A cross-country review of data collected on non-small cell lung cancer (NSCLC) patients in cancer registries, databases, retrospective and non-randomized prospective studies

Abstract

Introduction:

An increased number of pharmacotherapies exist to treat advanced NSCLC. This necessitates a review of the available information on routine-care treatment patterns, the outcome of treatment, and resource utilization for patients diagnosed and treated with advanced NSCLC that could inform evidence-based treatment decisions and aid decisions on the most cost-effective treatment alternatives.

Methods:

PubMed and the Health Economic Evaluations Database were searched for retrospective or non-randomized prospective studies between January 2000 and May 2012 that included information on treatment patterns, treatment outcomes including health-related quality-of-life (HRQoL), and resource utilization. In addition, registries and databases were identified from retrieved publications and internet searches. Data collected in registries and databases was summarized for eight European countries (Belgium, France, Germany, Italy, Sweden, Turkey, the Netherlands, the UK), Australia, and Canada.

Results:

The literature search resulted in 410 studies, whereof 87 studies met the study inclusion criteria. In total, 49 were retrospective chart reviews or database analyses, 30 non-randomized prospective studies, and eight HRQoL studies. Two studies compared treatment patterns and/or treatment outcomes across countries. Altogether, 181 cancer registries in the countries studied were identified. Clinical cancer-specific patient registries were identified in Australia and Germany. Databases or linkage systems that enable retrieval of complete information of patient disease history were found in Australia, Canada, the Netherlands, Sweden, and the UK. Cancer registries and databases were found to collect information on NSCLC patient demographics, NSCLC or lung cancer diagnosis, disease stage, performance status, treatment, treatment outcomes, and resource use. Differences existed between country registries and databases in whether information was collected on each of these data points.

Conclusion:

The literature review revealed few published NSCLC studies on treatment, treatment outcomes, and resource use in routine clinical practice and on HRQoL. Registries and databases were found to collect some of this information, however not systematically.

• Non-small cell lung cancer
• Treatment patterns
• HRQoL
• Literature review
• Registries
• Patient databases
• Retrospective and prospective studies

Introduction

Lung cancer is the leading cause of cancer-related mortality worldwide and accounts for more than 1 million deaths annually1. Approximately 85% of lung cancers are classified as non-small cell lung cancer (NSCLC), and the majority of patients present with locally advanced or metastatic disease, the most common form of lung cancer. Prevention and treatment of NSCLC remains a major clinical challenge given an increasing number of smokers, particularly in developing countries, late diagnosis, and an overall 5-year survival rate of less than 15%2–4. Even though traditional cytotoxic therapy for advanced NSCLC has evolved lately, it has resulted in modest improvements in health-related quality-of-life (HRQoL) and overall survival (OS)5. Efforts are now being directed at developing targeted therapies which have been found to improve progression-free survival and OS in advanced NSCLC5.

In an effort to improve cancer control in response to the World Health Assembly resolution on cancer and prevention and control (WHA58.22) the WHO has requested each country to develop National Cancer Control Programmes (NCCPs)6,7. The aim of the NCCPs is to reduce cancer incidence and mortality and improve quality-of-life of cancer patients, through a systematic implementation of evidence-based strategies for prevention, early detection, diagnosis, treatment, and palliation, and to make efficient and cost-effective use of available resources. In line with the WHO initiative, the European Commission (EC) has agreed on a European Partnership for Action Against Cancer8, and one of the objectives for action is to ensure accurate and comparable data on cancer incidence, prevalence, morbidity, cure, survival, and mortality in the European Union by 2013. The collection of data and information will enable, mainly through cross-country comparisons, the identification and promotion of evidence-based best practice in cancer prevention and control. In order to be able to make the best possible decisions on the most cost-effective treatment alternatives for different types of cancer, it is of importance that national and regional population-based registries contain high quality information and, above all, adequate information to be able to make evidence-based treatment decisions.

In concordance with the initiatives taken by both the WHO and the EC and given the increasing availability of new treatment alternatives, we sought to explore if country-specific and comparative information on routine clinical practice treatment patterns, treatment outcomes, resource use, and HRQoL would be available from databases, registries, retrospective, and non-randomized prospective studies, in order to make evidence-based treatment decisions and to make the best possible decisions on the most cost-effective treatment alternatives in advanced NSCLC patients. Sources of information were identified and presented that contain data on real-life current treatment patterns, clinical outcomes, HRQoL, as well as resource use and economic burden associated with advanced NSCLC.

Methods

Literature review

The literature review included a structured literature search in English language with the aim to collect information on previous retrospective chart reviews and prospective non-randomized observational studies describing routine care treatment patterns, resource utilization, and treatment outcomes including HRQoL measured with the disease generic EuroQol-5D (EQ-5D) instrument for patients with advanced NSCLC, defined as stage IIIb and stage IV NSCLC patients. PubMed and the Health Economic Evaluations Database (HEED) were searched for studies published between January 2000 and May 2012 using the search strings presented in Table 1.

Table 1.  Search strategy for literature review.

