The Cancer Registry of Norway (CRN) captures nearly 99% of new cancer diagnoses in Norway . The collected data includes patient demographics, tumour characteristics, treatment codes (surgical, radiotherapy) and causes of death. For bladder cancer, histopathology of specimens from transurethral resection of bladder tumour (TURB), cystectomy and biopsies of metastases are registered, along with the corresponding dates for the procedures.
The Norwegian Patient Registry contains individual administrative, demographic, and coded medical information (diagnoses, procedures, chemotherapy) from all patients’ contacts with public hospitals. This data was linked to the CRN by the personal identification number assigned to all residents of Norway .
We included patients undergoing RC (2008–2015) with or without NAC who were diagnosed with MIBC (urothelial carcinoma) without known distant metastases between 2008 and 2012. The pre-RC status of regional lymph node metastases was unknown (cNx). Patients with a pre-RC histology verifying muscle-invasion and patients without such verification but treated with NAC were considered as having MIBC. We chose this period since we had quality ensured histopathological information from this period and to ensure sufficient follow-up time for survival analysis. Patients with pre-RC radiotherapy were excluded.
For the evaluable patients, the research team reviewed all available clinical notifications and histology reports at the CRN. The presence of MIBC was confirmed in the histology reports from TURB specimens. Muscle-invasion was defined as tumour invasion into the muscularis propria (≥T2). From the histology reports from cystectomy specimens, the histopathological T and N category (pT; pN)  without sub-classification into a and b for pT2-pT4 were confirmed. All MIBC are high-grade .
We identified relevant specified intravenous chemotherapy codes (e.g., gemcitabine and cisplatin, methotrexate, vinblastine, doxorubicin and cisplatin (MVAC), carboplatin) and codes for intravenous administration of non-specified chemotherapy from the Norwegian Patient Registry. We excluded chemotherapy events concurrently registered with ICD-10 codes for a different cancer than C67. NAC was defined as any chemotherapy administered intravenously between diagnosis of bladder cancer and RC.
Based on the available data and definitions used in similar studies [12, 14, 19], we defined downstaging of the primary tumour (DS) as pT0/pTa/pTis/pT1 with the subunit of pT0 as complete response (CR), identified independent of the use of NAC. Patients without DS (non-DS) were characterized by having residual muscle-invasive disease (pT2-pT4) in the specimen. Downstaging can occur after TURB and NAC. Nodal downstaging could not be assessed because information about cN was not available.
The observation time started at the date of RC until death, emigration, or end of study (December 31st, 2019), whichever came first. Time in years from date of RC was used as timescale in all analyses.
We applied descriptive statistics (mean, median, interquartile range (IQR), proportions) to present pre- and post-operative characteristics in all patients as well as in the NAC and NoNAC group. The association between NAC and DS was estimated using logistic regression adjusted for all available pre-operative variables: age at diagnosis (≤59, 60–69, 70–79, ≥80), sex, type of hospital (academic vs. community), geographical health region (Southeast, West, Central, North) and the year of RC (2008–2009, 2010–2011, 2012–2015). OS was presented by Kaplan-Meier curves and the difference between them was evaluated with the log-rank test. The associations between DS with OS, as well as NAC with OS (total effect) were assessed with a Cox regression model adjusted for all available pre-operative variables.
The association between NAC and OS was additionally investigated by applying a mediation approach adjusted for available pre-operative variables. We applied a causal inference approach [20, 21] implemented in the R mediation package . The total effect of NAC on OS (unadjusted for DS) evaluated with a Cox regression model was decomposed into two parts : the indirect effect between NAC and OS mediated by DS, and the direct effect between NAC and OS (not through DS) This approach allowed us to assess the indirect effect of NAC on OS through DS.
In order to overcome the confounding by indication bias induced by missing information of factors leading to the decision of treatment, we applied an instrumental variable approach to estimate the causal effect of NAC on OS . We used type of hospital as the instrumental variable and G-estimation [24,25,26] with adjustment for the remaining pre-operative variables.
Quantities reported from the model-based analyses are odds ratios (ORs) and hazard ratios (HRs) including 95% confidence intervals (CI) and p-values. The statistical significance level was set at 0.05. Statistical analyses were performed using Stata 17 (StataCorp, College Station, TX) and R (version 4.1.4).