Gastric cancer survival (in)equity from 2002 to 2021: examining demographic and clinical characteristics among Māori and non-Māori

In New Zealand ethnic disparities in gastric cancer are stark, with incidence and death rates more than twice as high among Māori than non-Māori.^1–3 These ethnic inequities have been widely documented and investigated in terms of risk factors, presentation and management.^4–7 However, there has been minimal analysis of trends in survival inequities by ethnicity, and how these might vary by socio-economic factors. This information is crucial for monitoring inequity and assessing progress towards better outcomes.^8–10 In New Zealand, earlier studies indicate gastric cancer mortality decreased between 2001 and 2004, but ethnic and income group differences (higher excess mortality among Māori and lower income groups) remained.⁹ This aligns with global trends in gastric cancer incidence and mortality, with ethnic (and Indigenous) group differences likely partly explained by socio-economic factors, Heliobacter pylori (H. pylori) infection and barriers to care.^11,12

Clinical characteristics and socio-demographic factors have been theorised to at least partially explain inequitable outcomes.^13,14 Changes in survival could therefore partly be explained by changes in clinical characteristics over time. Although cancer grade and stage at diagnosis are broadly similar among Māori and non-Māori,^4,14 Māori tend to be younger at diagnosis and are more likely to be diagnosed with distal gastric cancer compared with non-Māori.⁴ A limitation with the New Zealand Cancer Registry (NZCR), however, is high levels of missingness of clinical characteristic data, particularly for cancer stage.^15,16 It is not known whether the level of missing clinical characteristic data for those with gastric cancer has improved over time.

Development of gastric cancer in the distal parts of the stomach is more likely to be associated with exposure to H. pylori, usually acquired during childhood due to household crowding.^5,6,17 Information on H. pylori infection history is not routinely collated nationally for the NZCR. However, H. pylori infection is more common in areas of higher socio-economic deprivation,^18–20 which may have a stronger link to gastric cancer survival than H. pylori infection (which should be more associated with incidence). Analysis of the 2001–2004 period revealed mortality differences between income bands among people diagnosed with gastric cancer in New Zealand.⁹ However, analysis of 1993–2003 registrations revealed no differences in relative survival rates for gastric cancer across area-level socio-economic deprivation quintiles.²¹ Relatedly, analysis of 2007–2016 registrations provided little evidence that gastric cancer–specific mortality hazard ratios (Māori relative to non-Māori) varied by area-level deprivation, but was highest in the lowest deprivation quintile.¹⁴

We provide information on gastric cancer diagnosis and survival over time among Māori and non-Māori in different periods during 2002–2021. We aimed to: a) describe the clinical characteristics of gastric cancer cases (site, extent of disease, grade, basis of diagnosis and missingness for these characteristics) for Māori and non-Māori over time; b) calculate gastric cancer mortality rates following a gastric cancer diagnosis (1 and 5 years post-diagnosis) for Māori and non-Māori, and examine these across 5-year diagnosis periods; and c) examine whether gastric cancer mortality rates for Māori and non-Māori, and the rate ratio (RR) for Māori:non-Māori, differed by demographic characteristics (age, sex, area-level deprivation and rurality) during 2002–2017/2021, and across 5-year diagnosis periods.

Methods

Study population

This study used a retrospective cohort of New Zealand residents diagnosed with gastric cancer from 2002 to 2021. All individuals with an International Classification of Disease 10th edition (ICD-10) primary diagnosis code indicative of gastric cancer (C160–C169) recorded in the NZCR between 1 January 2002 and 31 December 2021 were included in this study. The NZCR was accessed within Stats NZ Tatauranga Aotearoa’s (Stats NZ) Integrated Data Infrastructure (IDI), a large database containing linked, deidentified individual-level microdata about people and households in New Zealand.²² This allowed for NZCR records to be linked to other key data sources. We used the June 2023 IDI refresh for this study.

Data sources and variables

Gastric cancer records were linked to the Ministry of Health – Manatū Hauora Mortality Collection, including the full date of death and the ICD-10–coded underlying cause of death. This was used to determine all-cause mortality within 1 to 5 years of the date of diagnosis based on the presence and date of a death record. Gastric cancer mortality was coded only where gastric cancer (C160–C169) was recorded as the underlying cause of death. Because cause of death information was only available up to 2018 at the time of this study, gastric cancer mortality could only be examined among cases diagnosed up to 2013–2017 (depending on the survival period).