Type of search	Search string
Standard search string (#1)	(‘advanced nsclc’ OR ‘advanced non-small-cell lung’ OR ‘advanced non-small cell lung’ OR ‘metastatic nsclc’ OR ‘metastatic non-small-cell lung’ OR ‘metastatic non-small cell lung’ OR ‘stage IIIb nsclc’ OR ‘stage IIIb non-small-cell lung’ OR ‘stage IIIb non-small cell lung’ OR ‘stage IV nsclc’ OR ‘stage IV non-small-cell lung’ OR ‘stage IV non-small cell lung’)
Retrospective or prospective observational studies (#2)	#1 AND (‘chart review’ OR ‘retrospective study’ OR ‘observational study’ OR ‘prospective study’ OR ‘retrospective analysis’ OR ‘prospective analysis’ OR ‘expanded access program’)
Resource utilization	#1 AND (‘resource utilization’ OR ‘resource utilization’ OR ‘resource use’ OR ‘resource usage’)
Cost and economics	#1 AND (econom* OR cost)
Treatment patterns	#2 AND (‘treatment pattern’ OR ‘treatment sequence’ OR ‘treatment line’ OR ‘treatment duration’ OR ‘treatment cycles’ OR ‘therapy duration’ OR ‘therapy cycles’ OR ‘best supportive care’ OR ‘supportive care’)
Treatment outcomes	#2 AND (survival OR ‘response status’ OR ‘treatment outcome’)
Quality-of-life	#1 AND (utility OR EQ-5D OR ‘EQ 5D’ OR ‘standard gamble’ OR ‘time trade off’)

Database and registry mapping

Databases and registries were identified from the retrieved publications, from the International Association of Cancer Registries (IACR) homepage and from internet searches. Data collected in registries and databases was summarized for 10 countries, including in Europe; Belgium, France, Germany, Italy, Sweden, Turkey, the Netherlands, and the UK, as well as Australia and Canada, in order to allow for cross-country comparison. The search for the presence of relevant data on advanced NSCLC in databases and registries were restricted to exploring the information that was displayed on the respective homepages, and in the case that no information was presented the respective databases or registries were contacted.

Results

Results of the literature search

The literature search resulted in 410 unique studies, of which 87 studies met the search criteria (see Figure 1 for a PRISMA flow diagram and Table 2 for a summary by search term). Forty-nine studies were retrospective chart reviews or database analyses, 30 non-randomized prospective studies, one was based on a Delphi panel discussion, and eight were health-related quality of life (HRQoL) studies (one of the prospective studies also collected EQ-5D data and was therefore counted in both categories). Only a few studies (n = 23) were found to publish data on routine clinical practice in advanced NSCLC patients, which included 14 studies on treatment patterns, 12 studies on treatment outcome, seven studies on resource use, and two studies collecting HRQoL information. A list of all studies meeting the search criteria and divided by search term is provided in Supplementary Appendix 1.

Figure 1.  PRISMA flow diagram of studies identified from the structured literature search.

Table 2.  Results of the literature review.

Search term	Number of studies found*	Whereof international	Routine clinical practice treatments	Comments/conclusions
Treatment patterns in advanced NSCLC	14	2	14	• Four studies12,17,19,20 restricted to presenting only the initial choice of treatment, and three10,15,18, five11,13,14,16,22 and two9,21 studies restricted to presenting 1st–2nd, –3rd, or –4th line of treatment, respectively, without describing the complete treatment history of the patients. • Two international studies describing general treatment patterns in 1st–2nd line10,15. Although one of the studies collected treatment pattern information from five countries, treatment data was not presented in a country-specific manner and no comparisons between countries were made10. The other international study only presented country-specific data for three selected 2nd line regimens15. • 8/10 studies describing treatment patterns in 2nd and/or subsequent lines collected patient data or were published at a time when drugs such as erlotinib and gefitinib had not been approved for 2nd- and subsequent lines of treatment for advanced NSCLC9–11,14,16,18,21,22. 13/14 studies describing treatment patterns in 1st line collected patient data or were published at a time when drugs gefitinib had not been approved for 1st line treatment of EGFR + NSCLC9–12,14–22. As a result, it can be concluded that current treatment practices are not represented in these studies.
Treatment outcomes	65	1	12	• Most studies describing treatment outcomes have focused on selected treatment options such as erlotinib, gefitinib, or docetaxel (n = 51)23–73, or specific patient groups (n = 6)19,74–77, and have not compared treatment outcomes between various treatment options used in general clinical practice. • 9 studies9–11,17,18,21,22,78,79 were found that presented survival data on treatments in routine clinical practice. However, survival data was only presented per line and not specified by type of regimens given except for one study where survival was presented for 1st line therapy groups10.
Resource utilization	15	0	7	• 8 of the studies reported on resource use either for specific patient groups20,82,85, specific therapies52,80,86,88, or were restricted to certain managing phases of the patients83. • 4 studies reported on the complete direct cost of managing patients with advanced NSCLC in the Netherlands, Canada, Spain, and Australia, respectively13,18,84,87. • In 1 study, a Markov model was developed, populated with data from routine management of newly diagnosed lung cancer cases in France81. • 1 retrospective chart review study aimed at comparing the direct cost of care of advanced NSCLC patients with and without disease progression following 1st line treatment82. • In 1 retrospective database analysis, the lifetime medical cost were calculated stratified on the basis of type of treatment patients received during the first 3 months following diagnosis19.
Health-related quality-of-life studies including the EQ-5D instrument	8	2	2	• 3 studies restricted to those based on the general public’s estimates of disease-related utilities, rather than estimates by patients (standard gamble interviews and EQ-5D-VAS89,91; EQ-5D-VAS90). • 2 studies reported on utility values for patients on specific treatment regimens included in clinical trials (EQ-5D health status score and VAS-scale92,93. • 3 studies reported on utility values for patients treated in routine clinical practice (EQ-5D health status score and VAS-scale10,94,95).

*87 studies met the inclusion criteria. Number of studies sum up to above 87 since more than one search term category might be studied in one publication.

