesical chemotherapy
The treatment protocol for patients receiving single-agent gemcitabine was based on that used for the phase II S0353 within the Southwest Oncology Group (SWOG).10 Patients received 2 g of gemcitabine in 100 mL of saline, instilled into the bladder via a catheter for a minimum of 60 minutes once weekly for 6 weeks.
Treatment for patients receiving sequential Gem/Doce was based on the protocol published by Steinberg et al.11 Patients received 1 g of gemcitabine in 50 mL of saline, instilled via catheter for 60 to 90 minutes. The bladder was then drained, and 37.5 mg of docetaxel in 50 mL of saline was instilled via a catheter for 60 to 120 minutes. This treatment sequence was administered weekly for 6 weeks. Patients with no evidence of tumor recurrence after induction were offered maintenance therapy at the discretion of the treating team. Most patients receiving maintenance therapy underwent either 1 instillation every 4 weeks or 1 instillation weekly for 3 weeks, repeated every 3 months.
2.3. Outcomes
The primary outcome was high-grade recurrence-free survival (HGRFS), defined as the time from initial chemotherapy instillation to documented high-grade bladder tumor recurrence, progression, cystectomy, or death. Secondary outcomes included adverse events (AEs) and the following time-to-event measures: recurrence-free survival (defined as time from initial chemotherapy instillation to documented bladder tumor recurrence, progression, cystectomy, or death), progression-free survival (defined as time from initial chemotherapy instillation to muscle-invasive bladder cancer, lymph node involvement, metastatic disease, or death), cystectomy-free survival (defined as time from initial chemotherapy instillation to cystectomy or death), cancer-specific survival (defined as time from initial chemotherapy instillation to death due to bladder cancer), and overall survival (defined as time from initial chemotherapy instillation to death due to any cause).
The prevalence and severity of AEs were documented using the National Cancer Institute’s Common Terminology Criteria for Adverse Events (version 5.0).12 Patients lost to follow-up were censored at their last documented urologic evaluation (for bladder outcomes) or last known contact (for overall and cancer-specific survival).
2.4. Statistical Analysis
Patient characteristics were summarized as median with interquartile range (IQR) for continuous variables and as frequency counts with percentages for categorical variables. Continuous and categorical variables were compared between the groups receiving single-agent gemcitabine and sequential Gem/Doce using the Wilcoxon rank sum test and Fisher exact test, respectively. Event-free survival was estimated using the Kaplan-Meier method, and differences between groups were compared using the log-rank test. Cancer-specific survival was analyzed using cumulative incidence functions, with the Gray test used for between-group comparisons treating death from other causes as a competing risk. Point estimates and 95% CIs were reported. Kaplan-Meier curves for high-grade recurrence-free survival regarding the entire cohort and sub-cohort were presented up to the latest time point at which the number at risk exceeded 5 in either group to avoid overinterpretation of estimates based on very sparse risk sets. All time-to-event analyses were conducted using the full available follow-up.
2.5. Instrumental Variable Analysis
We conducted an instrumental variable analysis to adjust for confounding by indication. The year of intravesical chemotherapy induction was selected as the instrumental variable based on Steinberg et al’s11 2020 publication of a practice-changing, retrospective, multi-institutional study on the use of sequential Gem/Doce for NMBIC. Given the significant association between year of treatment and treatment group (Supplemental Table), we empirically tested the relevance criterion in the first-stage model of the instrumental variable analysis. The model yielded an F statistic of 19.7, exceeding the conventional threshold of 10, indicating that the instrument met the relevance criterion.
We then conducted a 2-stage residual inclusion method within a Cox proportional hazards framework. Along with other covariates, we adjusted for BCG cohort (BCG-naïve, BCG-exposed, and BCG-unresponsive), a known confounder of the primary outcome (HGRFS), and for BCG shortage, which may have mediated the relationship between the BCG cohort and treatment group and the receipt of maintenance chemotherapy. These adjustments supported the exchangeability assumption by accounting for confounding and helped protect the exclusion restriction by mitigating other pathways between the instrument and outcome.
Statistical analyses were conducted in Posit team (2025, RStudio version 4.4.1: Integrated Development Environment for R Posit Software, PBC; http://www.posit.co/). All tests were 2-sided and P values <.05 were considered significant. Missing or unknown values were reported as unknown and excluded from analysis.
3. Results
3.1. Patients
During the study period (2018–2023), 312 patients with high-risk NMIBC received intravesical gemcitabine monotherapy or sequential Gem/Doce. After excluding 16 patients for incomplete surveillance (n=4) or receipt of nonstandard regimens (n=12), 296 patients were included in the final analysis: 173 received single-agent gemcitabine, and 123 received sequential Gem/Doce.
Baseline demographic and clinical characteristics are shown in Table 1. Patients treated with gemcitabine were more likely to be current or former smokers (72% vs 59%, P=.02) and to have treatment plans influenced by BCG shortage (45% vs 28%, P=.003). They were less likely to have BCG-unresponsive disease (18% vs 47%, P<.001), to have received postoperative single-dose gemcitabine (17% vs 29%, P=.02), or to be receiving maintenance chemotherapy (19% vs 40%, P<.001). The distribution of variant histologic findings was similar between groups (7% vs 8%, P=1.00). The treatment group varied significantly by year, with gemcitabine use declining and sequential Gem/Doce use increasing over time (Supplemental Table, P<.01). One patient with a missing tumor grade and stage was excluded from that specific analysis; no other data were missing.
Table 1.
