Patterns of age-specific socioeconomic inequalities in net survival for common cancers in Taiwan, a country with universal health coverage

Cancer Epidemiology 53 (2018) 42–48

Contents lists available at ScienceDirect

Cancer Epidemiology journal homepage: www.elsevier.com/locate/canep

Patterns of age-specific socioeconomic inequalities in net survival for common cancers in Taiwan, a country with universal health coverage

T

Li-Hsin Chiena,1, Tzu-Jui Tsengb,1, Fang-Yu Tsaic, Jie-Huei Wangc, Chao A. Hsiunga, ⁎ ⁎ Tsang-Wu Liuc, , I-Shou Changa,b,c, a

Division of Biostatistics and Bioinformatics, Institute of Population Health Sciences, National Health Research Institutes, 35 Keyan Road, Zhunan, Miaoli County, 35053, Taiwan b Center of Biomedical Resources, National Health Research Institutes, 35 Keyan Road, Zhunan, Miaoli County, 35053, Taiwan c National Institute of Cancer Research, National Health Research Institutes, 35 Keyan Road, Zhunan, Miaoli County, 35053, Taiwan

A R T I C L E I N F O

A B S T R A C T

Keywords: Age at diagnosis Cancer in Taiwan Deprivation gap Net survival Socioeconomic inequalities Survival

Introduction: In high-income countries, advances in early diagnosis and treatment have improved cancer survival. However, socioeconomic inequalities in survival have persisted or increased for some adult cancers. Materials and methods: We assessed net survival for the 20 most common adult cancers in Taiwan. They were stratified into six age groups and three socioeconomic groups. Results: Out of 120 cancer site and age group combinations, 49 showed improvements in 5-year net survival from 2000–2004 to 2005–2010. Only cervix uteri cancer in the 35–49-year age group showed a deterioration. During 2000–2010, 13 of the 20 cancer cases experienced socioeconomic inequalities for all age groups combined, and the deprivation gaps varied with cancer site and age at diagnosis. For the five most common cancers – liver, colon and rectum, lung, breast, and oral – there were socioeconomic inequalities, and 5-year net survival improved for most or all of the six age groups from 2000–2004 to 2005–2010. Conclusion: Reducing socioeconomic inequality in survival may lead to improvements in survival overall. We should focus on the age groups with large deprivation gaps. Our results are useful for prioritizing cancer sites and age groups for in-depth socioeconomic disparity studies and for proposing interventions for health disparity reductions and net cancer survival improvements.

1. Introduction Socioeconomic inequalities in cancer survival have been observed extensively for more than two decades through studies of many cancers in numerous populations using various methods of determining socioeconomic status (SES) and survival events [1–16]. Although the overwhelming majority of these studies were performed in Western countries, a few were conducted in Asia [17,18]. The goals of these studies were to determine whether there are differences in survival for different socioeconomic groups. Differences in population-based cancer survival between SES groups as well as between countries or regions are increasingly being used to drive improvements in health services [19,20]. Because population-based cancer survival is a key measure of the overall effectiveness of the health system in managing cancer, it is important to examine trends of socioeconomic inequalities in terms of net survival [12,16].

Several studies have suggested that socioeconomic inequalities have persisted and even increased for many common cancers in adults, despite efforts to reduce them. Cancer survival improved significantly for almost all cancers, but patients from lower SES groups may not have benefited as much from advances in early diagnosis and treatment [9,11,12]. Furthermore, a recent study showed that the magnitude and patterns of age-specific socioeconomic inequalities in survival were different for patients diagnosed with lung, breast, and colon cancers. It has therefore been suggested that efforts should be made to ensure the availability of optimal treatment and appropriate management for lung cancer patients in all age groups, and for older patients in deprived groups with breast or colon cancer [21]. Furthermore, it was shown that identifying the age at diagnosis when socioeconomic inequalities in cancer survival are apparent may provide useful information for targeting cancer site-specific interventions and tailoring guidelines for patients at high risk.

⁎

Corresponding authors at: National Institute of Cancer Research, National Health Research Institutes, 35 Keyan Road, Zhunan, Miaoli County 35053, Taiwan E-mail addresses: [email protected] (L.-H. Chien), [email protected] (T.-J. Tseng), tatufi[email protected] (F.-Y. Tsai), [email protected] (J.-H. Wang), [email protected] (C.A. Hsiung), [email protected] (T.-W. Liu), [email protected] (I.-S. Chang). 1 These authors contributed equally to this work and are considered co-first authors. https://doi.org/10.1016/j.canep.2018.01.006 Received 6 June 2017; Received in revised form 4 December 2017; Accepted 10 January 2018 1877-7821/ © 2018 Elsevier Ltd. All rights reserved.

Cancer Epidemiology 53 (2018) 42–48

L.-H. Chien et al.

Net cancer survival is the probability of surviving the cancer under study in the absence of other causes of death. Estimates of net survival for a cancer provide useful measures for comparing cancer survival between diagnostic periods, ethnic groups, and registries [22,23]. Net survivals are studied in two frameworks: the relative survival (RS) framework, which does not require knowledge of the cause of death, and the cause-specific survival (CSS) framework, which does. Although estimates in the CSS framework require reliable information regarding the cause of death, which may not be accurately derived from the death certificate [24], for estimates in the RS framework the expected allcause survival of a comparable group of cancer-free individuals is often derived from life tables of a population, which may cause concern if SES is relevant. For example, in the United States, estimates in the RS framework are often computed using life tables of the general population; they are therefore overestimated for the high-SES group and underestimated for the low-SES group. This causes exaggeration of the SES gradient [25]. To deal with the estimation bias in the RS framework when SES is considered, Howlader and colleagues developed the Surveillance, Epidemiology, End Results program’s cause-specific death classification variable (SEERDCV), which utilizes information from the death certificate, the sequence of tumor occurrence, the site of the original cancer diagnosis, and comorbidities to assign the cause of death. They showed that estimates in the CSS framework utilizing SEERDCV provide more accurate estimates of net survival across the SES gradient by addressing differing rates of mortality resulting from other causes of death [25]. In this study, we report the age-group-specific and SES-specific net survival for the 20 most common cancers in Taiwan. We hope our study – which is the first in Taiwan to examine the relationship between age, SES, and net survival – will be useful for prioritizing cancer sites in more thorough and in-depth socioeconomic inequality studies that will eventually contribute to global cancer surveillance. Because lowering cancer mortality has become a priority of the Taiwan government [26], our findings may have a timely impact. To improve the reliability of our results, we also assessed the agreement between estimates in the RS and CSS frameworks based on data for those in Taiwan, and compared the levels of agreement based on data from patients in Taiwan with those based on SEER data.