Treatment patterns in advanced NSCLC

Fourteen published studies were identified that described routine clinical practice treatment patterns for patients with advanced NSCLC, whereof 12 studies described treatment patterns in one specific country only (one in Brazil, one in France, three in Germany, one in Japan, one in the Netherlands, one in Spain, and four in the US, respectively)9–22 (Table 3). Furthermore, four of these studies were restricted to presenting first line treatment, without describing the complete treatment history of the patients12,17,19,20. To date, only two international studies on advanced NSCLC describing general treatment patterns was found10,15. The first study, which is called ‘Assessment of Cost and outcomes of chemotherapy In an Observational setting in patients with advanced NSCLC’ (ACTION), enrolled 925 patients with stage IIIb/IV disease from five European countries (Finland, Germany, Portugal, the Netherlands, and the UK). Patients were followed from initiation of first line chemotherapy until death or for a maximum of 18 months. Although this study collected treatment pattern information from five countries, treatment data was not presented in a country-specific manner, and no comparisons between countries were made. The second study, called the ‘Survey in European Lung Cancer Evaluating Choice of Treatment and Tolerability in Observed NSCLC’ (SELECTTION), enrolled 1012 advanced NSCLC patients from 11 countries (Denmark, Finland, France, Germany, Israel, the Netherlands, Peru, Portugal, Romania, Sweden, and the UK) receiving second line treatment. Patients were followed from initiation of second line therapy until death or for a maximum of 12 months. Treatment cohorts were constructed based on the patients’ distribution across second line treatments that were assigned by physician decision (pemetrexed (46.2%), docetaxel (22.9%), erlotinib (20.4%), and ‘other’ (10.5%)). Country-specific treatment pattern data was only presented for the three treatment cohorts.

Table 3.  Summary of studies describing treatment patterns in routine clinical care for patients with advanced NSCLC.

Country	Reference	Type of study	No of patients	Study period	Treatment types described	Main findings
Brazil	Naime et al.17	Retrospective chart review	564	1990–2003	1st line chemotherapy	There was a great heterogeneity in the regimens of drugs to treat metastatic NSCLC patients.
France	Girard et al.11	Retrospective chart review	173	2000–2006	1st until 3rd line chemotherapy	From a population of 613 patients treated with 1st line treatment, a total of 173 patients received 3rd line treatment (cytotoxic chemotherapy in 131 patients; EGFR tyrosine kinase inhibitors in 42 patients). Patients with advanced NSCLC may benefit from 3rd line treatment. The best candidates can be identified using standard prognostic factors, such as PS, and disease control after 1st- and 2nd-line treatments.
Germany	Koeppler et al.14	Retrospective chart review	212	1995–2007	1st until 3rd line chemotherapy	1st line treatment consisted of a platinum/taxane regimen in 95 pts (49%), a platinum/non-taxane combination in 35 pts (18%), a non-platinum combination in 22 pts (11%), and a monotherapy (gemcitabine, vinorelbine, or docetaxel) in 42 pts (22%). 126/194 pts (65%) received a 2nd line treatment and a 3rd line chemotherapy was given to 75/194 pts (39%).
Germany	Zietemann et al.21	Prospective observational study	416	2003–2007	1st until 6th line chemotherapy	At least 50% of patients with advanced NSCLC receiving systemic 1st line therapy are subsequently being treated with a 2nd-, at least 25% with a 3rd-, and at least 10% with a 4th-line systemic chemotherapy.
Germany	Zietemann et al.22	Prospective observational study	406	2003–2010	1st until 3rd line chemotherapy	The majority of patients (84%) were treated with platinum-based doublets at 1st line (56% carboplatin-containing). Of the 406 individual patients 52%, 27%, and 11% received a 2nd- (31% docetaxel), 3rd- (28% erlotinib), and 4th- line therapy, respectively. Of the patients receiving 2nd and 3rd line treatment, 89% and 94% were pre-treated with platinum-based doublets at 1st line, respectively.
Japan	Asahina et al.9	Retrospective hospital registry analysis	599	2002–2006	1st until 4th line chemotherapy	Significant differences in age, performance status at the start of 1st line chemotherapy, and histology were observed between patients who received 3rd line chemotherapy and those who did not. Docetaxel, gefitinib, and S-1 were the most frequently used regimens for 3rd- or 4th line. As many as 38% of patients with advanced NSCLC who received 1st line chemotherapy could receive 3rd line chemotherapy.
The Netherlands	Pompen et al.18	Retrospective chart review	102	2001–2005	1st and 2nd line chemotherapy	Overall, the management of unresectable advanced NSCLC appeared to follow current treatment guidelines in the Netherlands.
Spain	Isla et al.13	Delphi panel	NA	2010	1st until 3rd line chemotherapy and best supportive care (BSC)	Patients with PS 0–2 are expected to receive anti-cancer pharmacotherapy, while patients with PS 3–4 receive mainly BSC. In 1st line treatment of patients with a good PS (0–1), dual combinations of anti-cancer pharmacotherapies containing cisplatin or carboplatin are mostly used (85%). The main treatments used as 1st line therapy in patients with PS 2 or the elderly are monotherapies with vinorelbine (30%), gemcitabine (15%), pemetrexed (15%), erlotinib (15%), and docetaxel (11%). The usual therapies in 2nd line NSCLC patients with PS 0–2 are docetaxel (squamous NSCLC, smoker) (30%), erlotinib (non-squamous NSCLC, non-smoker, EGFR mutation positive) (30%), and pemetrexed (non-squamous NSCLC, smokers) (40%). Erlotinib is the most commonly used anti-cancer therapy in 3rd line treatment (65%), although some patients (35%) receive chemotherapy (docetaxel, pemetrexed, gemcitabine, or vinorelbine in monotherapy) depending on the 1st- and 2nd line treatments received.
USA	Ramsey et al.19	Retrospective database analysis	14,875	1994–2001	1st line chemotherapy	Although chemotherapy prolongs survival in community settings, the majority of patients did not receive chemotherapy and a marked variation in prescribing patterns for initial chemotherapy was observed across various geographic regions.
USA	Stokes et al.20	Retrospective database analysis	5138	1998–2002	1st line chemotherapy	The 1st line chemotherapy regimens used by patients with afebrile and febrile neutropenia differed from those used by patients without neutropenia (p < 0.001 for both comparisons). The duration of 1st line chemotherapy was longer for patients with afebrile neutropenia and febrile neutropenia than for patients without neutropenia (mean [SD] months 5.5 [7.0], 4.5 [4.8], and 3.6 [4.9], respectively, p < 0.001 for both comparisons).
USA	Murillo et al.16	Retrospective chart review	417	2000–2003	1st until 3rd line chemotherapy	An increased availability of new chemotherapeutic agents has resulted in an increased use of chemotherapy near the end of life. At time of death, the majority of patients received 2nd line chemotherapy.
USA	Greer et al.12	Retrospective chart review	50	2003–2005	1st line chemotherapy	Initial chemotherapy included a platinum-based doublet (74%; e.g., carboplatin and paclitaxel), an oral EGFR–TKI (14%; e.g., gefitinib), or a parenteral single agent (12%; e.g., vinorelbine). The majority of patients (60%) received 2nd line chemotherapy, and many (38%) underwent 3rd line therapy or greater.
Multi-country*	Bischoff et al.10	Prospective observational study	975	2003–2004	1st and 2nd line chemotherapy	As 1st line treatment, most patients (85.8%) received a third-generation compound (gemcitabine, taxanes, or vinorelbine), generally in combination with platinum. Physician’s perception of efficacy was the most common factor influencing choice of treatment. 2nd line chemotherapy was planned for 29.2% of patients, and monotherapy of any of the above regimens was more common than combination regimens.
Multi-country**	Moro-Sibilot et al.15	Prospective observational study	1012	2006–2008	1st until 2nd line chemotherapy	Pemetrexed, then docetaxel and erlotinib were the most frequently prescribed 2nd line treatments, which is in line with international guidelines. Erlotinib was most commonly prescribed to patients with adenocarcinoma, never-smokers, and females. Pemetrexed was more frequently prescribed than docetaxel, with physicians most commonly choosing to prescribe the former agent because of its tolerability profile.