Patient Characteristics at Baseline (N=296)a
| Characteristic | Single-agent gemcitabine (n=173) | Sequential gemcitabine/docetaxel (n=123) | P value |
|---|---|---|---|
| Age, median (IQR), y | 75 (69–81) | 75 (69–80) | .7 |
| Sex, female | 29 (16.8) | 21 (17.1) | 1 |
| Race | .068 | ||
| White | 163 (94.2) | 117 (95.1) | |
| Other | 10 (5.8) | 6 (4.9) | |
| Smoker | .017 | ||
| Yes (current or former) | 125 (72.3) | 72 (58.5) | |
| Never | 48 (27.7) | 51 (41.5) | |
| Location | .4 | ||
| Arizona | 57 (32.9) | 35 (28.5) | |
| Florida | 65 (37.6) | 42 (34.1) | |
| Minnesota | 51 (29.5) | 46 (37.4) | |
| Stage at original diagnosis | .4 | ||
| LG Ta only | 27 (15.6) | 17 (13.8) | |
| HG Ta only | 76 (43.9) | 53 (43.1) | |
| Carcinoma in situ with or without Ta | 14 (8.1) | 19 (15.4) | |
| T1 only | 46 (26.6) | 30 (24.4) | |
| T1 + carcinoma in situ | 9 (5.2) | 4 (3.3) | |
| Unknown | 1 (0.6) | 0 (0.0) | |
| Pretreatment stage | .19 | ||
| HG Ta only | 103 (59.5) | 59 (48.0) | |
| Carcinoma in situ with or without Ta | 49 (28.3) | 43 (35.0) | |
| T1 only | 17 (9.8) | 19 (15.4) | |
| T1 + carcinoma in situ | 4 (2.3) | 2 (1.6) | |
| Pretreatment, CIS-containing tumor | .3 | ||
| Yes | 53 (30.6) | 45 (36.6) | |
| No | 120 (69.4) | 78 (63.4) | |
| BCG status | <.001 | ||
| Naïve/intolerantb | 117 (67.6) | 54 (43.9) | |
| Exposedc | 25 (14.5) | 11 (8.9) | |
| Unresponsived | 31 (17.9) | 58 (47.2) | |
| Number of prior BCG induction courses | .009 | ||
| 0 | 97 (56.1) | 47 (38.2) | |
| 1 | 42 (24.3) | 39 (31.7) | |
| 2+ | 34 (19.7) | 37 (30.1) | |
| BCG shortage influenced treatment plan | .003 | ||
| Yes | 77 (44.5) | 34 (27.6) | |
| Not documented | 96 (55.5) | 89 (72.4) | |
| Completed chemotherapy induction | .8 | ||
| No (failed induction) | 10 (5.8) | 8 (6.5) | |
| Yes (5+ doses) | 163 (94.2) | 115 (93.5) | |
| Received chemotherapy maintenance | <.001 | ||
| No | 141 (81.5) | 74 (60.2) | |
| Yes | 32 (18.5) | 49 (39.8) | |
| Previous single-dose postoperative intravesical gemcitabine | 29 (16.8) | 35 (28.5) | .021 |
| TURBT with blue-light cystoscopy immediately before first chemotherapy instillation | 25 (14.5) | 27 (22.0) | .12 |
| Time from initial diagnosis of bladder cancer to chemotherapy induction, median (IQR), mo | 9.0 (1.6–38.3) | 13.6 (3.4–39.2) | .086 |
Abbreviations: BCG, bacille Calmette-Guérin; CIS, carcinoma in situ; HG, high-grade; LG, low-grade; TURBT, transurethral resection of bladder tumor.
Data are displayed as No. (%) unless indicated otherwise. Continuous and categorical variables were compared between groups using the Wilcoxon rank sum test and Fisher exact test, respectively.
BCG-naïve disease was defined as the presence of disease without BCG administration in the preceding 24 months. BCG-intolerant disease was defined as persistent disease resulting from an inability to receive adequate BCG due to treatment-related toxicity.
BCG-exposed disease was defined as meeting any of the following criteria: 1) persistent or recurrent CIS or HG Ta disease at 3 months after induction BCG; 2) initial complete response 3 months after induction BCG, followed by HG recurrence within 24 months; 3) initial complete response after reinduction or maintenance BCG, followed by HG recurrence outside the BCG-unresponsive window: between 6 and 24 months for Ta/T1 disease and between 12 and 24 months for CIS.
BCG-unresponsive disease was defined as meeting any of the following criteria: 1) HG T1 disease present 3 months after induction BCG; 2) recurrent or persistent CIS or HG Ta/T1 after reinduction or maintenance BCG therapy; 3) initial complete response after reinduction or maintenance BCG, followed by HG recurrence: within 6 months of the last BCG dose for HG papillary tumors and within 12 months for CIS.
The median (IQR) follow-up for overall survival was 56.8 (34.5–63.8) months in the gemcitabine group and 23.9 (18.4–32.6) months in the sequential Gem/Doce group. In the gemcitabine group, 27 patients (16%) were lost to follow-up vs 10 patients (8.1%) lost to follow-up in the sequential Gem/Doce group.
Among patients receiving gemcitabine, 91 (53%) had high-grade recurrence, 15 (9%) had progression to muscle-invasive disease, 7 (4%) had metastases, 30 (17%) died of all-cause mortality), and 7 (4%) died of bladder cancer. Among those treated with Gem/Doce, 63 (51%) had high-grade recurrence, 5 (4%) had progression to muscle-invasive disease, 12 (10%) had metastases, 13 (11%) died of all-cause mortality, and 4 (3%) died of bladder cancer.
