1        Introduction

Cancer incidence is increasing globally¹. Among all cancers diagnosed in women, breast cancer (BC) is the most frequent, representing the leading cause of death. In 2018, there were over 2 million new BC cases worldwide². World Health Organization (WHO) estimates that almost 627,000 women died in 2018 because of BC, which constitutes approximately 15% of all cancer deaths amongst women³.

BC is also responsible for a large part of the cost of oncological treatment. For example, in the United States in 2014, approximately $18.1 billion can be attributed to female BC, out of $137.4 billion of the national expenditures for cancer care (i.e. 13%)⁴. The total costs of cancer not only vary by tumour type but also depend on the stage of disease: treatment of the advanced stages of cancer is often more intensive or invasive, most costly, and less successful^5,6. Sun⁶ showed that the mean treatment costs of stage II, III and IV (at diagnosis) exceeded those of stage I by 32%, 95%, and 109%, respectively. As can be seen, the increase is substantial for stages III (locally advanced breast cancer) and stage IV (metastatic breast cancer), referred to as an advanced breast cancer (ABC), i.e. BC that has spread to another part of the body³. These stages are also most costly among all stages due to intensive treatment and associated indirect costs.

BC affects relatively young patients. WHO estimates that approximately 30% of new cases of BC in 2018 affected women under 50 years². For example, in females aged 25-49 in the United Kingdom, BC is the most common cancer, accounting for more than 4 out of 10 (44%) of all cancer cases⁷. Therefore, working population may be largely affected and BC may generate substantial indirect cost (i.e. opportunity cost of foregone product)⁸. Even though indirect costs are not associated with actual cash flows, they measure the disruption to the economy caused by the illness and are considered as an important element from the societal perspective. Estimating the magnitude of ABC costs can help to determine its economic significance (not undermining the clinical importance). Understanding the components of these costs can help to optimize healthcare spending, e.g., by informing the cost-effectiveness analyses of a treatment or diagnostic technologies or deciding on investment in a particular health care setting (primary or hospital care) or type of care (prevention, curation, palliation).

We aimed to assess the direct medical costs and the indirect costs of ABC in selected European countries: Bulgaria, Croatia, Czech Republic, Estonia, Greece, Israel, Latvia, Poland, Romania, and Slovak Republic. The international scope of the analysis has at least two benefits. First, the results can be juxtaposed and the credibility of final estimates can be concluded. Second, in case of missing information for particular country, if the same estimation methodology is used throughout the study, the missing data can sometimes be imputed based on available values in other countries.

2       Methods

A unified questionnaire was used for all countries, enabling comparison of intermediary results and allowing for filling missing data based on average values reported in other countries, if needed. The questionnaire encompassed sections on epidemiology, mortality, treatment patterns (which intrinsically differ depending on the moment of diagnosis) using various types of resources, also the end-of-life treatment, unit cost information, and other economic data (e.g. economic activity) (questionnaire template is given in Online Resource 1). The questionnaires were filled based on available registers and databases, literature, published data, and local clinical experts’ experience, etc. (see Online Resource 2)..

2.1       Epidemiology and mortality

The data were collected split by the stage of disease; if split data were not available, the data for stages III and IV jointly or for the whole BC population were collected.

As treatment patterns may evolve with time, and considering an average patient may be cumbersome in cases where experts’ opinion was used, we separately considered the newly diagnosed (more recently than 12 months) and the remaining patients (i.e. after progression or diagnosed >12 months ago), expecting that it is easier to estimate the costs separately in two clinically distinct groups.

To understand the resource consumption, we collected data on the percentage of patients in whom the procedure was used and the average number of procedures used per year (among patients who used it at least once). 

2.2     Direct cost estimation

In cost estimation, we used the prevalence-based approach, i.e. we multiplied the number of patients (as measured at a given moment) by the average annual cost⁹. The number of patients was defined irrespectively of the disease onset, except the split for the newly and previously diagnosed, as described above. To calculate the average annual cost, the product of the percentage of patients receiving a given procedure, the average number of units of the procedure received per year, and the unit cost was used. The study encompasses three categories of direct medical costs: diagnostics, treatment, and other medical services. The treatment costs were divided into: surgery, breast reconstruction, external breast prosthesis, radiation therapy, chemotherapy, hormonal therapy, and targeted therapy.