National Health Insurance Research Database (NHIRD) during 2000–2012; these data have proved to be valuable resources for health science research [31–33]. The TCR was linked with the NHIRD and TCOD for the current study. For patients in the TCR, the TCOD provided survival information and the NHIRD provided SES information; we also used the NHIRD to confirm the survival information. This study was approved by the institutional review board of the National Health Research Institutes in Taiwan. The TCR, launched in 1979, is a population-based cancer registry that collects information about newly diagnosed cancer patients at all hospitals with more than 50 beds in Taiwan. The quality of the TCR has been improving and was recently reviewed [32,34]. More detailed description of data linkage and cleaning for the data used in this study is given in the Supplementary material. Cancer sites were included in this study if the number of patients aged between 50 and 64 years at diagnosis and the number of those aged 65 years or older were > 1000 during 1992–2004; these data were used for comparisons with data from patients in the United States [25]. In addition, these data were used to determine the 20 most common invasive cancer sites. This facilitated the evaluation of the performance of SEERDCV and the agreement of RS and CSS estimates in Taiwan. Taiwan’s National Health Insurance (NHI) program (implemented on March 1, 1995, by the Bureau of National Health Insurance, which is now the National Health Insurance Administration) provides compulsory universal health insurance and covers all healthcare services for more than 99% of Taiwan’s population [35]. The TCOD includes cause-of-death information for individuals in Taiwan since 1971. It is maintained by the Department of Statistics of the Ministry of Health and Welfare in Taiwan; its quality has been previously described [31]. Based on the national identification card number, sex, birth date, death date, and cause of death, there are 3,249,784 unique death records for the period between January 1, 1985 and December 31, 2011. The TCOD adopted the national identification card number in 1985. The original TCOD contains 3,255,505 records for this period; < 0.2% of the data were excluded during the data cleaning process.

2. Materials and methods

The SES of each cancer patient was defined according to insurable monthly income in the NHI registry archives during the year or month of diagnosis, which is the manner in which the NHI administration calculates beneficiary premiums [36]. The Registry for Beneficiaries in the NHIRD provides this information for those who registered in 2000 and after. If the insurable monthly income was unavailable at the time of diagnosis, then the insurable monthly income of the previous year or following year was used. According to the TCR, a total of 724,992 patients between 15 and 94 years of age were diagnosed with invasive cancers during 2000–2010; among them, 724,770 had survival information recorded in the TCOD and NHIRD. Among the patients having survival information, 723,810 had their SES assigned in this manner. The SES for 93% of these patients was based on the year or month of diagnosis. All patients were classified into one of the following three categories: low-income (monthly income less than Taiwan’s minimum wage level of 15,840 New Taiwan dollars [NTD]; n = 178,083); mediumincome (monthly income between 15,840 and 57,779 NTD; n = 500,280); and high-income (monthly income at or above the highest rank of insurable income, 57,800 NTD; n = 45,447) [36]. The exchange rate at the time of this study was approximately 1 United States dollar = 32 NTD. Subjects categorized as poor were classified by the local municipal authority as living below the local lowest living index and receiving social welfare subsidies. The poor category was small and was included in the low-income category. The number of patients included for the period 2000–2010 in this study by cancer site, age group, and SES are reported in Table 1. More

2.3. Socioeconomic status

2.1. Relative survival and cause-specific survival In this work, estimates of net survival in the RS framework are referred to as RS estimates. The RS estimates we considered were Ederer II estimates and Pohar Perme estimates (PPE) [27–29]. Estimates of net survival in the CSS framework that we used were the actuarial estimates, which are referred to as CSS estimates [30]. Because of the lack of SES-specific life tables in Taiwan, RS estimates using life tables for the general population may cause concern in a study of socioeconomic inequalities in cancer survival. However, because Taiwan’s National Health Insurance (NHI) program reduces financial barriers to health care, it is likely that SES-specific RS estimates may not be seriously biased. By using SEERDCV, we examined whether SES-specific RS and CSS estimates in Taiwan were in good agreement. Specifically, we compared RS and CSS estimates of net survival for the 20 most common cancer sites in Taiwan by age at diagnosis, SES, and periods of diagnosis. For each combination of cancer site and age group, we reported the differences between the net survival for high-, medium-, and low-SES groups in terms of CSS estimates. 2.2. Study population This study used data from the Taiwan Cancer Registry (TCR) during the period 1992–2010, the data from the Taiwan Cause of Death Database (TCOD) during 1992–2011, and the data from the Taiwan 43

Cancer Epidemiology 53 (2018) 42–48

L.-H. Chien et al.

Table 1 Number of patients included in this study by cancer site, age group, and socioeconomic status for the period 2000–2010. Cancer \ Age

15–34

SES

High SES/Medium SES/Low SES

All cancer Digestive systema Lymphomab Breast Cervix Uteri Colon and Rectum Corpus and Uterus, NOS Esophagus Kidney and Renal Pelvis Larynx Leukemia Liver and Intrahepatic Bile Duct Lung and Bronchus Non-Hodgkin Lymphoma Oral Cavity and Pharynx Other Biliary Other Non-Epithelial Skin Ovary Pancreas Prostate Stomach Thyroid Urinary Bladder

1168/21768/6463 165/3556/1168 100/1510/472 136/2557/593 24/650/270 66/1487/451 8/372/105 < 5/48/21 23/240/69 0/20/10 79/1532/490 67/1463/531 17/448/139 68/1008/318 97/2540/933 < 5/22/5 44/717/160 54/938/270 < 5/123/32 0/5/ < 5 27/435/133 180/2999/590 9/147/38

a b

35–49

11173/99924/24046 2767/25148/6734 309/2115/471 2851/22360/3871 266/5428/1357 1164/7633/1766 270/2578/487 69/2504/1014 261/1221/295 28/491/172 249/1724/405 1065/11091/2976 603/5121/1335 294/1963/424 1229/17573/5675 24/253/75 313/1699/333 279/2403/453 115/800/248 43/90/20 354/3120/730 820/4946/702 153/1086/244

50–64

65–74

75–84

85–94

13386/164101/40450 4982/63011/15121 359/3323/868 1752/19972/5430 198/5315/1312 2099/20870/5245 295/3837/1073 201/4785/1417 340/2925/651 87/1441/363 221/2239/508 1910/28325/6105 1275/16092/4082 349/3178/831 1101/19478/4690 94/1068/253 366/3443/764 215/2368/723 224/2576/701 565/2733/708 548/6455/1653 396/3691/733 333/3627/825

9087/116308/43365 3776/50975/17889 219/2262/980 618/4844/2533 202/2751/802 1522/18251/6847 82/740/379 127/2159/859 214/2483/921 72/979/433 157/1793/728 1466/21980/6479 1329/17919/6897 215/2190/952 341/5982/1807 77/1034/341 257/3977/1163 80/796/301 193/2416/1056 694/6876/3358 468/6169/2648 89/1058/384 263/3751/1453