3.2. High-grade recurrence-free survival
In the overall cohort (Figure 1), the median (IQR) HGRFS was 22.6 (15.4–36.7) months in the gemcitabine group and 19.5 (12.6–23.8) months in the sequential Gem/Doce group among patients who remained recurrence-free. Event-free survival estimates for HGRFS were as follows (Table 2):
At 6 months: 81% (95% CI, 75%−87%) with gemcitabine and 79% (95% CI, 72%−87%) with sequential Gem/Doce.
At 1 year: 61% (95% CI, 54%−68%) and 62% (95% CI, 54%−71%), respectively.
At 2 years: 49% (95% CI, 42%−57%) and 40% (95% CI, 30%−51%), respectively.
At 3 years: 43% (95% CI, 36%−52%) and 27% (95% CI, 15%−49%), respectively.
Among patients without high-grade recurrence, median (IQR) follow-up was 45.8 (25.3–56.2) months in the gemcitabine group and 21.0 (15.9–24.2) months in the sequential Gem/Doce group.
Figure 1.
High-Grade Recurrence-Free Survival. Survival curves were truncated (dotted line) after 36.0 months, when the number of patients at risk in the sequential gemcitabine and docetaxel group decreased to fewer than 5.
Table 2.
Unadjusted Survival Estimates Stratified by Pretreatment and Treatment Groups
| Estimated survival rate (event-free survival) or cumulative incidence (event-specific survival), % (95% CI) | Follow-up, median (IQR), mob | ||||||
|---|---|---|---|---|---|---|---|
| Outcome | Cohort | Treatment (No.)a | 6 mo | 1 y | 2 y | 3 y | |
| RFS | All | Gemcitabine (173) | 77 (71–83) | 51 (44–59) | 37 (31–45) | 29 (23–38) | 46.1 (23.8–57.0) |
| Gem/Doce (123) | 79 (72–86) | 59 (51–69) | 36 (27–48) | 23 (11–47) | 21.0 (16.2–23.9) | ||
| BCG-N/I | Gemcitabine (117) | 82 (75–89) | 56 (47–66) | 43 (35–53) | 34 (26–44) | 48.2 (28.1–58.4) | |
| Gem/Doce (54) | 87 (79–97) | 72 (61–85) | 48 (34–68) | 48 (34–68) | 21.0 (16.6–23.0) | ||
| BCG-UR | Gemcitabine (31) | 80 (67–96) | 47 (32–69) | 32 (19–55) | 28 (15–51) | 23.5 (14.6–55.5) | |
| Gem/Doce (58) | 68 (57–82) | 48 (37–63) | 25 (15–42) | 25 (15–42) | 20.6 (14.5–24.3) | ||
| HGRFS | All | Gemcitabine (173) | 81 (75–87) | 61 (54–68) | 49 (42–57) | 43 (36–52) | 45.8 (25.3–56.2) |
| Gem/Doce (123) | 79 (72–87) | 62 (54–71) | 40 (30–51) | 27 (15–49) | 21.0 (15.9–24.2) | ||
| BCG-N/I | Gemcitabine (117) | 84 (78–91) | 66 (58–75) | 56 (47–66) | 48 (39–59) | 47.4 (28.4–56.4) | |
| Gem/Doce (54) | 87 (79–97) | 76 (65–88) | 56 (42–75) | 56 (42–75) | 21.0 (16.4–24.1) | ||
| BCG-UR | Gemcitabine (31) | 87 (76–100) | 53 (38–75) | 39 (25–62) | 35 (21–58) | 45.9 (14.8–53.9) | |
| Gem/Doce (58) | 70 (59–83) | 50 (39–65) | 25 (15–42) | 25 (15–42) | 20.6 (14.5–24.3) | ||
| BCG-UR; any CIS | Gemcitabine (17) | 82 (66–100) | 53 (34–83) | 41 (23–73) | 33 (16–68) | 49.9 (29.1–56.2) | |
| Gem/Doce (32) | 62 (47–81) | 49 (34–70) | 8.4 (1.6–44) | NE | 16.8 (14.1–21.1) | ||
| BCG-UR; HG Ta/T1 | Gemcitabine (14) | 92 (79–100) | 54 (33–89) | 36 (17–77) | 36 (17–77) | 14.8 (9.0–33.4) | |
| Gem/Doce (26) | 80 (66–97) | 51 (35–76) | 41 (25–68) | 41 (25–68) | 22.9 (17.5–27.6) | ||
| PFS | All | Gemcitabine (173) | 99 (97–100) | 96 (94–99) | 87 (82–92) | 79 (73–86) | 48.2 (27.3–59.2) |
| Gem/Doce (123) | 94 (90–99) | 91 (86–96) | 80 (72–89) | 70 (58–84) | 20.6 (15.8–25.3) | ||
| BCG-N/I | Gemcitabine (117) | 100 (NE) | 98 (96–100) | 91 (86–96) | 81 (73–89) | 48.3 (28.5–57.1) | |
| Gem/Doce (54) | 96 (91–100) | 93 (86–100) | 85 (74–98) | 69 (49–96) | 20.5 (16.3–24.4) | ||
| BCG-UR | Gemcitabine (31) | 97 (91–100) | 90 (80–100) | 72 (57–91) | 67 (51–88) | 45.9 (25.4–56.8) | |
| Gem/Doce (58) | 93 (86–100) | 89 (81–98) | 75 (63–90) | 68 (51–89) | 19.9 (14.9–27.5) | ||