Due to data availability, a modified approach was used in Croatia and the Czech Republic. In Croatia, we used the incidence-based approach, i.e. we multiplied the number of new patients per year by the average number of procedures in life-long horizon. In the Czech Republic, we directly multiplied the total annual resource consumption for patients with ABC identified in Czech National Cancer Registry by the unit cost (information extracted from the National Registry of Reimbursed Health Services).

2.3      Indirect cost estimation

In the indirect cost estimation,  the deaths from a single calendar year were assigned the stream of lost future productivity (i.e. the fact that the deceased person does not generate the product in the future). In a sense, this way has some incidence-based approach elements (where death is the event the number of which per year we measure). A purely prevalence-based approach would rather artificially require estimating the total number of people who would be alive in a given moment if they had not died because of ABC.

We used the human capital approach, i.e. accounted for the whole period of illness-related absence from workforce, and not the friction cost method (FCM), in which it is assumed that market adjustments (e.g. new people being hired) will make up for an absent person after some friction time (see, e.g., van den Hout 2010¹⁰ for a more detailed comparison of the two). Firstly, we believe it is more suitable approach of obtaining a single number that expresses the overall disruption to the economy: for example, under FCM the death of a 20-year old would generate the same cost as the death of a 60-year old (ignoring differences in salaries, as both deaths impact the economy only within a short friction period). Secondly, using FCM would require additional assumptions in the multi-country setting: e.g., how the job markets function, how quickly replacements can be found, what is the degree of complementarity/substitution between the employees, country specific friction time, etc.

To estimate the indirect costs of ABC, the country-specific information on the economic activity (like employment rate, sick leaves, average monthly gross salary) was necessary. We used  two sources of indirect costs: the productivity lost due to premature mortality (i.e. before the expected death of a general population, restricted to pre-retirement age) and the productivity lost due to reduced employment rate (because of morbidity). We have collected data on mortality (yearly number of deaths) and age structure (at death; for the age ranges: ≤20 years, 21-30 years, 31-40 years, …, ≥81 years). Based on these data, the annual number of deaths for age ranges was calculated (assuming that all deaths occurred in the middle of the analysed age ranges). The number of potential years of work lost were estimated as follows: restricted mean survival time ranges multiplied by the annual number of deaths in the age ranges and by the employment rate. Finally, indirect costs of premature mortality were calculated as potential years of work lost multiplied by average annual gross salaries.

Productivity lost due to the reduced employment was calculated as follows. Based on the number of ABC patients and age structure (for prevalence; for the age ranges: 21-30 years, 31-40 years, 41-50 years, and 51-60 years) we estimated the size of the ABC population in the working age. The indirect costs of productivity lost due to reduced employment rate was calculated based on these data, as the number of working age population multiplied by decrease in employment rate (i.e. the difference between the general population employment rate and the sick population employment rate) and by the average annual gross salaries.

2.4     Data collection and analysis

Calculating both the direct and indirect cost a two-stage approach was used. First, data from countries were validated and answers for all parts of the questionnaire were analysed separately and compared between the countries. In case of missing data for epidemiology and mortality, this allowed the use of average values from the other countries. This method was not used for the other elements of questionnaire. Analysed countries differ in terms of economic development, price levels, and the scope of public payer health care coverage (i.e. which procedures, drugs etc. are covered by the public payer and which need to be covered by patients by out-of-pocket payments). The costs vary across countries, also due to the differences in general prices levels, clinical practice, and available treatment methods. Therefore, we believe that transferring such data cannot be performed credibly.

Second, the direct and indirect costs were estimated based on the available data. All costs were converted to Euro (using standard exchange rate from European Central Bank of 30.05.2019)¹¹ for comparability. We assumed that unit costs from the patient perspective are zero, if not explicitly reported otherwise.

3       Results

3.1          Epidemiology and mortality

As shown in Tab. 1, the average BC prevalence in women for studied countries amounts to 1.06% (approx. 1,060/100,000 women). The prevalence is highest in the Czech Republic (1,647/100,000 women), and lowest in Romania (506/100,000 women). ABC constitutes on average 20% of total BC patient population.

