Contra-intuitive findings in retrospective epidemiological studies should always make the reader alert. Moreover, if there are important well-known prognostic risk factors that cannot be corrected for, even though they are imperative for selection of the intervention under evaluation, all red flags should be raised.
The majority of patients with high risk non-muscle invasive bladder cancer (NMIBC) can be safely treated by conservative means with transurethral resection and adjuvant BCG [Citation1]. There are, however, several very well-known risk factors that make radical cystectomy the more obvious and oncologically safe choice. The reason for recommending this more radical treatment is an inherent risk of understaging or rapid progression to muscle-invasive bladder cancer (MIBC) if conservative treatment is chosen in selected patients with the highest risk. These risk factors include tumour size, variant histology, multiplicity, lymphovascular invasion, concomitant CIS etc. [Citation1]. Moreover, in these highest risk NMIBC patients, upstaging in almost 25% to MIBC has been described in the cystectomy specimens, thus, underlining the need for radical treatment up front [Citation2]. It should be rather clear that these patients with the highest risk have a higher risk for progression than other ‘high risk’ patients if treated by the same method – hence the name ‘highest risk patients’.
In the present Swedish nationwide analysis, the authors finds that prognosis of Swedish NMIBC patients is worse in patients undergoing cystectomy compared to patients undergoing BCG instillations only [Citation3]. However, if Swedish urologists in this period have been offering BCG to the conventional high risk patients and cystectomy to patients with the highest risk – as recommended in guidelines – this difference is a natural consequence of these risk factors per se. However, most of these known risk factors cannot be adjusted for in this material according to the authors. Therefore, this publication has unfortunately confounding by indication written all over it.
Basically, there is one simple way to adjust for these risk factors in the present material: treatment choice of cystectomy instead of BCG is a very excellent surrogate marker of these higher risk factors. However, it is naturally not possible to use the treatment modality both as marker of the important currently uncorrected risk factors and at the same time use it as exposure variable. Therefore, the authors should instead consider not to publish these data with the known very important uncorrected biases in the current format. The risk of publications like these is that some less academic urologists might take the message of this article literally and start to use BCG in patients with more risk factors than the good prognosis patients undergoing BCG in this historical cohort.
Basically, the rather small difference in prognosis of the two treatment groups of this cohort study is a very good marker that aggressive treatment (i.e. cystectomy) in patients with highest risk is not only indicated but also effective. Otherwise, the difference would have been higher.
In theory, the authors could have evaluated the prognosis of Swedish patients with MIBC undergoing either cystectomy alone or systemic chemotherapy alone and found a huge survival difference in favour of cystectomy, especially if not correcting for presence of metastases. But concluding that patients with metastatic MIBC therefore potentially could benefit from cystectomy would not be correct either. Patients are most often treated according to their specific stage and not just on coincidental selection by the urologist.
The study elegantly shows that Swedish urologists have been very good at selecting the right patients for either cystectomy or BCG. Other attempts for conclusions are hampered by the lack of details regarding risk factors and selection bias in the current study design.
Causal inference as supplement to correlation alone is an important statistical method described by e.g. Judea Pearl with emphasis on the probabilities of counterfactuals [Citation4]. Interested readers should be encouraged to revisit this subject.
Department of Urology, Aarhus University Hospital, Aarhus, Denmark
Department of Clinical Medicine, Aarhus University, Aarhus, Denmark
[email protected]
Disclosure statement
The author has no financial disclosures to this publication
References
- Babjuk M, Burger M, Compérat EM, et.al. European Association of Urology Guidelines on non-muscle-invasive bladder cancer (TaT1 and carcinoma in situ) – 2019 update. Eur Urol. 2019;76(5):639–657.
- Dalbagni G, Vora K, Kaag M, et al. Clinical outcome in a contemporary series of restaged patients with clinical T1 bladder cancer. Eur Urol. 2009;56(6):903–910.
- Wang EYH, Larsson U, Gårdmark T, et al. Radical cystectomy compared to intravesical BCG immunotherapy for high-risk non-muscle invasive bladder cancer – is there a long-term survival difference? – A Swedish nationwide analysis. Scand J Urol. 2020.
- Pearl J. An introduction to causal inference. Int J Biostat. 2010;6:7.