Predictors of the regional variation of prostatectomy or radiotherapy: evidence from German cancer registries

Curative treatment of prostate cancer remains controversial building on the alternative options of prostatectomy or radiotherapy. Randomized trials found a comparable efficacy of treatments in terms of overall survival, while patterns of adverse events differ between both options (Hamdy et al. 2016). Respecting these findings, the preference of the informed patient should guide the choice of treatment.

For certain patient cohorts, ‘active surveillance’ and ‘watchful waiting’ are further options to target prostate cancer. The former is restricted to patients with a very low risk and entails an active patient monitoring. In contrast, ‘watchful waiting’ is a passive approach, which abstains from either an interventional treatment or ‘active surveillance’, but would only consider treatment when the disease becomes symptomatic. Patients with a restricted life expectancy are candidates for this approach.

From a public health perspective, treatment environments differ in terms of the availability and established practice in the administered treatment. Externalities ranging from the neighborhood economic status to the availability of treatment units might affect preferences of physicians and patients as it was found in a recent population-based study (Tomic et al. 2018). Such effects might ultimately lead to substantial differences in terms of survival between regions within a country (Skyrud et al. 2016). While external factors explain survival differences, it remains unclear how treatment preferences differ in the first place independently from such externalities. This is especially relevant in entities where there are treatment options with comparable outcomes as it is the case for prostate cancer (see above).

German population-based, that is epidemiological, cancer registries cover almost all incident cancer cases among residents of their respective jurisdictions. They offer a unique possibility to compare regions in terms of the administered treatment on a population-based level (Schubert-Fritschle et al. 2017). It is the objective of this study to identify treatment preferences in a population-based sample.

In summary, we found that no active treatment, applying neither radiotherapy nor surgery, is more favored in East than in West Germany. Likewise, patients living in districts with both, a radiotherapy and urology treatment unit, or with a certified cancer center are more likely to receive no active treatment. Furthermore, patients are especially likely to receive prostatectomy compared to radiotherapy when they live in a district where only a radiotherapy institution is present or in East Germany.

To our knowledge, this is the first study that examines treatment preference in East and West Germany based on a population-based sample focusing on prostate cancer. A population-based approach entails the advantage of a nationwide picture that is not limited by data that might themselves affect treatment choices such as data from health care insurances or health care providers.

Coming to the clinical consequences of our findings, the identification of factors that lead to treatment clusters becomes a key objective of health care research. In a randomized trial (randomized between usual care and additional information from a decision aid), there was evidence that the visited hospital was a predictor for treatment, but not age and tumor characteristics (van Tol-Geerdink et al. 2013). These findings are well in line with our results where hospitals as they are part of a larger regional framework of East and West Germany were associated with treatment choice.

Treatment preferences are a complex interplay between individual patient characteristics and external factors such as the socioeconomic characteristics across areas. On a patient level, further studies found that factors such as risk attitudes (López-Pérez et al. 2017), age, or tumor stage were related to the preferred treatment (Sommers et al. 2008). Another study in a collective of 11,892 men with localized prostate cancer found a strong inter-institutional treatment variation in treatment preference ranging, e.g., for external radiotherapy, from 2 to 33% (Cooperberg et al. 2010). This differences could not be explained by case-mix variability or known patient factors, while institutional practice sites accounted for these variations to a varying degree (13% for androgen deprivation monotherapy; 74% for cryoablation) (Cooperberg et al. 2010).

An analysis based on SEER (Surveillance, Epidemiology, and End Results) data addressed treatment patterns in patients above 66 years diagnosed with T3 and T4 prostate cancer (Lowrance et al. 2012). Here, the authors reported an increase in multimodal therapy with radiotherapy and androgen deprivation therapy (ADT) between 1998 and 2005. Only a minority of 15% of the patients received no active treatment, while 36% received ADT only. As ADT is dependent on the application of radiotherapy, we abstained from accounting for ADT in statistical models in order to avoid bias from strongly correlated variables and collider bias (bias resulting from conditioning on variables that are dependent on two parent variables, e.g., radiotherapy and cancer stage for ADT in our case).

Returning to treatment outcomes in Germany, we found previously that people living in East Germany have a comparable mortality from prostate cancer as their West German counter-parts (Medenwald et al. 2017). Mortality rates in East and West Germany aligned only during the last 2 decades after the year 2000. Thus, the difference in treatment pattern seems not to result in a measurable difference in mortality. However, because of the favorable survival prospect of patients suffering from prostate cancer in an early stage screening programs might affect mortality from prostate cancer most strongly (Schröder et al. 2014).

Germany is characterized by a high rate of patients treated with surgery when compared to. e.g., USA (Hager et al. 2015). Hager et al. computed a rate of prostatectomy of 36.1% in the USA and 66.2% in Germany, while the rates of radiotherapy were 38.4% and 11.8%, respectively. Again data from the US-American and German registries showed an increased utilization of prostatectomy in locally advanced cases of prostate cancer (Hager et al. 2017). However, in both studies, the authors did not differentiate between East and West Germany and did not account for differences in the age composition and in economic factors. In our data, shifting only slightly from the preference for prostatectomy, radiotherapy was more likely when there was a radiotherapy unit but no urologic institution.

As our results emanate from multivariate adjusted models, causes leading to the mentioned treatment choices descend from independent origins. The finding that an active treatment is less likely in districts that are equipped with both, a radiotherapy and urology institution, or a certified cancer centers might reflect a better infrastructure for ‘active surveillance’ or ‘watchful waiting’. Still, other causes might underlie the higher likelihood for no active treatment in East Germany. As ‘active surveillance’ is a feasible treatment in a selective group of patients with an early stage, regions with a high density of treatment units might offer a better screening leading to a higher frequency of early cases suitable for ‘active surveillance’. Although, such an effect is unlikely as we adjusted for stage, there are parameters in the definition of early suitable cases, such as the degree of punch biopsy, which our data fail to record. This reasoning needs further investigation by prospective studies that, however, have the disadvantage of not covering the entire population.

In summary, we found that the administered treatment of prostate cancer varies systematically between East and West Germany according to the availability of radiotherapy institutions.