8440/79423/50956 3548/33471/19922 224/1584/1296 334/1739/1042 157/1634/510 1542/13231/7826 31/221/161 87/1040/901 191/1483/1012 53/540/439 148/1513/1016 1162/11571/5491 1306/14178/9922 217/1534/1267 175/2149/1130 87/830/490 327/3959/1872 49/441/203 212/1983/1386 786/5987/6168 545/5646/4318 61/492/213 352/3096/2099

2193/18756/12803 922/7899/4997 43/349/293 57/321/277 47/421/159 400/3348/2088 10/39/25 36/243/223 41/317/202 11/78/90 39/347/275 210/2204/1085 341/3110/2532 42/341/283 38/366/236 20/208/146 124/1547/649 6/72/37 77/519/388 151/1246/1287 199/1585/1213 14/86/42 111/866/580

Including esophagus, stomach, colon and rectum, liver and intrahepatic bile duct, and pancreas. Including non-Hodgkin lymphoma and Hodgkin lymphoma.

than 3% and if their 95% confidence intervals (95%CIs) did not overlap. To compare net survival between two periods or two socioeconomic groups, we estimated their differences non-parametrically.

detailed information and those included for the periods 2000–2004 and 2005–2010 are reported in Table S0 (Supplementary material). Those included for the period 1992–2004 are found in Table S5A. The five most common cancers for the period 2000–2010 were of the liver (105,384 patients), colon and rectum (95,968 patients), lung (86,766 patients), breast (71,352 patients), and oral cavity and pharynx (65,624 patients). The sixth most common cancer was stomach cancer (36,302 patients).

2.5. Socioeconomic inequalities Since we consider three SES groups in this study, we denote by H–L the net survival in the high-SES group minus that in the low-SES group; by H–M that in the high-SES group minus that in the medium-SES group; M–L that in the medium-SES group minus that in the low-SES group. Based on these, we say socioeconomic inequality exists if all the 95%CIs of H–L, H–M, and M–L consist of positive numbers. For each cancer site and age group combination, we studied the existence of socioeconomic inequality based on the time period 2000–2010. We found we need data from longer time periods to look at whether these socioeconomic inequalities change over time.

2.4. Statistical analysis We used both Ederer II estimates, as described by Cho and colleagues [27], and PPE to obtain RS estimates [29]. The Taiwan life tables for the period 1992–2010 were obtained from the Human Mortality Database website (http://www.mortality.org/cgi-bin/hmd/country. php?cntr=TWN&level=1) [37]. Although it was reported that both PPE and Ederer II estimates have practical merit, and although there was little difference between these two estimates in Western countries [38–40], considering 5-year net survival, we took this opportunity to check their differences in Taiwan. We found only small differences between 5-year RS using the PPE and Ederer II estimates (Table S1, Supplementary material). Hereafter, RS estimates are referred to as Ederer II estimates unless otherwise specified. For individuals who had more than one instance of cancer, survival estimates were calculated only for the first cancer. Based on data in the TCR and TCOD, the SEERDCV was used to assign the cause of death to each individual in this study [25,41]. The CSS estimate was calculated using the actuarial method and the SEERDCV. Survival times were calculated in months and censored either at the date of death when the underlying cause was not the cancer under study or at the end of follow-up on December 31, 2011. Standard errors were calculated using Greenwood’s formula. The computer codes for Ederer II and CSS estimates are written in R and are included in the Supplementary material; PPEs were implemented with the R package “relsurv” [42]. Following the work of Howlader et al. [25], we regarded the differences between RS and CSS estimates to be “large” if they were more

3. Results We found that the effects of using SEERDCV in Taiwan were similar to those in the United States, and that RS and CSS estimates were in much better agreement in Taiwan than in the United States. Because the main purpose of this work was to determine SES inequalities, we present the results about the effects of using SEERDCV and the agreement between RS and CSS estimates in the Supplementary material, parts of which are briefly mentioned in subsection 3.4. 3.1. Improvement in net survival over time We report in Table 2 the increments in 5-year net survival from 2000–2004 to 2005–2010 for each cancer and age-group combination and their 95%CIs. These were based on CSS estimates. Results based on RS estimates are shown in Tables S5C. It follows from Table 2 that 49 cancer-site and age-group combinations out of 120 combinations showed improvements in 5-year net survival. Only cervix uteri cancer in the 35–49 age group showed deterioration in 5-year net survival; other combinations did not show any 44

Cancer Epidemiology 53 (2018) 42–48

L.-H. Chien et al.

Table 2 5-year net survival difference between the two periods 2000–2004 and 2005–2010 in cause-specific survival framework. Net Survival difference between 2 periods (2005–2010 and 2000–2004) % (95% CI) Age group

15–34

Cancer site

D-CSS(95%CI)

All cancer Digestive systema Lymphoma b Breast Cervix Uteri Colon and Rectum Corpus and Uterus, NOS Esophagus Kidney and Renal Pelvis Larynx Leukemia Liver and Intrahepatic Bile Duct Lung and Bronchus Non-Hodgkin Lymphoma Oral Cavity and Pharynx Other Biliary Other Non-Epithelial Skin Ovary Pancreas Prostate Stomach Thyroid Urinary Bladder a b c

c

5.12(4,6.2) 6.04(3,9) 4.22(0.4,8.1) 6.29(3.4,9.2) −0.91(−5.9,4.1) 7.24(2.6,11.9) 4.22(−2.7,11.2) 3.35(−19.2,25.9) −1.77(−11.5,8) −14.28(−45.9,17.4) 16.1(11.6,20.6) 4.22(−0.1,8.6) 3.7(−3.9,11.3) 4.91(−0.1,10) 4.44(1.3,7.6) NA 0.49(−2.6,3.6) −1.88(−5.8,2) −15.05(−32.7,2.6) NA 1.12(−7.7,9.9) −0.01(−0.7,0.6) −4.56(−14.6,5.5)

35–49

50–64

65–74

75–84

85–94

D-CSS(95%CI)

D-CSS(95%CI)

D-CSS(95%CI)

D-CSS(95%CI)

D-CSS(95%CI)

3.6(3,4.2) 3.31(2.2,4.4) 5.67(2.1,9.3) 2.8(1.9,3.7) −2.24(−4.3,−0.2) 3.08(1,5.2) −0.05(−2.4,2.3) 2.16(−0.5,4.8) 1.39(−2.8,5.6) 1.7(−6.5,9.9) 9.31(5,13.7) 4.32(2.8,5.9) 4.84(2.7,7) 5.87(2.1,9.6) 3.63(2.3,5) −7.34(−18.9,4.2) 0.33(−1.9,2.6) 1.21(−2.4,4.8) −0.43(−4.9,4) −0.85(−17,15.3) −1.33(−4.6,1.9) 0.56(0,1.1) −0.22(−4.1,3.7)