Sex, ethnicity and age at diagnosis were obtained from the NZCR. Up to three ethnicities per person are recorded in the NZCR, selected by an algorithm if present on at least 20% of a person’s records across the National Health Index (person-level health identifier) database, the Mortality Collection and the National Minimum Dataset of hospital discharges. The algorithm is periodically re-run to account for new data.²³ We coded a Māori/non-Māori indicator as well as total response ethnicity information by Level 1 ethnicities (Māori, Pacific, Asian, MELAA, European) among the Māori and non-Māori members of our gastric cancer cohort. Prior work has shown an undercount of Māori within health data,²⁴ but our comparison to a collated ethnicity data source in the IDI (available in the Appendix) suggested no relative undercount in the NZCR data.

Cancer site was coded to a four-level variable of cardia (C160), proximal (C161, C162), distal (C163, C164) and other (C165, C166, C168). Extent of disease describes the stage of development that the tumour has reached based on information obtained up to 4 months following the date of diagnosis, and is guided by the National Cancer Institute’s Surveillance, Epidemiology and End Results (SEER) summary staging system for all diagnoses from 1999.²⁵ Extent was coded as localised, regional, or distant. Tumour grade was classified according to the World Health Organization’s international Classification of Tumours and describes the appearance of tumour cells under the microscope.²⁶ Tumour grade was coded as well differentiated, moderately differentiated or poorly differentiated/undifferentiated. Finally, the source of diagnosis describes the single, most valid basis of diagnosis for a primary malignant tumour, and was categorised into three categories based on the level of evidence with three codes: 1) death certificate or clinical only (suspicion of cancer), 2) clinical investigation or exploratory surgery/autopsy (some evidence of cancer), and 3) haematological, cytological or histological report (confirmed cancer).

Area-level socio-economic deprivation and rurality were derived from meshblock codes (containing approximately 60–120 residents) linked to addresses in the address notification table where available, otherwise domicile codes in the NZCR (see the Appendix for full coding information). Socio-economic deprivation was obtained by mapping meshblock/domicile to the 2018 New Zealand Index of Deprivation (NZDep) that assigns geographic areas into one of 10 deciles.²⁷ This study uses NZDep quintiles, with quintile 1 representing the 20% of areas with the lowest levels of socio-economic deprivation and quintile 5 the 20% with the highest. A code representing the rurality of usual residence for each individual was obtained by mapping meshblock/domicile codes to the Geographical Classification for Health (GCH). There are five categories of the GCH; U1 represents the major metropolitan areas and U2 the large urban areas, with R1–R3 being used to indicate increasing rurality with R3 used for the most remote areas.²⁸

Analysis

Analyses were performed within Stats NZ’s IDI environment using SAS Enterprise Guide 8.4 and R version 4.4.0. In accordance with Stats NZ’s microdata output guide, all counts were randomly rounded to base three and counts less than six were suppressed. Crude and age-standardised mortality rates (gastric cancer and all-cause) were produced using STATA 19, along with RRs with 95% confidence intervals (CIs). For age standardisation, data were standardised using the 2018 Census Māori population and presented as per 100 individuals. Equivalent analyses for all-cause mortality are presented in the Appendix. Analyses for 2- and 3-year mortality are also displayed in the Appendix.

Ethics approval

This project was approved by the University of Otago Human Ethics Committee (HD18/064).

Results

Demographic details of gastric cancer cohort

Table 1 displays key demographic details of all individuals in the NZCR with gastric cancer diagnosed from 2002 to 2021 for Māori and non-Māori. A total of 7,851 individuals were diagnosed with gastric cancer between the start of 2002 and end of 2021, of which 1,452 (18.5%) were Māori. Of Māori, 26% also identified with European ethnicity, followed by 2.3% who also identified with Pacific ethnicity. Among non-Māori, European ethnicity was most often reported (81%), followed by Pacific (10%) and Asian (9%) ethnicity.