*The Netherlands, Germany, UK, Portugal, and Finland.

**France, Germany, Portugal, Finland, Denmark, UK, Sweden, the Netherlands, Israel, Romania, and Peru.

Treatment outcomes

Most studies describing treatment outcomes have focused on selected treatment options such as erlotinib, gefitinib, or docetaxel23–73, or specific patient groups19,74–77, and have not compared treatment outcomes between various treatment options used in general clinical practice. Nine studies were found that presented survival data on treatments in routine clinical practice. Eight of the nine studies only present survival data per line of treatment and not specified by the various types of chemotherapy regimens given9,11,17,18,21,22,78,79, while one study presents survival data by 1st line chemotherapy regimen10. In the nine studies that presented survival data on treatments in routine clinical practice, median overall survival (OS) from initiation of 1st line therapy varied from 8.3 months (95% CI = 7.5–9.2) for Brazilian stage IV NSCLC patients17 to 16.4 months for French stage IIIb/IV patients that received at least three lines of therapy11. The only international study retrieved, and which included patients from Finland, Germany, the Netherlands, Portugal, and the UK, reported on a median OS, both for the overall treatment group (9.3 months; 95% CI = 8.6–10.3) and by 1st line chemotherapy regimen (between 7.9 months (for 2nd generation agents + other) and 9.9 months (for 3rd generation agents + platinum)10.

Resource utilization and cost of care

Fifteen studies, whereof 13 were retrospective chart reviews or database analyses, one was a prospective observational study, and one was based on a Delphi panel discussion, were found to explore the cost of treating advanced NSCLC. Resource utilization were in these studies estimated for patients treated in Australia, Canada, France, Italy, the Netherlands, Poland, Spain, the UK, and the US, respectively (Table 4)13,18–20,52,63,80–88. The studies estimated either total life-time costs or were restricted to presenting the cost of certain phases in the management of the disease, such as the cost of treating patients after the completion of chemotherapy until death. Seven of these studies described resource utilization and costs for treatment regimens used in routine clinical practice without any selection of specific patient groups, e.g., restricting the study population to older patients13,18,19,81,84,87. Three of those studies were retrospective chart reviews that reported on the complete direct cost of managing patients with advanced NSCLC in Canada, the Netherlands, and Australia, respectively. Types of resources utilized were either measured from completion of primary chemotherapy87, from the time of diagnosis18 or first consultation84, until the time of death or the end of the evaluation period. All three studies concluded that hospitalization is the major cost driver for this patient group. In a study by Isla et al.13 a two-round Delphi consensus panel of clinical experts was carried out to describe use of resources and costs associated with treating advanced or metastatic NSCLC patients in Spain. Results demonstrated that healthcare costs are impacted by line of treatment, patient performance status, type of administration of therapy, and adverse event management. In another study, a Markov model was developed, populated with data from routine management of newly diagnosed lung cancer cases in France, with the aim to assess the mean cost of the clinical management and the cost of the different management phases81. It was concluded that the cost of 1st line chemotherapy and the percentage of actively treated patients impacted the mean cost more than did the cost of 2nd or 3rd line chemotherapy regimens or the cost of palliative care. Another retrospective chart review study on US patients aimed at comparing the direct cost of care of advanced NSCLC patients with and without disease progression following first line chemotherapy, and concluded that patients who experience progression have significantly higher costs than similar patients with stable disease82. In a retrospective database analysis, Ramsey et al.19 calculated the lifetime medical cost for advanced NSCLC patients stratified on the basis of type of chemotherapy agents patients received during the first 3 months following diagnosis. Medical care costs were considerably higher for those patients receiving chemotherapy, with the highest cost found for patients receiving a platinum compound in combination with a taxane.