The annual disease specific mortality rate was calculated as the number of deaths per year divided by total number of patients with BC in an individual country. The highest annual disease specific mortality rate of BC was observed in Romania (approx. 7%). For other countries, the annual BC mortality rate ranges from 2.2% (the Czech Republic) to 4.6% (Israel) (see Online Resource 3). The data on the age structure for mortality are given in Online Resource 4. In the age range 41-50 years (this range was selected as it matters for indirect costs due to large prevalence and non-negligible number of remaining life years before retirement), the highest mortality is in Romania (8.9%), while in other countries it ranges from 5.3% to 6.4%. Other data seem to be consistent between countries. In countries where prevalence was higher for early disease stage (BC stage I and II), the mortality rate tends to be lower. For example, in the Czech Republic, almost 45% of patients are in the stage I of BC and the mortality rate is low; in Romania, almost 25% of patients are in the stage III of BC, and the mortality rate is the highest of all participating countries. 

3.2      Cost

Our results show that the annual direct cost of ABC per 100,000 women is the highest in Slovak Republic (2.7 million EUR) and the lowest in the Czech Republic (0.4 million EUR). The direct costs are not presented for Estonia because of the limited information.

In all the countries, except for Greece, the direct costs resulted mainly from the costs of treatment (Fig. 1, as reported in available data). 

Fig. 1. The cost structure. 

The employment rate (full-time and part-time jointly) in ABC population was only available for Latvia, and it amounts to 39% there. Based on this information, we estimated the employment rate in ABC population in other countries assuming the same relation of employment rate in ABC and general populations (see Online Resource 5 for results).

The results cvering years of potential life lost and of the productive lost are summarized in Tab. 2. The number of years of potential life lost ranges between approx. 2,000 (Estonia) and 55,000 (Poland), also leading to the productive years loss: between 250 (Estonia) and 4,000 (Poland). The high annual disease specific mortality in Poland and Romania results in a high number of years of potential life lost in these countries.

Tab. 2. The years of potential life lost and potential years of work lost in the specific country.

In absolute terms, the indirect cost was estimated as approx. 6 million EUR in Latvia and 121 million EUR in Poland (see Tab. 3). The indirect cost of lost productivity due to premature mortality is related to the number of potential years of work lost (see Tab. 2), which is the highest in Poland and the lowest in Estonia and Latvia. The indirect cost of lost productivity due to reduced employment rate is closely related to the number of the women working age population with ABC in individual countries. In Estonia and Latvia there are about 2,000 women with ABC, while in Poland almost 41,000 (see Tab. 1). In Bulgaria, Poland, and Romania, the percentage of patients in working age (20-60 years) is higher than in the other countries (more than 50%), while in the Czech Republic and Latvia less than 30% women with ABC are aged 20 to 60. Based on the data above, the working age population with ABC is the largest in Poland (about 24,000 women) and the smallest in Latvia (700 women). All indirect costs are linked to average monthly gross salary, which is the highest in Israel (2 628 EUR), and the lowest in Bulgaria and Romania (about 600 EUR). In other countries, the average monthly gross salary is quite similar and ranges between 926 EUR (Latvia) to 1 530 EUR (the Czech Republic).

The estimates of the indirect costs per 100,000 women is rather consistent between countries and ranges between 0.4 million EUR (Romania) to 1 million EUR (Estonia and Israel) (see Tab. 3). To a significant extent, ABC occurs in young patients of working age. Hence, premature deaths prevent patients from contributing to the economy and incur economic burden on society. As a result, the indirect costs weigh heavily up to 55% of the total costs of ABC (see Fig. 1).

Finally, the average (for all countries) total costs per 100,000 women amounts to 1.8 million EUR. This finding complements the fact that BC among all the cancers has one of the highest economic costs per country in the European Union13. Luengo-Fernandez13 showed that lung cancer had the highest economic cost (18.8 billion EUR, 15% of overall cancer costs in the European Union in 2009), followed by breast cancer (15 billion EUR, 12%), colorectal cancer (13.1 billion EUR, 10%), and prostate cancer (8.43 billion EUR, 7%). The results of estimation of total annual cost of ABC are summarized in Tab. 3. 