Comparing 5-year diagnosis periods, there was an increase in cases during 2017–2021, consistent with overall trends in growth, ageing and diversity in the New Zealand population. Higher proportions of Māori were diagnosed at younger ages compared with non-Māori. For example, 18.0% of Māori were under the age of 45 at gastric cancer diagnosis compared with 5.1% of non-Māori. Additionally, 5.2% of Māori were diagnosed under age 30, compared to 0.7% of non-Māori. A much higher proportion of the Māori gastric cancer cohort were living in areas of higher socio-economic deprivation at diagnosis (e.g., 49.6% in the most deprived quintile, compared with 20.1% of non-Māori). Māori with gastric cancer were also more likely to be living in rural areas (i.e., 27.9% vs 19.4% across R1–R3).

View Table 1–5, Figure 1–2.

Data missingness over time

Figure 1 displays rates of missing data in the NZCR for three key clinical characteristics (site, extent of disease and tumour grade) by each year of diagnosis for Māori and non-Māori (for counts and percentages for all cases, see Appendix Table 1). Rates of missingness in these characteristics were generally similar between Māori and non-Māori but varied over the years. For Māori and non-Māori, missing tumour grade information decreased over the years. Missing extent of disease data increased over the years and was consistently higher among non-Māori until recent years (2020). In 2021, approximately 30–50% of information across the key clinical characteristics was missing for Māori and 20–50% was missing for non-Māori.

Where disease characteristics were known (i.e., of the total gastric cancer cohort: 57.3% for extent of disease, 62.7% for tumour grade and 72.4% for cancer site) data are presented for Māori and non-Māori for diagnoses in 2002–2021 and as separate 5-year periods in Table 2. Over the full period, the most common gastric cancer site among Māori was distal (34.9% of cases), or “other” (i.e., non-cardia, non-proximal and non-distal; 27.2%). For non-Māori, cardia was the most common location (48.7%) followed by distal (19.9%). The proportion of distal cases among Māori was highest in the 2002–2006 period (40.5%) and generally lower in each subsequent diagnosis period, although the largest decrease relative to 2002–2006 occurred in 2007–2011 (33.3% of cases distal). By contrast, the percentage of distal cases among non-Māori remained within 17.7–22.1% over the 20-year period.

The extent of disease was classified as distant for more than half of Māori (56.5%) and non-Māori (57.7%) during the 2002–2021 period. However, the proportion of cases classified as distant was higher in more recent diagnosis periods compared with the 2002–2006 period. The proportion was 17 percentage points higher for Māori and 14 percentage points higher for non-Māori by 2017–2021. Tumour grade for diagnoses did not change consistently over the period examined, nor did the basis of diagnosis. Yet a higher proportion of diagnoses among Māori compared with non-Māori were associated with a poorly differentiated/undifferentiated tumour (72.4% vs 60.0% over the 2002–2021 period).

Gastric cancer mortality

Age-standardised gastric cancer mortality rates (ASRs) were higher among Māori compared with non-Māori (with the exception of 3-year and 5-year mortality during the 2007–2011 diagnosis period; see Table 3 for exact details). There was a general downward trend in 1-year gastric cancer ASRs for Māori and non-Māori from 2002 to 2013 (see Figure 2). Five-year gastric cancer ASRs decreased consistently across time periods for Māori but not non-Māori. Overall, the difference in mortality between Māori and non-Māori decreased modestly over the years (for 1-year mortality, the Māori:non-Māori age-standardised rate ratio (ASRR) decreased from 1.15, 95% CI 0.87–1.53, in 2002–2006 to 1.05 in 2017–2021). However, 95% CIs provided no evidence of statistically significant differences in ASRs in any period between Māori and non-Māori. Trends for all-cause mortality were broadly similar and are displayed in Table 3.

Gastric cancer mortality by demographic factors

Table 4 displays gastric cancer mortality within 1 and 5 years of diagnosis, by age groups (RR both within age groups and total age-standardised). The largest differences between Māori and non-Māori were observed for 1-year mortality in the 45–64-year age band (RR 1.15, 95% CI 0.98–1.34) and 85+ age group (RR 1.17, 95% CI 0.70–1.85). However, in part related to the wide CIs, there was insufficient evidence of statistically significant differences for any of the RRs. For the <45 and 45–64-year age groups, Māori:non-Māori crude RRs for mortality within 1 year of diagnosis decreased over time over 2002–2017 (see Appendix Table 2), but the 95% CIs were wide and overlapping.

