Lung cancer is one of the most frequent cancers worldwide. According to the Robert Koch Institute (RKI) data on NSCLC in Germany, approximately 35,000 male and 20,300 female patients are newly diagnosed with lung cancer per year, with a low five year survival rate of 21% for female and 16% for male patients in 2012 [20
Approximately 40–50% of patients with NSCLC are tested in stage IV, while NSCLC includes 85% of all newly diagnosed lung cancers [21
]. Taking into account the data from the RKI and the prognosis for 2016, 85% of 55,300 patients, i.e., 47,005 patients, are likely to be diagnosed with NSCLC each year. Of those patients, 18,802–23,503 will be diagnosed in stage IV (i.e., 40–50%). Approximately half to one third of those patients will be eligible for ALK testing.
As shown recently, ALK testing methods require significant improvement: von Laffert and colleagues have shown that, in up to 10 percent of clinical cases, ALK diagnostics were incorrect in up to 40% of pathology departments, although in most cases FDA-approved methods were used [22
]. This diagnostic challenge is complicated by the fact that there is an increasing number of identified ALK variants in NSCLC for which the prevalence, the treatment response, and clinical significance remain unknown [1
]. Whilst the overall positivity rate of mutated ALK in NSCLC is published as being between 2 and 7 percent, the response rate to the ALK inhibitor Crizotinib was described as 57%, with a remarkably high rate of progression-free survival of 72% after 6 months of ALK-inhibitor therapy [17
]. Thereby, the response rate ranges between full remission and total lack of response, a phenomenon that can most likely be attributed to the high number of different variants. Unfortunately, this topic has not yet been addressed, and only the median progression-free survival during ALK-inhibitor therapy has been published, with an increased rate of 7.7 vs. 3 months, i.e., a median improvement of 4 months [25
]. Therefore, it was our aim to analyze to what extent these internationally accepted data also apply to the situation in Germany in general and to the clinical cohort in our hospital in particular.
Our cohort consisted of patients in Cologne, and the region within approximately 100 km of Cologne (Table 1
). The patients coming to our hospital were either transferred to our hospital because of a pre-diagnosed NSCLC or because of clinical symptoms of lung cancers. Therefore, no limitations existed with respect to age, sex, or ethnic background; thus, the cohort is not a “study-cohort”, but reflects daily hospital practice. In our cohort in 2014, 645 patients were diagnosed with lung cancer, of which 574 suffered from NSCLC. The total number of NSCLC during the entire observation period was 1722 registered patients with NSCLC. Of those, 860 patients were in NSCLC stage IV. A total number of 60 patients out of the 860 NSCLC stage IV patients tested positive for ALK mutation, and thus in principle were eligible for Crizotinib therapy. This latter number corresponds to a percentage of 3.8% ALK-positive NSCLC in the entire cohort, thus being in a range comparable to other cohorts [26
]. Out of the 60 patients who were, in principle, eligible for Crizotinib therapy, only 22 received Crizotinib therapy, whilst the remaining patients received alternative or palliative therapy.
Out of the 22 patients receiving Crizotinib therapy, 14 were male (63.6%) and 8 were female (36.4%). The mean age was 62.68 years, ranging between 48 and 79 years, with a median age of 62.5 years. Seven patients were smokers, 11 patients were former smokers, 3 patients had never been smokers, and for one patient the smoking status remained unknown. The therapy duration had a mean of 5.36 months, whilst the median therapy duration was 2 months, and thus significantly lower than in other published studies. The majority of patients had therapy durations of max. 4 months (16 out of 22, i.e., 72.7%); only 13.6% were still receiving therapy at the cut-off date of this study. Serious differences were also observed in the clinical courses during treatment. Three patients died within the month when the Crizotinib therapy was initiated, while another patient died 28 months after the therapy start. In one case (patient 1), the therapy had to be stopped due to side effects, and in one case the daily dose was reduced after two weeks. Six patients received an additional therapy (chemotherapy or radiation therapy). Only one patient had a stable remission, and continued to receive Crizotinib therapy even 25 months after therapy onset.
In the case of the cohort described above, the cost for the ALK-inhibitor therapy was 118 months × 6087.77 = 718,356.86 € (Table 2
). No further differences in therapy costs could be identified between ALK-positive and ALK-negative patients. It was surprising that in our patient cohort the median and the average therapy duration were below those of published studies. Thereby, it has to be mentioned that our cohort was not a “controlled” study cohort but an observational study, and reflected the “real life” daily patients coming to a German hospital specialized in the therapy of lung diseases.