4       Discussion

In this study we estimated the economic burden of ABC: the direct medical costs (defined from the public-payer and patient perspectives) and the indirect costs (societal perspective). This multitude of perspectives sheds more light on the overall economic burden of the illness and demonstrates how various components weigh overall. On the other hand, the multinational context of our analysis allowed us to detect potential problems with data (where values differed substantially between the countries) or replace the missing data (in case of epidemiology, where we believed the transferring data can be done rather credibly). Although we have included quite many countries, we managed to maintain a unified approach to data collection and analysis, with a few exceptions as indicated above.

As in some cases data of sufficient quality were inaccessible (e.g. information about the treatment related to AE or other complications), some cost categories may be inaccurately estimated. The differences in the level of total costs of ABC between countries do not necessarily mean that the costs differ so much, but rather that the access to reliable data or the nature of this data differs. Only for Bulgaria, the Czech Republic, Israel, Latvia, Romania, and Slovak Republic all components of direct costs are known. As missing categories were typically omitted, we tend to treat our results as a lower bound of the actual economic burden.

Still, our results show that the estimated total cost of ABC is rather consistent among the countries. The total costs of ABC per 100,000 women ranges between approx. 1 million EUR (Romania) to 3.4 million EUR (Slovak Republic). As the data were consistent, we believe that these numbers are one of the major findings of our research. Because of the possible downward bias due to data unavailability, in future analyses it may be worth considering a country result but also an average result (for every 100,000 women in the general population, ABC generates approximately 1.8 million EUR annually).

Regarding the cost structure, even though ABC occurs frequently in the elderly (almost 60% of patients are over 60 years old), the indirect costs constitute a large part of the total: on average, they are responsible for 55% of the cost. The earlier assessments confirm our estimate: in a Swedish study the indirect cost was assessed to be 50%14, while in the Netherlands the total cost of BC was estimated at 1.27 billion EUR, of which 768 million EUR (60%) is the healthcare expenditure, 260 million EUR (20%) is the indirect cost of morbidity, and 243 million EUR (19%) is the indirect mortality cost15. Owing to this high share, omitting indirect costs in burden of illness studies may not reveal the complete picture. We also conclude that these findings confirm the importance and additional benefits of early diagnosis.

Importantly, our indirect cost estimates are conservative, as they do not include the cost of sick leaves or of presenteeism (reduced productivity while present at work).

Regarding the direct cost component, in all the countries (except for Greece) the direct costs resulted mainly from the costs of treatment.

Early detection of BC is also financially beneficial in terms of direct cost. The cost of treatment is much smaller in the early stages of disease. This finding seems to be in line with other analyses presented in the literature: it was found that treating advanced- versus early-stage BC is associated with increases in costs (costs increased with increased stage of cancer)¹⁶.

Obviously, our study is subject to several limitations. Burden of disease studies bear lot of limitations due to data collection as well as inherent differences among countries (related to delivery, financing and organization of health care as well as cultural differences). Chronic diseases, including cancer, are highly country-specific, thus comprehensive and uniform approach to resource use and costing are challenging. With study pilot we could identify but not fully adjust for these diversities. Especially the coverage and funding methods are very specific, as different care items could be contracted separately, pooled, or financed within various budgets.

Due to the wide scope of the requested data, the collection proceeded in an iterative way, with data being scrutinized, compared across countries, and amended, if needed. As mentioned above, we find this to be a difficulty but also an advantage of multinational studies. As mentioned, the variety of available format of data as well as the data quality differs among the countries. Therefore, the comparability of the individual cost components between the countries is rather limited. Fortunately, the final aggregated results are fairly consistent. We conclude that using these total cost estimates is rather well-grounded and safe. This is especially the case for indirect costs, where there are fewer parameters used in the estimation. Apart from the estimates, the present study indicates there are still issues with data availability or quality. For example, in most of the countries the number of patients with BC split by the stage of disease, the information about employment rate in ABC/BC population or data on sick leaves were not available. In all the countries, the information about treatment related to AE or other complications were limited and insufficient to calculate related costs. In Estonia and Greece, information on unit cost of diagnostic and other medical services was limited. The cost of other medical services was omitted in three counties (Croatia, Greece, and Poland) because it was not possible to either obtain data from existing databases or get reliable data through the interviews with experts. For example, implementing national registers would allow for more accurate estimates to be obtained in the future, which could result in more informed decisions on resource allocation. Finally, the retrospective, bottom-up like design, input data driven by the quality of specific epidemiological data justify careful consideration of our research findings. We were also unable to project future burden, which is likely to double in the next 15 years¹⁷.