6.32(5.8,6.8) 6.02(5.3,6.8) 5.25(2.1,8.4) 3.27(2.2,4.3) 0.24(−1.9,2.4) 4.76(3.5,6.1) 2.98(0.6,5.4) 1(−1.1,3.1) 4.75(1.6,7.9) 3.52(−1.1,8.1) 8.04(4,12.1) 6.21(5.2,7.3) 6.91(5.7,8.2) 5.22(2,8.5) 3.42(2,4.8) 3.58(−2.1,9.2) 1.19(−0.5,2.9) 7.85(3.9,11.8) −0.21(−2.3,1.9) 5.36(2.3,8.4) 0.52(−1.8,2.8) 2.4(0.9,3.9) 2.08(−0.4,4.5)

5.95(5.4,6.5) 5.78(5,6.6) 8.56(4.8,12.3) 2.51(0.5,4.6) −0.3(−3.4,2.8) 6.05(4.7,7.4) 6.56(0.9,12.2) 2.76(−0.2,5.7) 3.04(−0.5,6.6) 5.87(0.6,11.1) 2.18(−2.1,6.4) 6.05(4.9,7.2) 5.24(4.3,6.2) 8.45(4.7,12) 3.94(1.4,6.4) 1.92(−3.4,7.3) 3.15(1.4,4.9) 3.8(−2.6,10.2) −0.16(−1.9,1.6) 3.67(1.9,5.5) 4.48(2.3,6.7) 9.9(5.2,14.6) 2.86(0.2,5.5)

4.35(3.8,4.9) 4.33(3.4,5.2) 5.38(1.5,9.2) 4.72(0.8,8.6) 1.2(−3.4,5.8) 4.51(3,6) 1.92(−9.2,13) 0.45(−2.8,3.7) 3.09(−1.3,7.5) 4.93(−2.1,12) −1.07(−5.4,3.2) 4.75(3.5,6) 1.46(0.7,2.3) 5.45(1.6,9.3) 9.71(5.9,13.6) 2.33(−2.4,7) 2.55(0.6,4.5) 2.55(−5.5,10.6) 0.65(−0.8,2.1) 4.33(2.4,6.3) 2.67(0.7,4.7) 9.7(1.8,17.6) 4.42(1.4,7.5)

2.39(1.2,3.6) 3.57(1.9,5.3) −0.48(−8.2,7.3) 0.32(−9.9,10.5) 5.8(−3.2,14.8) 4.3(1.2,7.4) −26.42(−57.7,4.9) 4.4(−1.5,10.3) −1.72(−11.4,7.9) −0.15(−20,19.7) 0.95(−8.3,10.2) 1.64(−0.6,3.9) −0.25(−1.7,1.2) 0.07(−7.8,7.9) 7.05(−2.1,16.2) 3.4(−3.1,9.9) 3.69(−0.4,7.8) 7.5(−8.5,23.5) 0.58(−1.9,3.1) 3.38(−1.8,8.6) 4.51(1.3,7.7) −4.75(−23.4,13.9) 1.3(−4.9,7.5)

Including esophagus, stomach, colon and rectum, liver and intrahepatic bile duct, and pancreas. Including non-Hodgkin lymphoma and Hodgkin lymphoma. “NA”: not available due to a small sample size. Bold when the CIs do not contain 0.

groups 35–49 and 50–64. With smaller magnitudes, deprivation gaps also appeared for the age groups 35–49 and 50–64 for breast cancer. More detailed information can be found in Tables S3A and B.

significant difference. Here improvements and deteriorations are based on the 95%CIs. Improvements occurred most often in the age groups 50–64 and 65–74, followed by the age group 75–84. For cancer of all sites and for cancer of the digestive system, improvements in 5-year survival occurred for every age group. For the five most common cancers, 5-year net survival improved for most or all age groups.

3.4. Agreement between RS and CSS The percentages of Taiwan patients experiencing cancer-specific death, using SEERDCV, are reported in Table S2, and we discuss the effects of using SEERDCV in Supporting Methods and Additional Results. Table S3 presents the RS and CSS estimates of net survival for each age- and SES-group combination and related inequalities. Using both RS and CSS frameworks, SES-specific estimates of net survival and related inequalities for all age groups combined are presented in Table S4. Age-group-specific estimates of net survival without considering SES are presented in Table S5. A comparison of Table 2 in the work by Howlader and colleagues [25] with our Table S5A indicates that RS and CSS estimates were in much better agreement in Taiwan than in the US.

3.2. Existence of socioeconomic inequalities For each of the 20 cancers, in Table 3 we present 5-year net survival according to SES and the corresponding socioeconomic inequalities for the age group 15–94 and the period 2000–2010. These were based on CSS estimates. Results based on RS estimates are given in Table S4A. Table S4B and S4C present similar results for the periods 2000–2004 and 2005–2010 respectively. It follows from Table 3 that SES inequalities existed for 13 of the 20 cancers. These were breast, colon and rectum, esophageal, kidney and renal pelvis, laryngeal, liver, lung, oral, other non-epithelial skin, stomach, and thyroid cancers, leukemia, and non-Hodgkin lymphoma.

4. Discussion

3.3. Age-group-specific deprivation gap

Among the 20 most common cancers in Taiwan, we found that 49 cancer-site and age-group combinations (out of 120 combinations) showed improvements between 2000–2004 and 2005–2010 in 5-year cause-specific survival (Table 2). We also found that 13 of these 20 cancers experienced socioeconomic inequalities during 2000–2010 for all age groups combined (Table 3), and that the magnitudes of the gaps varied with both cancer site and age at diagnosis (Table 4). For the five most common cancers – liver, colon and rectum, lung, breast, and oral – SES inequalities existed for all of them (Table 3); between 2000–2004 and 2005–2010, 5-year net survival improved in most or all of the six age groups (Table 2). Therefore, for these cancers, although net survival improved, deprived patients may not benefit equally. These findings are in agreement with those in the literature [9,11,12]. However, it is important to examine whether these observations extend to other cancers when more data from later years become available. Indeed, data from

In view of the fact that deprivation gaps existed for the majority of the 20 cancers for all age groups combined for the period 2000–2010, we now report the age groups in which they continue to exist. Because SES inequality did not appear for any cancer for the age groups 75–84 and 85–94, we considered only age groups younger than 74 years in Table 4, which reports the H–L for 5-year net survival and its 95%CI for each cancer and age-group combination. When all H–L, H–M, and M–L and their 95%CIs consisted of positive numbers, we mark the interval with an asterisk. It follows from Table 4 that SES inequalities existed for the following: liver and oral cancers, with all gaps being nearly > 20% for all age groups younger than 64; colon and rectum, esophageal, kidney and pelvis, and stomach cancers, with all gaps being > 10% for the age 45

Cancer Epidemiology 53 (2018) 42–48

L.-H. Chien et al.