Gastric cancer mortality rates 1- and 5 years post-diagnosis, by socio-economic deprivation, rurality and sex are displayed in Table 5. The largest ASRR estimates were obtained for Māori men compared with non-Māori men (1-year ASRR 1.18, 95% CI 0.94–1.49) and Māori compared with non-Māori living in areas in the second quintile of the deprivation index (1-year ASRR 1.42, 95% CI 0.87–2.30, 5-year RR 1.22, 95% CI 0.72–2.01) but CIs were wide and crossed the null value. There was no clear pattern of change in RR of mortality within 1 year of diagnosis by socio-economic deprivation, rurality or sex over time (See Appendix Table 3). Equivalent statistics for all-cause survival can be found in Appendix Tables 4–7.

Discussion

This study used up to 20 years of cancer registry data to assess trends in gastric cancer survival inequity between Māori and non-Māori, as well as the socio-demographic and clinical characteristics of diagnoses. Consistent with global trends in mortality¹² and survival,^29,30 gastric cancer mortality among those diagnosed decreased across the period. The relative difference in mortality between Māori and non-Māori appeared to narrow, particularly 1 year post-diagnosis (the ASR for Māori was 15% higher than non-Māori in 2002–2006 vs 5% in 2012–2017). However, the level of precision of these estimates is low, as evidenced by the wide CIs (95% CI −13%–53% and −23%–42%, respectively).

There are multiple reasons to expect gastric cancer survival to improve over time, including ongoing changes in the treatment of gastric cancer.³¹ For example, guidelines have been updated on genetic testing and prophylactic total gastrectomy for hereditary diffuse gastric cancer,³² which contributes to the higher gastric cancer rates among Māori.³³ However, the clinical response to guidelines for diffuse gastric cancers, including prophylactic surgery, has varied over time and by region in New Zealand. Prophylactic surgeries started in just one hospital (Tauranga in 1998), and by 2024 there were seven clinical centres providing prophylactic gastric cancer surgery. Further research on access to treatment and diagnosis pathways could provide more information on gastric cancer survival and ethnic inequities in survival.

Consistent with past research,³⁴ Māori diagnosed with gastric cancer during 2002–2021, compared with non-Māori, were younger at diagnosis, more likely to live in a deprived area and more likely to live rurally. Those living in the least deprived areas had the lowest mortality rates (all-cause and gastric cancer–caused) 1–5 years following diagnosis, as did those living in the most urban areas. However, differences in survival within these sub-groups between Māori and non-Māori were generally consistent and not statistically significant, suggesting ethnic inequities are not specific to particular socio-demographic groups. Moreover, there was no trend of change in these differences over time.

For those with clinical characteristics recorded, diagnoses among Māori were more likely to be of the distal type and graded as poorly differentiated/undifferentiated. The proportion of diagnoses with location classified as distal among Māori appeared notably higher during 2002–2006 (41%) compared with later periods (around 33%), and the proportion with extent of disease classified as distant increased over 2002–2021 for both Māori and non-Māori.

Monitoring clinical characteristics over time is important, as characteristics such as extent of disease and differentiation are central to prognosis.^13,16 However, completeness of data on clinical characteristics varied over time. Missing data for cancer site and tumour grade decreased over the 2002–2021 period, whereas missing data for cancer stage/extent of disease increased. Gurney et al. previously documented that 38% of gastric cancer cases were not staged in the NZCR between 2006–2008.¹² Our findings show this has now increased to 48–50% in 2020–2021. These high rates of missing data impact assessments of mortality associated with clinical characteristics^14,35 and the ability of this routinely collected information to improve service delivery and reduce inequitable outcomes. Data may be missing for multiple reasons, including that the data are collected but not used in clinical settings (creating a disconnect between collectors and users), and that there is no requirement for complete data on submission (in contrast to other national collections such as the death registration process). There are also no standardised clinical staging definitions for gastric cancers, in contrast to cancers such as breast and cervical cancer. Clinical guidelines for hereditary diffuse gastric cancer include information requirements, but these are not yet routinely adopted in clinical practice in New Zealand.