Table 4.  Summary of studies describing resource utilization for patients with advanced NSCLC.

Country	Reference	Type of study	No of patients	Study period	Resources analysed	Main findings
Australia	Kang et al.84	Retrospective chart review	210	2005–2008	To describe the current pattern of resource use and direct medical costs associated in managing lung cancer in South Western Sydney, Australia.	The median cost for managing NSCLC was A$10,675 (range A$669–612,789). Hospitalization and cancer treatment, particularly chemotherapy, accounted for the major components of direct medical costs.
Canada	Navaratnam et al.87	Retrospective chart review and database analysis	150	1997–2000	To examine the costs of management of NSCLC after completion of chemotherapy until death in a population of patients in Manitoba, Canada.	The largest overall component of cost after the end of chemotherapy was hospitalizations. Effective new therapies that reduce the episodes of hospitalizations would have a significant impact on decreasing aggregate costs.
Canada	Bradbury et al.80	Retrospective review of a trial database	731	2001–2003	Economic analysis of erlotinib treatment in a clinical trial and to determine the cost-effectiveness of treating various patient populations with erlotinib.	Erlotinib treatment for patients with previously treated advanced NSCLC is marginally cost-effective. The use of molecular predictors of benefit for targeted agents may help identify more or less cost-effective subgroups for treatment. Erlotinib treatment was the main cost driver followed by hospitalizations.
France	Chouaid et al.81	Retrospective chart review	428	1998–1999	To evaluate the mean cost of managing patients with lung cancer according to histological type and initial stage, during different phases of disease management.	The cost of 1st line chemotherapy and the percentage of actively treated patients impacted more on the mean cost than did the cost of 2nd- or 3rd-line chemotherapy regimens or the cost of palliative care.
Italy	Gridelli et al.83	Prospective observational study	104	2004	To assess economic data regarding the home assistance burden for advanced NSCLC patients.	The principal caregiver support was the main cost item representing 74% of total assistance cost during the 3-month observation period. Regression analysis showed a positive correlation between severity of symptoms and the cost of assistance. Caregiver burden was higher in patients with bone and/or brain metastases.
The Netherlands	Pompen et al.18	Retrospective chart review	102	2001–2005	To explore the cost of treatment of advanced NSCLC, and comparing patients receiving 2nd line treatment consisting of only BSC (group A) to patients on active chemotherapy in addition to BSC (group B).	The average total treatment cost per patient per year in the Netherlands was €32,840 in group A and €31,187 in group B. In both groups, hospitalization was the major cost driver.
Poland	Skowron et al.88	Retrospective chart review	87	1997–1999	To estimate the direct costs of three different chemotherapy regimens for advanced NSCLC: cisplatin/etoposide, cisplatin/vinorelbine, and cisplatin/gemcitabine	The cost of drugs and the inpatient administration of chemotherapy were the main sources of expenditure during chemotherapy for advanced NSCLC.
Spain	Isla et al.13	Delphi panel	NA	2010 (cost year 2009)	To describe treatment patterns, use of resources, and the costs associated with treating advanced or metastatic NSCLC patients in Spain (preparation and administration of anti-cancer pharmacotherapy; management of adverse events associated with anti-cancer pharmacotherapy; and best supportive care).	The total cost per patient with advanced or metastatic NSCLC from starting anti-cancer therapy until death was estimated to be between €11,301 and €32,754 depending on the number of treatment lines received. Healthcare costs are impacted by the line of treatment, patients’ PS, the type of administration of therapy, and the management of adverse events.
UK	Maslove et al.86	Retrospective chart review	194	1999–2000	To assess the cost incurred with the use of chemotherapy for patients with advanced NSCLC (through analysis of a large multi-centre randomized trial, which in part compared BSC with or without cisplatin-based chemotherapy	Patients that receive cisplatin-based chemotherapy had an average of 3.4 more inpatient bed days than the mean of 11.9 days for patients receiving BSC alone, and more outpatient visits. The mean total cost for patients receiving chemotherapy was £5355 (cost associated with the administration of chemotherapy was £1233 and non-chemotherapy costs were £4122) compared with £3595 for the BSC group.
USA	Ramsey et al.19	Retrospective database analysis	14,875	1994–2001	To determine the lifetime medical care costs in relation to initial treatment of advanced NSCLC.	Lifetime costs were highest for those patients receiving platinum + taxane combination therapy, exceeding other regimens by more than $10,000 per patient.
USA	Lang et al.85	Retrospective database analysis	3130	1997–2002	To identify total lifetime medical-care costs and costs associated with first-line chemotherapy treatment among older patients with advanced NSCLC treated with commonly used two-drug (‘doublet’) chemotherapy.	Total lifetime medical-care costs and on treatment costs among study patients were US$70,000 and US$30,000, respectively. These costs were driven by hospitalization and physician costs. Although doublet therapy with platinum and a taxane was the most frequently utilized regimen, it was associated with the highest lifetime and on-treatment costs.
USA	Stokes et al.20	Retrospective database analysis	5138	1998–2002	To evaluate (a) costs and medical resource use for chemotherapy-related afebrile and febrile neutropenia in an elderly population with advanced NSCLC, and (b) costs unrelated to neutropenia and total all-cause healthcare costs during first-line chemotherapy.	For elderly patients undergoing first-line chemotherapy for NSCLC, neutropenia, particularly febrile neutropenia, is associated with substantially higher total all-cause healthcare costs.
USA	Fox et al.82	Retrospective chart review and database analysis	306	2001–2006	To compare direct costs of care of advanced NSCLC patients with and without disease progression following first line chemotherapy.	Patients who experience progression have significantly higher costs than similar patients with stable disease. Multivariate regression to estimate predictors of costs after progression. Costs were found to vary only with younger age.
USA	Stepanski et al.63	Retrospective chart review	45	2004–2007	To describe treatment use patterns for patients treated with single-agent erlotinib as either 2nd- or 3rd line therapy from 4 community oncology clinics. Resource use was assessed by the number of clinic visits and hospitalizations, and the use of concomitant medication.	Although nearly 25% of patients had ≥1 hospitalization during erlotinib treatment, none of these hospitalizations was documented by the treating physicians as being caused by complications from erlotinib treatment. Since erlotinib is administered orally rather than i.v. it could reduce clinic costs of administering the treatment.
USA	Nadler et al.52	Retrospective chart review	610	2006–2009	To describe clinical and economic outcomes among patients receiving 2nd line monotherapy with erlotinib, docetaxel, and pemetrexed for NSCLC in a large network of outpatient community clinics in the US.	Relative to pemetrexed, total adjusted costs per patient per month (PPPM) was $1579 lower for docetaxel and $1584 lower for erlotinib (p < 0.05). Outpatient visits, laboratory procedures, and acute care visits were also less frequent with erlotinib relative to pemetrexed (−2.6 PPPM, p < 0.05). Erlotinib and docetaxel were associated with a statistically significant lower costs and resource use relative to pemetrexed.