| BCG-UR; any CIS | Gemcitabine (17) | 100 (NE) | 88 (74–100) | 64 (44–92) | 56 (36–87) | 53.9 (27.3–57.0) | |
| Gem/Doce (32) | 90 (79–100) | 90 (79–100) | 74 (57–95) | 65 (46–92) | 20.3 (14.0–36.9) | ||
| BCG-UR; HG Ta/T1 | Gemcitabine (14) | 92 (79–100) | 92 (79–100) | 83 (64–100) | 83 (64–100) | 36.6 (18.1–53.8) | |
| Gem/Doce (26) | 96 (89–100) | 88 (76–100) | 79 (65–97) | 79 (65–97) | 19.8 (17.3–23.2) | ||
| CFS | All | Gemcitabine (173) | 98 (96–100) | 95 (92–99) | 88 (83–93) | 78 (71–85) | 45.9 (26.0–57.1) |
| Gem/Doce (123) | 96 (92–100) | 92 (87–97) | 85 (79–92) | 82 (73–92) | 21.6 (16.6–18.5) | ||
| CSS | All | Gemcitabine (173) | 0 (NE) | 0.58 (0.05–3.00) | 0.58 (0.05–3.00) | 2.7 (0.9–6.4) | 56.8 (34.5–63.8) |
| Gem/Doce (123) | 0 (NE) | 0.83 (0.07–4.10) | 3.1 (0.8–8.1) | 3.1 (0.8–8.1) | 23.9 (18.4–32.6) | ||
| OS | All | Gemcitabine (173) | 99 (98–100) | 98 (96–100) | 94 (90–97) | 87 (82–93) | 56.8 (34.5–63.8) |
| Gem/Doce (123) | 99 (98–100) | 98 (95–100) | 91 (86–97) | 89 (82–97) | 23.9 (18.4–32.6) | ||
Abbreviations: BCG, bacille Calmette-Guérin; CFS, cystectomy-free survival; CSS, cancer-specific survival; Gem/Doce, sequential gemcitabine/docetaxel; HGRFS, high-grade recurrence-free survival; NE, not estimable; N/I, naïve/intolerant; OS, overall survival; PFS, progression-free survival; RFS, recurrence-free survival; UR, unresponsive.
Number in each cohort at the beginning of induction.
Median follow-up for patients surviving the outcome of interest.
3.3. Secondary outcomes
At 2 years, recurrence-free survival (RFS) was 37% (95% CI, 31%−45%) for gemcitabine and 36% (95% CI, 27%−48%) for sequential Gem/Doce (Table 2). Progression-free survival was 87% (95% CI, 82%−92%) for gemcitabine and 80% (95% CI, 72%−89%) for sequential Gem/Doce. Overall survival was 94% (95% CI, 90%−97%) and 91% (95% CI, 86%−97%), respectively. The 2-year cumulative incidence of cancer-specific death was 0.6% (95% CI, 0.01%−3.0%) in the gemcitabine group and 3.1% (95% CI, 0.8%−8.1%) in the sequential Gem/Doce group.
Cystectomy-free survival at 2 years was 88% (95% CI, 83%−93%) for gemcitabine and 85% (95% CI, 79%−92%) for the sequential Gem/Doce group. Sixteen patients in the gemcitabine group and 9 in the sequential Gem/Doce group underwent cystectomy. Indications for cystectomy included recurrence (10 vs 7 patients), disease progression (5 vs 1 patient), and treatment intolerance (0 vs 1 patient). Pathologic findings revealed T2 or more advanced disease in 7 patients treated with gemcitabine and 3 patients treated with sequential Gem/Doce. Lymph node involvement was identified in 1 patient (pT4aN2) treated with gemcitabine and in no patients treated with sequential Gem/Doce.
3.4. Adverse events
AEs were reported by 109 patients (63%) treated with gemcitabine and 68 patients (55%) treated with sequential Gem/Doce (P=.19), with most events being grade 1 or 2 (Table 3). The most frequently reported grade 1 and 2 events included incomplete dwell time (31%), bladder spasms (28%), urinary urgency/frequency (21%), dysuria (16%), and hematuria (11%). Grade 3 AEs occurred in 4 patients (2.3%) in the gemcitabine group and 3 patients (2.4%) in the sequential Gem/Doce group (P=1.00). Induction was discontinued due to AEs in 10 gemcitabine-treated patients (5.8%) and in 8 sequential Gem/Doce–treated patients (6.5%) (P=.80). Treatment schedules were modified for 22 patients (13%) in the gemcitabine group and 10 patients (8.1%) in the sequential Gem/Doce group (P=.30). No grade 4 or 5 AEs were reported.
Table 3.