Table 3 5-year net survival by socioeconomic status and corresponding inequalities for the 20 most common cancers in Taiwan, in cause-specific framework for the period 2000–2010 and age group 15–94. Net Survival difference between SES groups% (95% CI)

Cancer site All cancer Digestive systema Lymphomab Breast Cervix Uteri Colon and Rectum Corpus and Uterus, NOS Esophagus Kidney and Renal Pelvis Larynx Leukemia Liver and Intrahepatic Bile Duct Lung and Bronchus Non-Hodgkin Lymphoma Oral Cavity and Pharynx Other Biliary Other Non-Epithelial Skin Ovary Pancreas Prostate Stomach Thyroid Urinary Bladder a b c

High SES

Medium SES

Low SES

H-L

CSS(95%CI)

CSS(95%CI)

CSS(95%CI)

D-CSS(95%CI)

59.22 (58.71,59.7) 45.64 (44.77,46.5) 65.64 (62.71,68.4) 89.69 (88.74,90.6) 72.72 (69.45,75.7) 63.19 (61.87,64.5) 87.22 (84.26,89.7) 21.82 (17.82,26.1) 74.03 (71.02,76.8) 79.73 (73.25,84.8) 46.38 (42.65,50) 33.56 (32.19,34.9) 20.37 (19,21.8) 64.16 (61.11,67) 73.39 (71.58,75.1) 24.94 (19.46,30.8) 93.54 (91.94,94.8) 66.82 (62.6,70.7) 9.84 (7.58,12.4) 80.06 (77.99,82) 42.86 (40.54,45.1) 96.55 (95.43,97.4) 70.93 (67.95,73.7)

50.85 (50.7,51) 36.46 (36.22,36.7) 57.24 (56.24,58.2) 85.23 (84.87,85.6) 77.4 (76.71,78.1) 58.68 (58.25,59.1) 84.65 (83.74,85.5) 15.21 (14.43,16) 63 (61.89,64.1) 66.43 (64.65,68.1) 40.82 (39.68,41.9) 23.69 (23.35,24) 12.94 (12.61,13.3) 55.19 (54.14,56.2) 60.46 (59.97,60.9) 27.15 (25.44,28.9) 90.64 (90.1,91.1) 64.68 (63.42,65.9) 6.69 (6.09,7.3) 71.62 (70.8,72.4) 36.7 (36.03,37.4) 94.12 (93.68,94.5) 69.59 (68.69,70.5)

c

14.67(14.1,15.24) 12.74(11.78,13.7) 16.23(12.94,19.52) 9.92(8.72,11.12) 4.37(0.91,7.83) 9.39(7.9,10.88) 8.16(4.89,11.43) 9.65(5.35,13.95) 15.69(12.24,19.14) 16.73(10.36,23.1) 10.84(6.7,14.98) 14.19(12.69,15.69) 9.21(7.74,10.68) 16.57(13.15,19.99) 23.08(21.09,25.07) 5.29(−0.96,11.54) 5.12(3.36,6.88) 8.93(4.21,13.65) 4.18(1.59,6.77) 3.46(1.28,5.64) 11.2(8.69,13.71) 6.85(5.27,8.43) 5.03(1.82,8.24)

44.55 (44.29,44.8) 32.9 (32.49,33.3) 49.41 (47.77,51) 79.77 (78.97,80.5) 68.35 (66.85,69.8) 53.8 (53.08,54.5) 79.06 (77.12,80.9) 12.17 (11.05,13.3) 58.34 (56.4,60.2) 63 (60.18,65.7) 35.54 (33.67,37.4) 19.37 (18.76,20) 11.16 (10.69,11.6) 47.59 (45.89,49.3) 50.31 (49.39,51.2) 19.65 (17.14,22.3) 88.42 (87.36,89.4) 57.89 (55.41,60.3) 5.66 (4.83,6.6) 76.6 (75.68,77.5) 31.66 (30.66,32.6) 89.7 (88.38,90.9) 65.9 (64.45,67.3)

H-M

M-L

D-CSS(95%CI)

D-CSS(95%CI)

8.37(7.84,8.9) 9.18(8.28,10.08) 8.4(5.38,11.42) 4.46(3.48,5.44) −4.68(−7.89,−1.47) 4.51(3.14,5.88) 2.57(−0.26,5.4) 6.61(2.39,10.83) 11.03(7.95,14.11) 13.3(7.3,19.3) 5.56(1.7,9.42) 9.87(8.46,11.28) 7.43(6,8.86) 8.97(5.82,12.12) 12.93(11.1,14.76) −2.21(−8.16,3.74) 2.9(1.37,4.43) 2.14(−2.09,6.37) 3.15(0.63,5.67) 8.44(6.29,10.59) 6.16(3.75,8.57) 2.43(1.37,3.49) 1.34(−1.66,4.34)

6.3(6,6.6) 3.56(3.08,4.04) 7.83(5.92,9.74) 5.46(4.6,6.32) 9.05(7.42,10.68) 4.88(4.04,5.72) 5.59(3.53,7.65) 3.04(1.65,4.43) 4.66(2.45,6.87) 3.43(0.17,6.69) 5.28(3.09,7.47) 4.32(3.62,5.02) 1.78(1.2,2.36) 7.6(5.61,9.59) 10.15(9.12,11.18) 7.5(4.4,10.6) 2.22(1.08,3.36) 6.79(4.05,9.53) 1.03(−0.04,2.1) −4.98(−6.2,−3.76) 5.04(3.84,6.24) 4.42(3.1,5.74) 3.69(2.01,5.37)

Including esophagus, stomach, colon and rectum, liver and intrahepatic bile duct, and pancreas. Including non-Hodgkin lymphoma and Hodgkin lymphoma. Bold when the CIs do not contain 0.

Table 4 Socioeconomic inequalities for 5-year net survivals by age group for the 20 most common cancers in Taiwan for the period 2000–2010, in cause-specific framework. Net Survival difference between SES groups% (95% CI) Cancer site/Age group

15–34

35–49

H-L(95%CI) All cancer Digestive systema Lymphomab Breast Cervix Uteri Colon and Rectum Corpus and Uterus, NOS Esophagus Kidney and Renal Pelvis Larynx Leukemia Liver and Intrahepatic Bile Duct Lung and Bronchus Non-Hodgkin Lymphoma Oral Cavity and Pharynx Other Biliary Other Non-Epithelial Skin Ovary Pancreas Prostate Stomach Thyroid Urinary Bladder *

50–64

H-L(95%CI) *

13.97(11.16,16.78) 20.25(11.71,28.79)* 13.97(6.45,21.49)* 6(−1.07,13.07) 8.95(−6.21,24.11) 16.23(2.83,29.63) −8.44(−39.48,22.6) NA 11.91(−7.53,31.35) NA 1.92(−11.14,14.98) 28.93(16.02,41.84)* 20.53(−6.28,47.34) 15.89(5.43,26.35) 22.78(14.45,31.11)* NA 8.15(1.45,14.85) 9.67(1.67,17.67) 55.26(37.22,73.3) NA −0.53(−22.31,21.25) 1.13(0.22,2.04) 11.52(−15.79,38.83)