This information deficit together with the absence of clinically important information such as Lauren classification³⁶ or H. pylori infection history highlight the need for a specific and comprehensive cancer registry for gastric and gastro-oesophageal cancers that can be used in the immediate clinical setting as well as public health cancer reduction policy. This could assist in placing greater emphasis on the collection of information that is known to be clinically or diagnostically relevant, in contrast to generalised information in a national register. National breast cancer registries are an example of this, and are managed by clinicians based on important information specific to breast cancer. Specific registries allow a closer relationship between collectors and users of the data, and issues of data completeness and variable range can be addressed at the point of collection. This would be enabled by immediate use of the data in the clinical setting where it is gathered, rather than going back to records long after detecting missing data.

Strengths and limitations

Providing information on gastric cancer survival over time is imperative for monitoring the equity of health outcomes. While the finding that Māori diagnosed with gastric cancer have poorer survival than non-Māori is consistent with past research,¹⁴ our analysis suggests ethnic differences in mortality decreased between 2002 and 2017, particularly for short (1-year) follow-up periods. Linked administrative data allowed us to examine all gastric cancer records in the NZCR during the period of interest, while also being able to access other health and socio-demographic data not recorded within the NZCR. By contrast, not all of those contacted for case-control studies may choose to participate.⁵ This can be especially problematic for cancers with rapid progression and poor survival, such as gastric cancer. However, we could not assess other relevant factors (e.g., H. pylori) that are not routinely collected and recorded. H. pylori infection is an important consideration as it is associated with distal disease, incidence of familial diffuse gastric cancer and poor post-diagnosis outcomes, yet it is a treatable infection. We also did not account for loss to follow-up by means of people leaving the country. Previous work suggests this may have a relatively small impact on gastric cancer survival estimates.³⁷

The relatively low number of cases per year required aggregation of age groups and time periods (years). Year-by-year analysis would have allowed for easier identification of trends. The ability to monitor changes over time requires a consistent and statistically robust approach with regards to time periods and age groups. It is also important to note that the trends only relate to the period examined here, and these may change in future years. For example, analysis of all-cause mortality, for which we could include more recent years, similarly exhibited a downward trend in the ASRR among diagnoses from 2002 to 2016, but an increase in the ASRR among those diagnosed from 2017. It may be possible that increases in the ASRRs for gastric cancer mortality would be observed for diagnoses beyond 2017.

Further, we report on trends in survival-based mortality here, but not mortality rates in the population, which are likely to differ. We also used a broad non-Māori comparison group for monitoring health equity between Māori and other New Zealanders using a standard equity-based approach aligned to the principles of the Treaty of Waitangi. However, gastric cancer incidence and outcomes vary by ethnicity within the non-Māori group. For example, gastric cancer incidence is higher among Pacific peoples compared with European New Zealanders;³⁸ in our study, 10.4% of non-Māori were Pacific. Future research could examine non-Māori ethnic groups in more detail, including examination of variability of cancer outcomes between specific Pacific populations.³⁷ As our analyses were descriptive, we cannot say why inequity may differ across time periods, such as whether detection and treatment have differed over time or whether the ethnic composition of the non-Māori population itself is changing. Finally, our approach is intended to describe gastric cancer mortality over time, so differs from previous work using hazard ratios over longer diagnosis periods. We have used a RR approach for fixed survival periods as this allows for the identification of changes in survival between shorter periods of time. Although we present RRs consistently across the time periods we examine, these results are not directly comparable with hazard ratios.

Conclusion

Using two decades of recent data, we examined trends in gastric cancer survival and inequity among Māori and non-Māori. Mortality risk was higher among Māori but reduced over 2002–2021. The downward trend in mortality risk was larger for Māori compared with non-Māori, but changes were small and risk remained higher among Māori in all diagnosis periods. There were generally no clear trends over time in survival by socio-demographic characteristics. However, there were changes in the clinical characteristics of diagnoses as well as missing data. These findings provide essential insights into how gastric cancer survival might be changing in New Zealand and important data for monitoring ethnic inequity in gastric cancer survival as well as support for a specific gastric cancer register linked to clinical practice.

View Appendix.