Adverse Events of Intravesical Gemcitabine With and Without Docetaxel
| Patients, No. (%) | |||
|---|---|---|---|
| Adverse events | Single-agent gemcitabine (n=173) | Sequential gemcitabine/docetaxel (n=123) | P valuea |
| Any | 109 (63) | 68 (55) | .19 |
| Adverse events causing induction failure | 10 (5.8) | 8 (6.5) | .8 |
| Adverse events affecting treatment schedule | 22 (13) | 10 (8.1) | .3 |
| Grade 3 adverse event (any type) | 4 (2.3) | 3 (2.4) | 1 |
| Event type (any grade) | |||
| Failure to retain chemotherapy for the full 1-hour dwell time | 54 (31) | 37 (30) | .9 |
| Bladder spasm | 43 (25) | 40 (33) | .15 |
| Urinary frequency/urgency | 40 (23) | 21 (17) | .2 |
| Dysuria | 30 (17) | 18 (15) | .6 |
| Hematuria | 17 (10) | 18 (15) | .3 |
| Fever/chills | 17 (10) | 9 (7.3) | .8 |
| Fatigue | 16 (9.2) | 7 (5.7) | .5 |
| Urinary retention | 11 (6.4) | 2 (1.6) | .11 |
| Urinary incontinence | 6 (3.5) | 7 (5.7) | .4 |
| Nausea/vomiting | 8 (4.6) | 5 (4.1) | 1 |
| Flank pain | 4 (2.3) | 7 (5.7) | .12 |
| Diarrhea | 5 (2.9) | 4 (3.3) | 1 |
| Urticaria/rash | 4 (2.3) | 1 (0.81) | .4 |
| Abdominal pain | 3 (1.7) | 2 (1.6) | 1 |
| Myalgias | 4 (2.3) | 0 | .14 |
| Headache | 1 (0.58) | 2 (1.6) | .7 |
| Dizziness | 1 (0.58) | 2 (1.6) | .7 |
Between-group comparisons were calculated using the Fisher exact test.
3.5. Instrumental variable analysis
Instrumental variable analysis using year of treatment as the instrument showed no significant association between treatment group and HGRFS (hazard ratio [HR], 1.09; 95% CI, 0.65–1.82; P=.80) after adjusting for measured and unmeasured confounders (Table 4). Substantial temporal separation occurred in treatment assignment, with gemcitabine predominating before 2021 and sequential Gem/Doce thereafter (Supplemental Table). Importantly, year of treatment was not independently associated with recurrence risk, supporting the validity of the exclusion restriction.
Table 4.
Instrumental Variable Analysis of High-Grade Recurrence-Free Survival
| Estimated high-grade recurrence-free survival | |||
|---|---|---|---|
| Covariate/level | No. | HR (95% CI) | P valuea |
| Treatment group | |||
| Gem/Doce | 123 | 1.09 (0.65–1.82) | .8 |
| Gemcitabine | 173 | Referent | |
| Age | 296 | 1.00 (0.98–1.01) | .8 |
| Sex | |||
| Female | 50 | 1.44 (0.98–2.12) | .07 |
| Male | 246 | Referent | |
| Race | |||
| White | 280 | 1.15 (0.56–2.38) | .7 |
| Other | 16 | Referent | |
| Smoker | |||
| Yes (current or former) | 197 | 0.87 (0.63–1.22) | .4 |
| No | 99 | Referent | |
| Location | |||
| Arizona | 92 | 1.41 (0.94–2.13) | .10 |
| Florida | 107 | 1.36 (0.93–2.00) | .11 |
| Minnesota | 97 | Referent | |
| Disease type | |||
| HG Ta/T1 | 198 | 0.71 (0.51–1.01) | .056 |
| Any CIS | 98 | Referent | |
| BCG status | |||
| Unresponsive | 89 | 2.16 (1.43–3.28) | <.001 |
| Exposed | 36 | 2.04 (1.28–3.24) | .003 |
| Naïve/intolerant | 171 | Referent | |
| BCG shortage influenced treatment plan | |||
| Yes | 111 | 1.15 (0.83–1.59) | .4 |
| Not documented | 185 | Referent | |
| Chemotherapy maintenance therapy | |||
| Yes | 81 | 0.50 (0.34–0.73) | <.001 |
| No | 215 | Referent | |
Abbreviations: BCG, bacille Calmette-Guérin; CIS, carcinoma in situ; HG, high-grade; HR, hazard ratio.
Values were derived from a 2-stage residual inclusion method within a Cox proportional hazards framework.
3.6. Sensitivity analysis of the BCG-unresponsive subgroup
Among the BCG-unresponsive subgroup, 31 patients received gemcitabine and 58 received sequential Gem/Doce. AEs resulted in discontinuing of induction for 6 of 58 BCG-unresponsive patients (10%) who received sequential Gem/Doce and 0 patients who received gemcitabine. The median (IQR) HGRFS was 14.4 (9.2–43.0) months for gemcitabine and 12.1 (6.5–20.5) months for sequential Gem/Doce. Among those with BCG-unresponsive carcinoma in situ (CIS), the median (IQR) was 12.2 (8.8–not estimable) months and 7.6 (5.7–17.6) months, respectively (Figure 2). The 2-year HGRFS estimates were 39% (95% CI, 25%−62%) for gemcitabine and 25% (95% CI, 15%−42%) for sequential Gem/Doce. Instrumental variable analysis again showed no statistically significant difference in HGRFS between treatment groups (HR, 2.24; 95% CI, 0.96–5.24; P=.06).
Figure 2.
High-Grade Recurrence-Free Survival Among Patients With BCG-Unresponsive Disease. A, Overall; B, Carcinoma in situ–Containing Disease; C, Ta/T1-Only Disease. Survival curves were truncated (dotted lines) when the number of patients at risk decreased to fewer than 5: at 22.5 months in panel A, at 14.4 months in panel B, and at 21.2 months in panel C.
4. Discussion
In this retrospective analysis of patients with high-risk NMIBC, treatment with intravesical sequential Gem/Doce was not associated with improved HGRFS compared with gemcitabine monotherapy, even after adjusting for known confounders (eg, treatment selection) using instrumental variable methods. This null association remained consistent across all secondary outcomes and all disease states, including the highest-risk subgroup: patients with BCG-unresponsive CIS. These findings underscore the need for further evaluation of the additive benefit of combination therapy, ideally through a randomized clinical trial.