65–74

H-L(95%CI) *

24.46(23.36,25.56) 27.6(25.26,29.94)* 18.25(11.51,24.99)* 13.81(12.04,15.58)* 14.11(9.09,19.13) 17.16(13.33,20.99)* 6.73(1.68,11.78) 28.19(15.32,41.06)* 17.63(10.34,24.92)* 33.74(18.72,48.76) 14.02(5.43,22.61) 27.49(23.9,31.08)* 17.94(12.89,22.99)* 20.1(12.99,27.21)* 30.57(27.76,33.38)* 20.49(−6.55,47.53) 7.84(3.3,12.38) 6.53(−0.92,13.98) 8(−1.24,17.24) 18.04(−8.31,44.39) 14.39(7.73,21.05)* 1.09(0.05,2.13) 14.58(7.24,21.92)

H-L(95%CI) *

16.19(15.13,17.25) 18.26(16.46,20.06)* 13.78(7.59,19.97) 7.88(5.76,10)* 4.25(−2.93,11.43) 10.88(8.15,13.61)* 8.51(3.85,13.17)* 13.46(6.14,20.78)* 19.18(13.35,25.01)* 20.25(10.33,30.17)* 15.45(6.59,24.31) 19.87(17.09,22.65)* 17.22(13.69,20.75) 13.14(6.82,19.46) 23.2(19.84,26.56)* 17.88(5.22,30.54) 6.09(3,9.18) 9.32(0.49,18.15) 6.57(0.22,12.92) 13.36(8.76,17.96) 17.58(12.37,22.79)* 5.19(2.13,8.25) 10.14(4.58,15.7)

4.93(3.66,6.2) 6.35(4.41,8.29) 5.53(−2.65,13.71) 2.26(−1.48,6) −2.67(−10.27,4.93) 5.07(2.06,8.08) 3.66(−7.52,14.84) 3.99(−3.89,11.87) 1.27(−6.7,9.24) 7.02(−5.7,19.74) 2.66(−6.86,12.18) 6.91(3.99,9.83) 4.26(1.6,6.92) 4.91(−3.35,13.17) 5.88(−0.57,12.33) −11.54(−22.29,−0.79) 4.99(1.45,8.53) 3.05(−10.69,16.79) 1.79(−2.52,6.1) −1.51(−5.23,2.21) 7.57(2.17,12.97) 13.44(5.27,21.61) 1.26(−5.31,7.83)

(1) Bold when the 95%CIs of H-L does not contain 0; (2) bold with * when the H-L, H-M and M-L are all positive numbers and their CIs do not contain 0. a Including esophagus, stomach, colon and rectum, liver and intrahepatic bile duct, and pancreas. b Including non-Hodgkin lymphoma and Hodgkin lymphoma. “NA”: not available due to a small sample size.

46

Cancer Epidemiology 53 (2018) 42–48

L.-H. Chien et al.

treatments and appropriate management for lung cancer patients. With more in-depth studies, we hope to provide suitable suggestions regarding socioeconomic inequalities in cancer survival in Taiwan.

later years are needed to evaluate more systematically the trends in the deprivation gap [11]. These findings suggest that reducing socioeconomic inequalities in cancer survival may lead to improvements in cancer survival overall, and that we should focus particular attention on age groups with large deprivation gaps. Our results are useful for prioritizing cancer sites and age groups for in-depth socioeconomic disparity studies and for proposing interventions for reductions in health disparity and improvements in net cancer survival. Those cancer-site and age-group combinations that showed the largest gaps deserve further study, especially those involving the five most common cancers. The possible underlying causes of SES inequality in cancer survival include factors related to tumors, patients, and the healthcare system. Because differences between SES groups in the stage of disease at diagnosis and in access to optimal treatment are known to have important roles in cancer survival disparity [1,11,43], the task of highest priority may be to clarify the relevance of each type of difference for the cancersite and age-group combinations that showed large SES inequalities in Taiwan. Because Taiwan’s NHI program provides comprehensive universal health coverage, researchers should investigate the relationship between the ways in which patients seek and obtain access to health services and the socioeconomic differences in cancer survival. Despite the NHI program, access to specialized cancer services varies among regions. Therefore, in addition to the SES effect, it is necessary to study regional effects on cancer survival [17]. We found that deprivation gaps were less prevalent among older people (Tables Table 4 and S3). This might be because certain therapeutic treatments may not be suitable for older people, and because NHI insurable income may not represent SES for certain retired people. Taiwan’s NHI program allows retired people to pay a lower premium if they are not dependents of their spouse or children. Although RS and CSS are in excellent agreement in general in this paper, and the above explanations are plausible for older people, we would like to point out a limitation of this study in this connection. It is well known that CSS may suffer from biases due to competing risks and misclassification of cause of death, which are common and serious among older people. SEERDCV helps reduce the misclassification of cancer-specific death, but it doesn’t deal with ‘cancer-consequent death’[44]. Because relative survival avoids these concerns, it is considered a method of choice when suitable life tables are available, and hence it is worth the effort to construct suitable life tables. Two good examples are the life tables specific for socioeconomic category in the UK [8,9] and the ethnic-specific life tables in New Zealand [45]. In addition to studying the net survival for lung cancer patients, the latter was also used to assess the bias in relative survival when using incorrect life tables. Excellent reviews of and comments on the choice between relative survival and cause-specific survival have appeared in the literature [44,46,47]. Since the RS in this study made use of general-population life tables, it is desirable to construct SES-specific life tables using Taiwan NHIRD linked to TCOD, which cover more than 99% of the Taiwan population, and use them in relative survival estimation and socioeconomic inequalities. Research in this direction is in hand. The definitions of high, medium, and low SES used in this study reflect the cost of living in Taiwan, with the high-SES group accounting for approximately 6.5% of the population. Researchers should explore whether there are more appropriate SES definitions for investigating socioeconomic inequalities in cancer survival. Nur and colleagues reported the impact of age at diagnosis on socioeconomic inequalities in survival for breast, lung, and colon cancers [21]. They found the following: for lung cancer, the deprivation gap narrows with age; for breast cancer, it widens with age; and for colon cancer, it has an intermediate pattern. Our findings confirmed their results for lung cancer, and our results regarding colorectal cancer were in line with their results for colon cancer (Table S3A). They noted that the reduction of socioeconomic inequalities in lung cancer survival requires action to extend to all age groups the availability of optimal