Health-related quality-of-life studies including the EQ-5D instrument

Eight studies were retrieved that used the EQ-5D instrument to measure HRQoL of advanced NSCLC patients (Supplementary Appendix 2). In three of these studies89–91 the general public in the UK was asked to complete the EQ-VAS instrument, and in addition standard gamble interviews where ratings were done on NSCLC health state descriptions were conducted in two of these studies89,91. The remaining five studies either reported utility values for patients on specific treatment regimens included in clinical trials or for patients treated in routine clinical practice (using the EQ-5D and EQ-VAS). The two studies measuring HRQoL on patients included in clinical trials either described chemonaïve stage IIIb/IV NSCLC patients from various countries receiving first line docetaxel/platinum vs first line vinorelbine/cisplatin92, or Dutch stage IIIa/IIIb/IV NSCLC patients ineligible for chemotherapy and that were administered two different radiotherapy schedules93, respectively. The three studies measuring HRQoL on patients treated in routine clinical practice either described chemonaïve stage IIIb/IV NSCLC patients on first and second line chemotherapy from various countries10, Italian NSCLC patients without any restriction on stage or previous therapy94, or stage IIIa/b/IV NSCLC German and Austrian patients on second line pemetrexed treatment95, respectively. The study by Trippoli et al.94 was the only patient-oriented study that reported on the impact of clinical variables on HRQoL by describing utility values for different health states. While the international study by Bischoff et al.10 aimed at evaluating patients at several time points from baseline (at initiation of first line chemotherapy) until death, only baseline utility values were presented, and, thus, no conclusions could be drawn about the impact of various disease health states and treatments on QoL. Although this study collected QoL data for patients from five countries including Finland, Germany, Portugal, the Netherlands, and the UK, country-specific results were not presented. The study by Schuette et al.95 presented utility values at baseline and after treatment cycle 2 and 6, but was restricted to second line pemetrexed treatment. Of the two studies reporting on utility values for unselected patients treated in routine clinical practice, only the study by Bischoff et al.10 described the type of agents used, however this was limited to the type of agents that were used during first line therapy.

Similar values for the stable disease health state (0.626 and 0.653, respectively)89,91 were reported in two out of the three studies where utility values were estimated from the general public. The third study reported lower values for the same stable disease health state (0.426–0.451)90. The three studies that explored utility values for patients treated in routine clinical practice and that included either chemonaïve advanced patients10, any NSCLC patient94, or patients treated with second line pemetrexed, presented similar baseline utility values that were not specific to the stable disease health state (values ranging from 0.6–0.7, 0.50–0.68, and 0.66, respectively) (Supplementary Appendix 2). Findings from three of the studies show that disease progression, the presence of metastases, and severe symptoms all have a negative impact on QoL89,91,94, with utility values being as low as 0.461 when patients experience cough, dyspnoea, and pain89 and 0.473 for patients experiencing progressive disease91. Second-line pemetrexed treatment was shown to have a positive effect on QoL, increasing utility values by 0.02 and 0.11 after two and six cycles of therapy, respectively95 (see Supplementary Appendix 2).