Our results contrast with prior retrospective observational cohort studies reporting favorable 2-year HGRFS rates with sequential Gem/Doce, particularly for BCG-unresponsive disease. Steinberg et al11 reported a 52% HGRFS rate in a multi-institutional cohort of 105 patients with BCG-unresponsive disease; Yim et al13 reported a 49% HGRFS in a cohort of 102 BCG-unresponsive patients; and Chevuru et al14 reported 50% HGRFS for 34 BCG-unresponsive patients. Recently, Taylor et al15 reported 62% HGRFS in a cohort of 95 BCG-unresponsive patients, and Scilipoti et al16 reported 54% HGRFS in a cohort of 36 BCG-refractory patients. We observed 2-year HGRFS rates of 40% with sequential Gem/Doce and 49% with gemcitabine alone; however, among BCG-unresponsive patients, the rates were 25% and 39%, respectively. The differences between treatment groups in our study and between our findings and those of prior reports may reflect a combination of clinical and methodological factors. The most plausible explanation is baseline risk imbalance. Patients selected for sequential Gem/Doce appeared to have more treatment-refractory disease, as evidenced by the higher prevalence of BCG-unresponsive disease in this group. Alternatively, patients treated during shortage periods (gemcitabine: 45%; Gem/Doce: 28%) may have been less likely to meet BCG-unresponsive criteria regardless of their true BCG-sensitivity because classification as BCG-unresponsive disease requires adequate prior BCG exposure. Although our use of calendar year as an instrumental variable helped mitigate confounding by indication, residual confounding likely remained.
Sequential Gem/Doce was not associated with improved 2-year HGRFS compared with gemcitabine monotherapy in a subset analysis restricted to patients with BCG-unresponsive CIS, a population widely regarded as the most difficult to treat. Although the sample size was small, this null association held in both unadjusted and adjusted models. It also extended across all secondary end points, including RFS, progression-free survival, and cystectomy-free survival, as well as across all disease states, including BCG-naïve or intolerant disease, BCG-unresponsive CIS, and BCG-unresponsive papillary-only tumors. Thus, these findings challenge prevailing assumptions about the additive benefit of docetaxel and emphasize the need for rigorous comparative effectiveness research in the treatment of BCG-unresponsive NMIBC.
In addition to selection bias, several other factors may explain the lack of observed benefit with sequential Gem/Doce in our study. Maintenance therapy, potentially critical to achieving durable responses, was administered to only 40% of patients in the Gem/Doce group and to just 19% of those receiving only gemcitabine monotherapy. By contrast, patients receiving Gem/Doce in the Steinberg et al11 study had monthly maintenance for up to 24 months; in the Yim et al13 study, two-thirds of patients received maintenance; and in the Chevuru et al14 study, 96% of responders received maintenance. Further, granularity regarding maintenance schedule subtype, duration, and agent-specific dwell time completion could not be consistently determined because intravesical therapy administration details were captured retrospectively. The underutilization of maintenance therapy in both of our groups, particularly in the gemcitabine cohort, likely attenuated any differences in durability between regimens and may partly explain the overall lower HGRFS observed for both groups.
Shorter follow-up duration in the sequential Gem/Doce group may also have contributed to underestimating long-term benefit. Among patients without high-grade recurrence, the median follow-up was 21.0 months in the sequential Gem/Doce group and 45.8 months in the gemcitabine group. This disparity reflects the more recent adoption of combination therapy in clinical practice, but it also limits the precision of survival estimates at longer time points. More favorable outcomes with Gem/Doce were reported from a single-institution observational study by Chevuru et al14 that had a longer follow-up duration of 49 months. Although time-to-event analyses appropriately account for censoring, the reduced number of patients at risk beyond 24 months diminishes statistical power and can bias later survival estimates toward the null, particularly in the absence of protocol-based maintenance therapy.
Differences in gemcitabine dosing may also have influenced outcomes. Patients treated with gemcitabine monotherapy received 2 g per instillation, consistent with the dosing strategy used in SWOG S0353, whereas those treated with sequential Gem/Doce received 1 g of gemcitabine followed by 37.5 mg of docetaxel.10, 11 Although both regimens reflect commonly accepted practice, the lower gemcitabine dose in the combination group may have resulted in reduced drug exposure and, consequently, diminished antitumor activity. To date, no comparative pharmacodynamic studies have clarified whether the effectiveness of combination therapy compensates for the reduced gemcitabine dose. Thus, the relative contribution of gemcitabine intensity vs the additive effect of docetaxel remains an open question.
It is notable that outcomes observed in our cohort for gemcitabine monotherapy compare favorably to those reported in the phase II SWOG S0353 trial. In that study, 1-year and 2-year RFS rates were 28% and 21%, respectively.10 By comparison, our study showed RFS rates of 51% at 1 year and 37% at 2 years among patients treated with gemcitabine alone. This discrepancy likely reflects differences in patient selection and institutional practices, although advances in surgical techniques, particularly more widespread use of blue light cystoscopy, may also have contributed to improved outcomes. These observations introduce the possibility that single-agent gemcitabine, particularly when integrated into contemporary surgical and surveillance practices, may offer greater effectiveness than previously thought, warranting renewed consideration for BCG-unresponsive patients.