The costs for ALK testing with immunohistochemistry (IHC) or fluorescence in situ hybridization range between 80 and 90 € per sample. This calculation is based on the list prices for ALK FISH assays from the German branches of vendors of ALK FISH assays (Abbott, Wiesbaden, Germany; ZytoVision, Bremerhafen, Germany; KreaTech, Amsterdam, The Netherlands). The cost for the novel assay was calculated by analogy to the costs of Respifinder and Meningofinder assays (multiplex assays making use of the same technology proposed here) by our co-author Guus Simons (CEO of Pathofinder, Maastricht, Germany). We calculated that a novel assay that can distinguish between all known ALK fusion variants, including other variants that determine eligibility for ALK-inhibitor therapies like ROS fusions, would also cost about 80 € per sample. Such an assay could be based on MLPA and preamplification as already used for other purposes by PathoFinder, and could include a software tool that enables therapy outcome predictions based on cell culture and correlated therapy data. Therefore, the initial development costs for such an assay would reach a maximum of 6 million € as a single investment (our exact calculation, based on a Horizon2020 grant proposal submitted to the European Commission, was 5,183,566 €). The proposed assay should undergo clinical evaluation; that is, the clinical outcome of therapy in relation to ALK fusion variants needs to be included in the interpretation software tool. This goal can be easily achieved by combining cell culture data with cell culture phenotyping assays, making use of the different variants and clinical observations. The in vitro phenotyping would either be based on cell lines with known ALK mutations, or these mutations would have to be included in the cells by recombination technology like CRISPR/Cas or Cre/loxP assays. The predictions should be based on the combinations of in vitro and in vivo data. If a variant is not treatable in vitro, it won’t be treatable in vivo. Thereby, dose effects and toxicity profiles have to be taken into account. The prediction software should be constructed by analogy with the Geno2pheno software used for HIV therapy outcome prediction, and make use of machine learning approaches.
Based on these latter specifications, we then assumed that such an assay could have predicted the outcome of Crizotinib therapy in our patient cohort, i.e., the 6 out of 22 patients (27.3%) treated could have been predicted as responders, whilst the remaining patients could have been identified as non-responders (72.7%); thus, the high-cost Crizotinib therapy could have been avoided in 72.7% of cases, resulting in a remarkable cost-saving potential (Table 2
The frequency of ALK translocations in NSCLC was published as 1–2% [27
]; i.e., based on the above-mentioned epidemiological data of approximately 47,005 newly diagnosed NSCLC patients [28
] per year, a minimum of between 470 and 940 patients per year would be eligible for ALK testing in Germany.
Taking international studies into account, the frequency of ALK translocations would be 4–6%, resulting in a maximum of 1880–2820 patients per year being eligible for ALK-inhibitor therapy.
Excluding 2 patients of our cohort of 22 ALK-positive patients, where in one case Crizotinib therapy was stopped and in the other, therapy was initiated at the cut-off point of this study, the remaining 20 ALK-positive patients were used for the subsequent calculations. Of these 20 ALK-positive patients, we calculated a number of 7.33 ALK-positive patients under Crizotinib therapy; i.e., our cohort reflects 20/940 = 2.13% to 20/470 = 4.26% of German patients with ALK therapy, i.e., (470/20) × 7.33 = 172 to (940/20) × 7.33 = 355 patients per year in Germany under Crizotinib therapy in Germany (data from Table 2
Based on these calculations, the average yearly therapy duration is 172 to 355 patients × 5.36 months = 921.92 to 1902.8 months, with a median yearly therapy duration of 172 to 355 × 2 months = 344 to 710 months.
Based on a cost of 6087.77 € per month for Crizotinib therapy [29
], the average costs for Crizotinib therapy in Germany could be estimated by 921.92 to 1902.8 months × 6087.77 € = 5,612,436.92 to 11,583,808.76 €, with median therapy costs of 344 to 710 months × 6087.77 € = 2,094,192.88 to 4,322,316.70 €. As the international published data for the prevalence of mutated ALK are 4–6% (see above), these costs have to be multiplied by a factor of 2, resulting in a maximum average cost for Crizotinib therapy of 23,167,617.52 €, and a median cost of 8,644,633.40 €, respectively [30
Using the data from Table 1
and Table 2
as a basis, the cost-saving potential for our patient cohort would have been 31 months × 6087.77 € = 188,720.87 €, if the non-responder rate of 72.73% had been predicted in advance of therapy. Extrapolated to the estimation for the entire German cohort, the cost-saving potential would be 4,081,772.31 €–8,424,588.15 € if the average therapy duration is used as the basis for the calculation, and 1,523,049.37 €–3,143,503.05 € if the median therapy duration is used as the basis for the calculation.
Taking into account the overall population of the European Union, which is estimated to be 510.1 million people [31
], the putative cost-saving potential would be 11,006,270.64 €–22,012,541.27 € per year based on the average therapy duration of non-responders, and 11,361,311.57 €–22,722,623.14 € based on the median therapy duration of non-responders, per year (Table 3