Databases and registries

In total, 181 cancer registries and five cancer databases were found to include relevant information on lung cancer in the countries studied (see Supplementary Appendix 3). Of these cancer registries or cancer databases, 117 are regional epidemiological population-based cancer registries. Nine out of 10 countries have various types of national organizations compiling statistics on incidence and mortality from these regional epidemiological cancer registries (Table 5). All of the epidemiological population-based registries present data on lung cancer at the level of cancer site using the ICD-O-3 coding system (International Classification of Diseases for Oncology, Third Edition). In Italy and the Netherlands, data is further presented on the level of the invasiveness of the lung cancer (invasive or in situ). In the Netherlands, it is also possible to retrieve data by lung cancer type including NSCLC. In all countries studied with the exception of France, Italy, and Turkey, cancer registration with these registries is mandatory on a national level. In France and Italy, national cancer registration coverage on a population level was found to be 18% and 38%, respectively96,97. In Turkey, the Izmir Cancer Registry, which was the first population-based cancer registry to be organized in Turkey, is considered to be the only well-defined cancer registry which aims at accessing the data of all cancer patients within the Izmir geographical area (∼5% of the Turkish population)98. The population-based epidemiological cancer registries were found to collect differing amounts of information pertinent to the objectives of this study. Resource use and HRQoL data is not collected by any registry. No registry collects information on all treatment options a patient receives, and only 60 of all 117 registries collect the initial choice of treatment. In terms of treatment outcome, all regional registries within all countries, with the exception of the Turkish registries, collect data on the date of death of the patients. Detailed information on treatment outcome by type of treatment and by tumour response using e.g., the Response Evaluation Criteria in Solid Tumours (RECIST) criteria was not identified in any of the registries. An outline of variables recorded by population-based epidemiological cancer registries in the different countries is presented in Table 6. Differences in the type of data collected in registries existed both within and among countries. Additional variables, (e.g., occupation, basis of diagnosis, number of primary tumours, previous cancer diagnosis, participation in clinical trials, number of births for women, and waiting times) are collected in the registries, which were not part of the investigation of this study and are therefore not presented in Table 6.

Table 5.  Cancer registries and databases.

Country	Number of regional population-based epidemiological cancer registries identified	National cancer registry or association that collects information from regional registries and compile incidence and mortality statistics from these data	Number of clinical cancer registries collecting more in depth information on the treatment history of the patients**	Patient databases or linkage systems that enables retrieval of information of patient disease history***	DRG/HRG-based hospital/inpatient care registries
Australia	8	National Cancer Statistics Clearing House	3^a	Yes^c	Yes
Belgium	4	Belgian Cancer Registry			Yes
Canada	14	Canadian Cancer Registry (CCR)		Yes^d	Yes
France	14	FRANCIM at Institut de Veille Sanitaire*			Yes
Germany	11	Association of Epidemiological Cancer Registries in Germany (GEKID)/GEKID-Atlas	60^b		Yes
Italy	29	Associazione Italiana Registri Tumori (AIRTUM)/L’istituto superiore di sanità (ISS)*			Yes
The Netherlands	8	Dutch Cancer Registry (NKR)/NKR Cijfers		Yes^e	Yes
Sweden	6	The Swedish National Cancer Registry		Yes^f	Yes
Turkey	12	–			Yes
UK	11	The National Cancer Intelligence Network (NCIN)/Cancer e-Atlas		Yes^g	Yes

* Nation-wide coverage in France; 18%, Italy; 38%, respectively. All other countries with the exception of Turkey have national registries that aim at having a complete nation-wide coverage of cancer cases.

** In comparison to the population-based epidemiological cancer registries that collect a restricted amount of treatment data for each cancer patient, these clinical registries collect data on the complete treatment history of the patient.

*** Databases or linkage systems that interrogate multiple data sources based on personal health numbers and can provide demographic-, drug prescription and usage-, clinical laboratory-, pathology- information and diagnostic codes for outpatient visits and hospital stays.

^aACCORD (The Australian Comprehensive Cancer and Research Database, a database system installed at several hospitals in different states of Australia, NSW Clinical Cancer Registry Pilot Project, Queensland Oncology Online (QOOL) including OASys (Oncology Analysis System).

^b60 clinical cancer registries located at regional tumour centres across Germany (majority of centres use the GTDS database system).

^cSA-NT DataLink (South Australia and the Northern Territory).

^dSaskatchewan Health Database.

^ePHARMO record linkage system.

^fSwedish National Board of Health and Welfare.

^gGPRD (the General Practice Research Database).

Table 6.  Variables recorded by population-based epidemiological cancer registries.

Type of data	Variable collected	Country
		Australia	Belgium	Canada	France	Germany	Italy	The Netherlands	Sweden	Turkey	UK
Demographical data	Patient identification information*	x (8)	x (4)	x (14)	x (14)	x (11)	x (29)	x (8)	x (6)	x (12)	x (11)
	Date of death	x (8)	x (4)	x (14)	x (14)	x (11)	x (29)	x (8)	x (6)	x (1)	x (11)
	Cause of death	x (8)	x (4)	x (13)		x (10)	x (26)		x (6)		x (11)
	Smoking history					x (2)			x (6)		x (8)
	Treating doctor/source of referral/treating hospital	x (6)	x (4)	x (14)	x (14)	x (7)	x (5)	x (8)	x (6)	x (1)	x (11)
Clinical information	Performance status at diagnosis		x (4)			x (1)			x (6)		x (8)
	NSCLC histology/morphology	x (8)	x (4)	x (14)	x (14)	x (11)	x (29)	x (8)	x (6)	x (12)	x (11)
	Disease stage	x (8)	x (4)	x (14)	x (14)	x (11)	x (4)	x (8)	x (6)	x (12)	x (11)
	Comorbidities					x (1)		x (1)			x (8)
	Recurrence date**		x (4)	x (1)	x (1)	x (2)
Treatment	Planned initial treatment		x (4)	x (2)	x (14)	x (11)	x (3)	x (8)	x (6)	x (1)	x (11)

Within brackets is the number of registries in that country that collect the specified variable.

*Including, e.g., name, sex, residential address, origin, date of birth (in the Dutch registries, the patient name is not collected).

**Date of diagnosis of residual disease.

While the majority of registries in all 10 countries collect basic information on treating doctor, source of referral, treating hospital, lung cancer diagnosis, and death, the availability of registries containing more detailed information including type of treatments prescribed, treatment outcomes, and smoking status is limited to clinical cancer registries in Australia, with data principally being collected using the Australian Comprehensive Cancer and Research Database (ACCORD) and in Germany using the Gießener Tumordokumentationssystem (GTDS) data collection systems, respectively. Data collection using the ACCORD system is still in its pilot stage.