Author contribution Conceptualization, I-Shou Chang, Tsang-Wu Liu, Chao A. Hsiung, Li-Hsin Chien, Tzu-Jui Tseng; Methodology, I-Shou Chang, Li-Hsin Chien, Tzu-Jui Tseng, Jie-Huei Wang; Formal Analysis, Li-Hsin Chien, Fang-Yu Tsai; Investigator, I-Shou Chang, Chao A. Hsiung, Tsang-Wu Liu; Writing-Original Draft, I-Shou Chang, Tzu-Jui Tseng, Li-Hsin Chien; Writing-Review & Editing, I-Shou Chang, Tsang-Wu Liu, Chao A. Hsiung; Supervision, I-Shou Chang, Tsang-Wu Liu. Conflict of interest statement The authors declare that they have no competing interests. Acknowledgments The data analyzed in this study were provided by the Health and Welfare Data Science Center, MOHW, Taiwan. This study was supported by the MOHW (Project grants MOHW103-TDU-212-114001, MOHW104-TDU-B-212-124-008, MOHW105-TDU-B-212-134013, and MOHW106-0324-01-10-05), Ministry of Science and Technology (MOST 106-2319-B-400-001) and National Health Research Institutes (CA-105-PP-08). Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.canep.2018.01.006. References [1] L.M. Woods, B. Rachet, M.P. Coleman, Origins of socio-economic inequalities in cancer survival: a review, Ann. Oncol. 7 (1) (2006) 5–19. [2] M. Kogevinas, M. Porta, Socioeconomic differences in cancer survival: a review of the evidence, IARC Sci. Publ. 138 (1997) 177–206. [3] L.G. Shack, B. Rachet, D.H. Brewster, M.P. Coleman, Socioeconomic inequalities in cancer survival in Scotland 1986–2000, Br. J. Cancer 7 (7) (2007) 999–1004. [4] S.O. Dalton, L. Ross, M. During, K. Carlsen, P.B. Mortensen, J. Lynch, C. Johansen, Influence of socioeconomic factors on survival after breast cancer–a nationwide cohort study of women diagnosed with breast cancer in Denmark 1983–1999, Int. J. Cancer. Journal international du cancer 121 (11) (2007) 2524–2531. [5] X.Q. Yu, D.L. O'Connell, R.W. Gibberd, B.K. Armstrong, Assessing the impact of socio-economic status on cancer survival in New South Wales, Australia 1996–2001, Cancer Causes Control 19 (10) (2008) 1383–1390. [6] T.E. Byers, H.J. Wolf, K.R. Bauer, S. Bolick-Aldrich, V.W. Chen, J.L. Finch, J.P. Fulton, M.J. Schymura, T. Shen, S. Van Heest, X. Yin, G. Patterns of Care Study, The impact of socioeconomic status on survival after cancer in the United States: findings from the National Program of Cancer Registries Patterns of Care Study, Cancer 113 (3) (2008) 582–591. [7] C.M. Booth, G. Li, J. Zhang-Salomons, W.J. Mackillop, The impact of socioeconomic status on stage of cancer at diagnosis and survival: a population-based study in Ontario, Canada, Cancer 116 (17) (2010) 4160–4167. [8] M.P. Coleman, P. Babb, A. Sloggett, M. Quinn, B. De Stavola, Socioeconomic inequalities in cancer survival in England and Wales, Cancer 1 (1 Suppl) (2001) 208–216. [9] M.P. Coleman, B. Rachet, L.M. Woods, E. Mitry, M. Riga, N. Cooper, M.J. Quinn, H. Brenner, J. Esteve, Trends and socioeconomic inequalities in cancer survival in England and Wales up to 2001, Br. J. Cancer 0 (7) (2004) 1367–1373. [10] M.P. Coleman, M. Quaresma, F. Berrino, J.M. Lutz, R. De Angelis, R. Capocaccia, P. Baili, B. Rachet, G. Gatta, T. Hakulinen, A. Micheli, M. Sant, H.K. Weir, J.M. Elwood, H. Tsukuma, S. Koifman, E.S. GA, S. Francisci, M. Santaquilani, A. Verdecchia, H.H. Storm, J.L. Young, Cancer survival in five continents: a worldwide population-based study (CONCORD), Lancet Oncol. 9 (8) (2008) 730–756. [11] B. Rachet, L.M. Woods, E. Mitry, M. Riga, N. Cooper, M.J. Quinn, J. Steward, H. Brenner, J. Esteve, R. Sullivan, M.P. Coleman, Cancer survival in England and Wales at the end of the 20th century, Br. J. Cancer 99 (Suppl 1) (2008) S2–10. [12] B. Rachet, L. Ellis, C. Maringe, T. Chu, U. Nur, M. Quaresma, A. Shah, S. Walters, L. Woods, D. Forman, M.P. Coleman, Socioeconomic inequalities in cancer survival in England after the NHS cancer plan, Br. J. Cancer 03 (4) (2010) 446–453. [13] A. Pokhrel, P. Martikainen, E. Pukkala, M. Rautalahti, K. Seppa, T. Hakulinen,

47

Cancer Epidemiology 53 (2018) 42–48

L.-H. Chien et al.

[14]

[15]

[16]

[17]

[18]

[19]

[20] [21]

[22] [23]

[24] [25]

[26] [27]

[28]