This study has several limitations in addition to those previously described. As a retrospective analysis, it was subject to documentation variability and potential misclassification of both exposures and outcomes. AEs and treatment intolerance were identified through routine clinical documentation rather than with standardized instruments, which may have led to underestimating toxicity and its effect on treatment adherence. Additionally, the study included patients treated across 3 academic centers with differing practice patterns and underlying patient populations. Although these factors reflect real-world heterogeneity, they may also contribute to variability in outcomes. Neither treatment group received pretreatment with urinary alkalinization (eg, bicarbonate), which has been proposed to enhance gemcitabine activity. However, because gemcitabine was administered to both groups, the absence of alkalinization is unlikely to have differentially influenced comparative outcomes, although its overall effect on treatment efficacy cannot be assessed from this study. In addition, calendar year was used as an instrumental variable, so the exclusion restriction cannot be definitively established. However, the relatively short study period (2018–2023) was selected to minimize secular changes, and there were no major systematic shifts in surgical technique, surveillance, or pathology practices across sites. Blue light cystoscopy was available throughout the study period without a discrete adoption phase, and pembrolizumab was infrequently used, making it unlikely that temporal factors meaningfully influenced recurrence risk.
Despite these limitations, we believe our findings offer a meaningful contribution to the literature. This study provides a direct comparison between 2 widely used intravesical salvage therapy regimens, drawn from a multicenter cohort defined by contemporary criteria for BCG unresponsiveness and adjusted rigorously for treatment selection bias. Although sequential Gem/Doce is often regarded as the preferred regimen in academic practice, recent survey data indicate substantial variation in real-world treatment patterns, with approximately 30% of respondents continuing to favor single-agent gemcitabine for BCG-unresponsive patients.17 This variability likely reflects the operational complexity of administering combination therapy, particularly in community settings where providers frequently encounter challenges related to infusion chair time, staffing, and drug reimbursement.7 In this context, our results offer a timely and pragmatic benchmark for clinicians navigating treatment decisions in the absence of prospective data. Until such trials are available, the academic community should be cautious in overstating the superiority of combination therapy and remain attentive to the practical constraints that shape intravesical chemotherapy delivery in the real world.
5. Conclusions
In this multicenter retrospective study of patients with high-risk NMBIC, sequential intravesical Gem/Doce was not associated with improved HGRFS compared with gemcitabine alone. This finding held across analytic approaches and disease states, including BCG-unresponsive CIS. Given its complexity and lack of demonstrated benefit, sequential therapy warrants evaluation in a prospective randomized trial.
Supplementary Material
Highlights.
Gem/Doce is increasingly used for NMIBC, especially after BCG failure or shortage.
A real-world direct comparison with gemcitabine alone was not previously available.
Median HGRFS was 22.6 months for gemcitabine vs 19.5 months for Gem/Doce (P=.25).
Gem/Doce was not associated with improved HGRFS on instrumental variable analysis.
No HGRFS difference was observed in the BCG-unresponsive subgroup.
Acknowledgments:
Marianne Mallia, ELS, MWC, Mayo Clinic, substantively edited the manuscript. The Scientific Publications staff, Mayo Clinic, provided proofreading and administrative and clerical support.
Funding:
This study was supported by grant number P30 CA015083 from the National Cancer Institute (NCI) and from the Mayo Clinic Robert D. and Patricia E. Kern Center for the Science of Health Care Delivery (to M.D.T.), the Christian Haub Family Career Development Award in Cancer Research Honoring Dr Richard Emslander (to M.D.T.), Eric and Gail Blodgett (to M.D.T.), and the Endras Family (to M.D.T.).
Abbreviations
- AE
adverse event
- BCG
bacille Calmette-Guérin
- CIS
carcinoma in situ
- FDA
US Food and Drug Administration
- Gem/Doce
gemcitabine and docetaxel
- HGRFS
high-grade recurrence-free survival
- HR
hazard ratio
- IQR
interquartile range
- NMIBC
non–muscle-invasive bladder cancer
- RFS
recurrence-free survival
- SWOG
Southwest Oncology Group
- TURBT
transurethral resection of bladder tumor
Footnotes
CRediT authorship contribution statement: Jacob A. Moyer: Methodology, Formal analysis, Data curation, Writing – original draft, Writing – review & editing, visualization. Mahnoor Ashraf: Data curation, Validation, Writing – review & editing. Lanyu Mi: Methodology, Validation, Formal analysis, Writing – original draft, Writing – review & editing. Andrew J. Zganjar: Writing – review & editing, Supervision. Timothy D. Lyon: Writing – review & editing, Supervision. Paras H. Shah: Writing – review & editing, Supervision Stephen A. Boorjian: Writing – review & editing, Supervision. Mark D. Tyson II: Conceptualization, Methodology, Writing – original draft, Writing – review & editing, Supervision, Project administration, Funding acquisition.
Conflict of interest: None.
Previous presentation: Presented as a poster/podium presentation at the Western Section AUA 101st Annual Meeting; November 2, 2025; Napa Valley, California.
Data access and responsibility:
The datasets generated and/or analyzed during this study are available from Mayo Clinic, but restrictions are mandated per institutional procedures. Data may be available from the authors on reasonable request and with documented permission from Mayo Clinic.