A total of five patient databases or linkage systems that enable retrieval of information of patient disease and treatment history, treatment outcome, and resource utilization (although not complete) by combining data from various sources were identified in Australia, Canada, the Netherlands, Sweden, and the UK (Table 5). Data that could be linked to the demographic and clinical history of the patient include, e.g., mortality, drug dispensing, hospital inpatient care, clinical laboratory, pathology, and general practitioner information. An explanation of the data that can be linked using the linkage systems in the respective countries is available in Supplementary Appendix 3.

Discussion

The aim of this study was to review the existing literature for retrospective and non-randomized prospective studies and the information in registries and healthcare databases, addressing routine care treatment patterns, the outcome of treatment, and resource utilization for advanced NSCLC patients in different countries. The literature review revealed few published studies on treatment patterns, treatment outcomes, resource use, and quality-of-life in routine clinical practice. Most studies were found to be conducted in small study populations, selected patient groups, or only consider certain pre-specified treatment alternatives, not focusing on general treatment patterns in routine clinical practice. The EQ-5D and EQ VAS have been scarcely used in the reviewed NSCLC literature.

Only two international studies describing general treatment patterns and outcomes for patients with advanced NSCLC in routine clinical practice were retrieved. Patient enrolment in the ACTION study was, however, limited to 2003, and the study is therefore not expected to reflect current treatment patterns in NSCLC10. In addition, given the prospective nature of the study, patients were followed from initiation of first line therapy until death or for a maximum of 18 months, and the complete treatment history of all patients could therefore not be documented in all cases. Thus, overall survival data was censored at the time of last follow-up for patients who were still alive. Although patient enrolment in the SELECTTION study occurred later between the years of 2006 and 2008, at a time when targeted therapies had been approved for 2nd line treatment of advanced NSCLC, the study was restricted to reporting 2nd line treatment patterns and only followed patients for 12 months. Generally, most retrieved studies on treatment patterns collected patient data or were published at a time when no targeted therapies were approved for use in advanced NSCLC, suggesting that current treatment practices are not fully represented in these studies. Only one study was current enough to capture possible treatment patterns for this group of patients; however, it was based on a Delphi panel discussion rather than routine care patient information, and was restricted to the Spanish healthcare system13.

Most published analyses on resource utilization show that hospitalizations constitute a major cost driver in the treatment of NSCLC patients18,80,84,85,87,88,99. This information can be useful when evaluating the economic impact of new therapies in NSCLC, and would have a significant effect in the evaluation of therapies that reduce the episodes of hospitalizations, and thus also the total cost of NSCLC treatment.

The retrieved HRQoL studies that presented utility values for different health states were mainly restricted to those estimated by the general public. The only patient-oriented study reporting on utility values for different health states was not specific to advanced NSCLC patients and did not discriminate between patients on various lines of treatment or patients experiencing the most common adverse events pertinent to NSCLC patients94.

In the search for information in national registries and databases, it was found that most registries and databases do not contain enough information to provide a complete picture of patients’ treatment histories and outcomes. Few registries collect coherent treatment information from the date of diagnosis until death of patient. The availability of this more detailed information is restricted to a few regional clinical cancer registries and/or databases in Australia, Canada, Germany, the Netherlands, Sweden, and the UK. Resource use is also not measured comprehensively enough to draw conclusions regarding resource allocation in the treatment of lung cancer and, only in the Netherlands, data is collected by the type of lung cancer, including NSCLC. Although initiatives are being taken in several countries (e.g., Australia and the UK) to extend the mandatory dataset that is being collected, it is not expected to come to pass in the near future100. However, the increasing availability of computerized databases or linkage systems (e.g., SA-NT DataLink in Australia, the Saskatchewan Health Database in Canada, the PHARMO record linkage system in the Netherlands, National Board of Health and Welfare in Sweden, and the GPRD in the UK) that can link cancer registries with other sources of patient level data has enabled the capture of information on cancer patients that goes beyond the more limited cancer registry dataset, e.g., to include details of treatment and clinical status through hospital information systems101.

Conclusion

In conclusion, currently available data sources do not seem to provide sufficient routine clinical practice information to make evidence-based treatment decisions and to make the best possible decisions on the most cost-effective treatment alternatives in advanced NSCLC patients. Few studies on treatment patterns, treatment outcomes, resource use, and quality-of-life in routine clinical practice have been published. Moreover, most studies have been conducted in small study populations, selected patient groups, or have only considered certain pre-specified treatment alternatives, not focusing on general treatment patterns in routine clinical practice. Variability in the type of data collected in cancer registries limit cross-country comparisons. While awaiting more comprehensive information to become available through databases and patient level linkage systems, patient chart reviews on unselected patients treated in routine clinical practice could offer one option to complement data collected in patient cancer registries or databases and allow for cross-country comparisons. Cross-country comparison may not only allow identification of evidence-based best practice relating to the treatment outcome, but also help to reduce inequalities and demonstrate how the primary site of care (hospital or outpatient setting) may influence resource use and treatment decisions.

Transparency

Declaration of funding

Funding for this study was provided by Boehringer Ingelheim GmbH.

Declaration of financial/other relationships

H.W.F. is employed by Boehringer Ingelheim Pharmaceuticals. A.D.G. and J.E. have disclosed they are employed by Optum Insight Life Sciences, a company that received funding by Boehringer Ingelheim GmbH to conduct the research pertinent to this study.

The peer reviewers on this manuscript have disclosed that they have no relevant financial relationships.

Acknowledgements

No assistance in preparation of this article is to be declared.