[29] M.P. Perme, J. Stare, J. Esteve, On estimation in relative survival, Biometrics 8 (1) (2012) 113–120. [30] P.W. Dickman, E. Coviello, Estimating and modeling relative survival, Stata J. 5 (1) (2015) 186–215. [31] T.H. Lu, M.C. Lee, M.C. Chou, Accuracy of cause-of-death coding in Taiwan: types of miscoding and effects on mortality statistics, Int. J. Epidemiol. 9 (2) (2000) 336–343. [32] C.J. Chiang, Y.C. Chen, C.J. Chen, S.L. You, M.S. Lai, Cancer trends in Taiwan, Jpn. J. Clin. Oncol. 0 (10) (2010) 897–904. [33] A.W. Hsing, J.P. Ioannidis, Nationwide population science: lessons from the Taiwan national health insurance research database, JAMA Intern. Med. 75 (9) (2015) 1527–1529. [34] C.J. Chiang, S.L. You, C.J. Chen, Y.W. Yang, W.C. Lo, M.S. Lai, Quality assessment and improvement of nationwide cancer registration system in Taiwan: a review, Jpn. J. Clin. Oncol. 5 (3) (2015) 291–296. [35] National Health Insurance Administration, Ministry of Health and Welfare, 2014–2015 National Health Insurance Annual Report, Taiwan, Taipei, (2014). [36] C.C. Hsu, C.H. Lee, M.L. Wahlqvist, H.L. Huang, H.Y. Chang, L. Chen, S.F. Shih, S.J. Shin, W.C. Tsai, T. Chen, C.T. Huang, J.S. Cheng, Poverty increases type 2 diabetes incidence and inequality of care despite universal health coverage, Diabetes Care 5 (11) (2012) 2286–2292. [37] The Human Mortality database. http://www.mortality.org/cgi-bin/hmd/country. php?cntr=TWN&level=1. (Accessed May 20, 2015). [38] L. Roche, C. Danieli, A. Belot, P. Grosclaude, A.M. Bouvier, M. Velten, J. Iwaz, L. Remontet, N. Bossard, Cancer net survival on registry data: use of the new unbiased Pohar-Perme estimator and magnitude of the bias with the classical methods, Int. J. Cancer. Journal international du cancer 132 (10) (2013) 2359–2369. [39] V. Jooste, P. Grosclaude, L. Remontet, G. Launoy, I. Baldi, F. Molinie, P. Arveux, N. Bossard, A.M. Bouvier, M. Colonna, Unbiased estimates of long-term net survival of solid cancers in France, Int. J. Cancer. Journal international du cancer 132 (10) (2013) 2370–2377. [40] P.C. Lambert, P.W. Dickman, M.J. Rutherford, Comparison of different approaches to estimating age standardized net survival, BMC Med. Res. Methodol. 15 (2015) 64. [41] National Cancer Institute, Surveillance Research Program, SEER Cause-specific Death Classification. http://seer.cancer.gov/causespecific/. (Accessed Mar.18 2015). [42] M. Pohar, J. Stare, Relative survival analysis in R, Comput. Methods Programs Biomed. 1 (3) (2006) 272–278. [43] M. Orsini, B. Tretarre, J.P. Daures, F. Bessaoud, Individual socioeconomic status and breast cancer diagnostic stages: a French case-control study, Eur. J. Public Health 6 (3) (2016) 445–450. [44] D. Sarfati, T. Blakely, N. Pearce, Measuring cancer survival in populations: relative survival vs cancer-specific survival, Int. J. Epidemiol. 9 (2) (2010) 598–610. [45] T. Blakely, M. Soeberg, K. Carter, R. Costilla, J. Atkinson, D. Sarfati, Bias in relative survival methods when using incorrect life-tables: lung and bladder cancer by smoking status and ethnicity in New Zealand, Int. J. Cancer. Journal international du cancer 131 (6) (2012) E974–82. [46] B. Rachet, M.P. Coleman, Commentary: estimating cancer survival–which is the right approach? Int. J. Epidemiol. 39 (2) (2010) 611–612. [47] P. Baade, S. Cramb, P. Dasgupta, D. Youlden, Estimating cancer survival – improving accuracy and relevance, Aust. N. Z. J. Public Health 0 (5) (2016) 403–404.

Education, survival and avoidable deaths in cancer patients in Finland, Br. J. Cancer 103 (7) (2010) 1109–1114. X.L. Du, C.C. Lin, N.J. Johnson, S. Altekruse, Effects of individual-level socioeconomic factors on racial disparities in cancer treatment and survival: findings from the National Longitudinal Mortality Study 1979–2003, Cancer 117 (14) (2011) 3242–3251. L. Jansen, A. Eberle, K. Emrich, A. Gondos, B. Holleczek, H. Kajuter, W. Maier, A. Nennecke, R. Pritzkuleit, H. Brenner, G.C.S.W. Group, Socioeconomic deprivation and cancer survival in Germany: an ecological analysis in 200 districts in Germany, Int. J. Cancer. Journal international du cancer 134 (12) (2014) 2951–2960. M. Soeberg, T. Blakely, D. Sarfati, Trends in ethnic and socioeconomic inequalities in cancer survival, New Zealand, 1991–2004, Cancer Epidemiol. 39 (6) (2015) 860–862. C.M. Chang, Y.C. Su, N.S. Lai, K.Y. Huang, S.H. Chien, Y.H. Chang, W.C. Lian, T.W. Hsu, C.C. Lee, The combined effect of individual and neighborhood socioeconomic status on cancer survival rates, PLoS One 7 (8) (2012) e44325. Y. Ito, T. Nakaya, T. Nakayama, I. Miyashiro, A. Ioka, H. Tsukuma, B. Rachet, Socioeconomic inequalities in cancer survival: a population-based study of adult patients diagnosed in Osaka, Japan, during the period 1993–2004, Acta Oncol. 53 (10) (2014) 1423–1433. C. Allemani, H.K. Weir, H. Carreira, R. Harewood, D. Spika, X.S. Wang, F. Bannon, J.V. Ahn, C.J. Johnson, A. Bonaventure, R. Marcos-Gragera, C. Stiller, G. Azevedo e Silva, W.Q. Chen, O.J. Ogunbiyi, B. Rachet, M.J. Soeberg, H. You, T. Matsuda, M. Bielska-Lasota, H. Storm, T.C. Tucker, M.P. Coleman, Global surveillance of cancer survival 1995–2009: analysis of individual data for 25,676,887 patients from 279 population-based registries in 67 countries (CONCORD-2), Lancet (London, England) 385 (9972) (2015) 977–1010. M.P. Coleman, Cancer survival: global surveillance will stimulate health policy and improve equity, Lancet (London, England) 383 (9916) (2014) 564–573. U. Nur, G. Lyratzopoulos, B. Rachet, M.P. Coleman, The impact of age at diagnosis on socioeconomic inequalities in adult cancer survival in England, Cancer Epidemiol. 9 (4) (2015) 641–649. F. Ederer, L.M. Axtell, S.J. Cutler, The relative survival rate: a statistical methodology, Natl. Cancer Inst. Monogr. 6 (1961) 101–121. J. Esteve, E. Benhamou, M. Croasdale, L. Raymond, Relative survival and the estimation of net survival: elements for further discussion, Stat. Med. 9 (5) (1990) 529–538. C.B. Begg, D. Schrag, Attribution of deaths following cancer treatment, J. Natl. Cancer Inst. 4 (14) (2002) 1044–1045. N. Howlader, L.A. Ries, A.B. Mariotto, M.E. Reichman, J. Ruhl, K.A. Cronin, Improved estimates of cancer-specific survival rates from population-based data, J. Natl. Cancer Inst. 02 (20) (2010) 1584–1598. Y.-J. Ma, V. Siew, New century health policy (In chinese), Taiwan Med. J. 51 (2008) 92–93. H.N. Cho Hyunsoon, Angela B. Mariotto, Kathleen A. Cronin, Estimating Relative Survival for Cancer Patietns from the SEER Program Using Expected Rates Based on Ederer I Versus Ederer II Method. Surveillance Research Programs, NCI, Technical Report 2011–01, (2011). A.B. Mariotto, A.M. Noone, N. Howlader, H. Cho, G.E. Keel, J. Garshell, S. Woloshin, L.M. Schwartz, Cancer survival: an overview of measures, uses, and interpretation, Journal of the National Cancer Institute, Monographs 2014 (49) (2014) 145–186.

48