5        Conclusions

ABC is associated with substantial healthcare costs and imposes a significant societal burden, as indicated by the high indirect costs. Early detection, timely intervention, and effective treatment of early stage BC can lead to the decrease of costs associated with ABC while improving the overall disease prognosis. Our findings may be used in informing decisions on resource allocation, improving cancer policies, and supporting national cancer plans. Better data availability would improve the quality of estimates and lead to more informed decision making. 

Funding 

The study was funded by Novartis. 

Supplement
Online Resource 1

Introduction

Thank you for agreeing to participate in the study; we appreciate your time and expertise! We ask for information to assess the cost of advanced breast cancer (ABC) in several European countries, including yours, in order to increase the awareness of this disease. By ABC, we understand stages III (locally advanced breast cancer) and IV (metastatic breast cancer). We aim to estimate the ABC cost split by the stage of the disease; thus, we will be grateful for filling the data separately for each stage. If relevant data are not publicly available, please try to estimate them (e.g. using experts’ opinions). If split is unavailable, provide answers for ABC jointly or, at worst, data for the overall BC population. In case of mortality, we ask for similar information in various ways, not knowing what kind of data may be available in your country. Based on data availability and quality we will choose the analytical strategy.

Once again, thank you for your valuable time!

Contact information

Epidemiology

Prevalence

Please provide 2017 data, if available; if using older data, provide info on the year. If relevant data are available, do fill all fields; we may use your data to fill in the gaps in other countries.

Annual disease specific mortality

Please provide 2017 data, if available; if using older data, provide info on the year. If relevant data are available, do fill all fields; we may use your data to fill in the gaps in other countries.

Age

Please provide data on the age structure for the age ranges below. If needed, use your own ranges. Provide the most recent data with info about the year.

Please provide the mean age at death, per stage. If not available, provide values for the III and IV jointly or for the whole BC population.

Economic activity

In order to estimate the indirect costs of ABC, we ask for info on the economic activity.

Instructions for the remaining part of the questionnaire

The remaining part consists of four sections: diagnostics, treatment, other medical services, and end of life management; in which we ask for different kind of resources. As treatment patterns may evolve in time, in each section, we separately ask for data for newly (≤12 months before) diagnosed patients and other patients (i.e. patients after progression to a given stage or having been diagnosed >12 months ago). We believe averaging values for so different patients could be cumbersome. To understand the resource consumption, we ask about the percentage of patients in whom the procedure is used and the average number of procedures used per year (amongst patients who use it at least once). We also ask about the unit cost (both from public payer and patient perspective). Please follow the suggestions below.

1.    Please try to provide data split by stage. If not possible, stages III and IV jointly. In some cases, we ask for overall BC population.

2.    Several medical procedures may be financed jointly within some broad category (e.g. a DRG) – in such cases please provide data on this broadest category only. Report all the procedures which are financed separately (e.g. are not included in hospitalization tariff) and ignore procedures which are included in hospitalization tariff etc., i.e. avoid double counting.

3.    In the diagnostics section, some procedures are used only in newly diagnosed (e.g. biopsy), while others are also used during monitoring of the disease. Please note that we ask for these groups separately (i.e. newly diagnosed vs other patients).

4.    In the treatment section, for some procedures (e.g. surgery) we need only data for stage III and IV BC (if split data are not available, then for ABC population jointly). However, for drug therapies we need data for each stage, ABC jointly, as well as the overall BC population. This will allow us to estimate relationship between cost in BC and ABC (split by stage) and will be further used for countries where detailed data are not available.

5.    If you have a publication with costs (e.g. hormonal therapy) calculated, please provide the average cost (with the information about the data source and year for which it was calculated). However, try also to provide the detailed data: the method of cost estimation used in the publication may differ from the method used in other countries and may not include all currently available drugs.

6.    Use your national currency. Whenever data are outdated and inflation should be accounted for, report this and provide details.

7.    Please provide references. This will be needed, e.g. when preparing publication.

8.    Add new rows if needed.

Diagnostics

Newly diagnosed patients

Proportion of patients

Resource consumption

Patients after progression or >12 months after the diagnosis

Proportion of patients

Resource consumption

Unit costs

Treatment

Newly diagnosed patients

Proportion of patients