References
- 1.Zhang LY, Wang P, Wang YB, He ZQ. Global, regional, and national burden of bladder cancer, 1990–2019: an age-period-cohort analysis based on the Global Burden of Disease 2019 study. Public Health. Nov 2024;236:193–203. doi: 10.1016/j.puhe.2024.07.027 [DOI] [PubMed] [Google Scholar]
- 2.Holzbeierlein JM, Bixler BR, Buckley DI, et al. Diagnosis and Treatment of Non-Muscle Invasive Bladder Cancer: AUA/SUO Guideline: 2024 Amendment. J Urol. Apr 2024;211(4):533–538. doi: 10.1097/JU.0000000000003846 [DOI] [PubMed] [Google Scholar]
- 3.Ritch CR, Velasquez MC, Kwon D, et al. Use and Validation of the AUA/SUO Risk Grouping for Nonmuscle Invasive Bladder Cancer in a Contemporary Cohort. J Urol. Mar 2020;203(3):505–511. doi: 10.1097/JU.0000000000000593 [DOI] [PubMed] [Google Scholar]
- 4.Daniels MJ, Barry E, Schoenberg M, et al. Contemporary oncologic outcomes of second induction course BCG in patients with nonmuscle invasive bladder cancer. Urol Oncol. Jan 2020;38(1):5 e9–5 e16. doi: 10.1016/j.urolonc.2019.05.018 [DOI] [PubMed] [Google Scholar]
- 5.Chun B, He M, Jones C, et al. Variation in Statewide Intravesical Treatment Rates for Non-Muscle Invasive Bladder Cancer During the Bacillus Calmette-Guerin Drug Shortage. Urology. Jul 2023;177:74–80. doi: 10.1016/j.urology.2023.02.044 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Abou Chakra M, Packiam VT, Duquesne I, Peyromaure M, McElree IM, O'Donnell MA. Combination intravesical chemotherapy for non-muscle invasive bladder cancer (NMIBC) as first-line or rescue therapy: where do we stand now? Expert Opin Pharmacother. Feb 2024;25(2):203–214. doi: 10.1080/14656566.2024.2310073 [DOI] [PubMed] [Google Scholar]
- 7.Packiam VT, Gabrielson AT, Kates M. Intravesical Gemcitabine and Docetaxel for Nonmuscle-Invasive Bladder Cancer. American Urological Association; 2022. https://auanews.net/issues/articles/2022/june-2022/intravesical-gemcitabine-and-docetaxel-for-nonmuscle-invasive-bladder-cancer [Google Scholar]
- 8.BCG-Unresponsive Nonmuscle Invasive Bladder Cancer: Developing Drug and Biological Products for Treatment (U.S. Department of Health and Human Services; ) (2024). [Google Scholar]
- 9.Roumiguie M, Kamat AM, Bivalacqua TJ, et al. International Bladder Cancer Group Consensus Statement on Clinical Trial Design for Patients with Bacillus Calmette-Guerin-exposed High-risk Non-muscle-invasive Bladder Cancer. Eur Urol. Jul 2022;82(1):34–46. doi: 10.1016/j.eururo.2021.12.005 [DOI] [PubMed] [Google Scholar]
- 10.Skinner EC, Goldman B, Sakr WA, et al. SWOG S0353: Phase II trial of intravesical gemcitabine in patients with nonmuscle invasive bladder cancer and recurrence after 2 prior courses of intravesical bacillus Calmette-Guerin. J Urol. Oct 2013;190(4):1200–4. doi: 10.1016/j.juro.2013.04.031 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Steinberg RL, Thomas LJ, Brooks N, et al. Multi-Institution Evaluation of Sequential Gemcitabine and Docetaxel as Rescue Therapy for Nonmuscle Invasive Bladder Cancer. J Urol. May 2020;203(5):902–909. doi: 10.1097/JU.0000000000000688 [DOI] [PubMed] [Google Scholar]
- 12.Common Terminology Criteria for Adverse Events (CTCAE). U.S. Department of Health and Human Services. https://dctd.cancer.gov/research/ctep-trials/for-sites/adverse-events/ctcae-v5-8x11.pdf [Google Scholar]
- 13.Yim K, Melnick K, Mott SL, et al. Sequential intravesical gemcitabine/docetaxel provides a durable remission in recurrent high-risk NMIBC following BCG therapy. Urol Oncol. Nov 2023;41(11):458 e1–458 e7. doi: 10.1016/j.urolonc.2023.06.018 [DOI] [PubMed] [Google Scholar]
- 14.Chevuru PT, McElree IM, Mott SL, Steinberg RL, O'Donnell MA, Packiam VT. Long-term follow-up of sequential intravesical gemcitabine and docetaxel salvage therapy for non-muscle invasive bladder cancer. Urol Oncol. Mar 2023;41(3):148 e1–148 e7. doi: 10.1016/j.urolonc.2022.10.030 [DOI] [PubMed] [Google Scholar]
- 15.Taylor J, Kamat AM, Annapureddy D, et al. Oncologic Outcomes of Sequential Intravesical Gemcitabine and Docetaxel Compared with Bacillus Calmette-Guérin in Patients with Bacillus Calmette-Guérin-Unresponsive Non-Muscle Invasive Bladder Cancer. Eur Urol Oncol. Apr 2025;8(2):469–476. doi: 10.1016/j.euo.2024.12.005 [DOI] [PubMed] [Google Scholar]
- 16.Scilipoti P, Zaurito P, Longoni M, et al. Impact of BCG failure patterns on efficacy of intravesical gemcitabine/docetaxel in high-risk non-muscle-invasive bladder cancer. BJU Int. Feb 25 2026;doi: 10.1111/bju.70182 [DOI] [PubMed] [Google Scholar]
- 17.Saylor BP. Survey: New NMIBC Treatments Face Slow Uptake. Urology Times. 2025;53(6):18–22. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets generated and/or analyzed during this study are available from Mayo Clinic, but restrictions are mandated per institutional procedures. Data may be available from the authors on reasonable request and with documented permission from Mayo Clinic.