Home >  Incite >  Issue 6 > 

What factors most influence survival and the odds of presenting with localised cancer in NSW patients?

Research updates

Issue 6, December 2010

Survival from cancer in NSW varies by cancer type, sex, age, other socio-demographic factors and summary stage at diagnosis.

Population-based cancer registries routinely report for each of these factors individually using relative survival.

Relative survival is a ratio of number of people who are diagnosed and followed up for a specified time, usually within a five-year period, adjusted for the background mortality of the population.

The purpose of this study is to look at factors that most influence survival using proportional hazards regression modeling and to examine the odds of presenting with localised versus non localised cancer while controlling for all other factors.

Method

All NSW cases of invasive cancer diagnosed between 1980 and 2003 followed to the end of 2004 were included. The NSW Central Cancer Registry allocates the cause of cancer death from coded ABS deaths and the full death certificate.  In addition, because detailed case information is derived from pathology reports and other notifications the final allocation of the cause of death is more specific.

Proportional hazards regression modeling of survival was undertaken including as predictor's age, sex, period of diagnosis, extent of cancer and each Area Health Service. Logistic regression analysis was also conducted to determine the odds of presenting with localised cancer while controlling for other covariates.

Results

There were 551,926 cancer patients in NSW diagnosed between 1980 and 2003, which were followed up to the end of 2004. The most important determinants of cancer death in order (highest to lowest) were extent of cancer at diagnosis or summary stage, age at diagnosis, period of diagnosis, sex, socioeconomic status, migrant status and urban versus rural area health service.

The factors that influenced survival, in order of strongest to weakest evidence of effect, were:

• Extent of disease at diagnosis: those with distant disease were nine times more likely to have died than those with localised disease.
• Age at diagnosis: those aged 80+ were three times more likely to have died than those aged 0-49.
• Year of diagnosis: those diagnosed more recently were less likely to have died than those diagnosed early in the study period.
• Sex: women were less likely to have died than men.
• Socioeconomic status: those in higher SES groups were less likely to have died than those in lower SES groups.
• Country of birth: those born outside Australia were more likely to have died than those born in Australia.
• The study did not consider Aboriginal and Torres Strait Islander status because available data are not of sufficient quality. (Table 1 and 2)

In summary, compared with residents in the rest of NSW, residents in rural area health service (A, B and C) had a 4 per cent higher, 7 per cent higher or 10 per cent lower likelihood of death while controlling for sex, age, extent of disease, country of birth, socioeconomic status and the year of their cancer diagnoses.

As extent of disease is the factor that most influences survival, it makes sense to look at what factors influence presentation with localised cancer. These factors, in order of strongest to weakest evidence of effect, were:

• Age: older age groups were less likely to be diagnosed with localised cancer than the youngest age group.
• There was a greater odds of presenting with localised cancer from 1994 onwards. This may be due to better reporting due to electronic notification of cancer. The higher odds of presenting with localised disease for people diagnosed in 1984 to 1988 appears to be unusual and requires further analysis by cancer type.
• Country of birth: those born outside Australia were less likely to be diagnosed with localised cancer than those born in Australia.
• Socioeconomic status: those in higher SES groups were more likely to be diagnosed with localised cancer than those in lower SES groups.
• Sex: women were more likely to be diagnosed with localised  cancer than men.
• Area Health Service of residence: those in rural Area Health Services were 7 per cent less likely to be diagnosed with localised cancer than those in urban Area Health Services. However, when different models were undertaken by cancer site the odds varied for different cancers. For breast and bowel cancer, rural AHS residents were no different to urban residents in the odds of being diagnosed with localised cancer; for prostate cancer, rural residents were 4 per cent less likely to be diagnosed with localised cancer, and notably, for lung cancer, rural AHS residents were 23 per cent less likely to be diagnosed with localised cancer.  (Table 3 and 4)

Discussion

Variation in cancer survival and the influence of age, stage, socioeconomic status, migrant background and period of diagnosis have been demonstrated in NSW ^1-3, and when NSW is compared with the UK.⁴

One of the limitations in this study is that survival differentials found between Area Health Services may represent co-morbidity differentials, risk factor behaviours or treatment differences. Examination of the NSW Health Survey data and the Chief Health Officer's report indicate that in those areas with lower survival there are higher rates of diabetes hospitalisation and chronic obstructive airways disease, coronary heart disease and stroke hospitalisations and deaths compared to the state. Co-morbidity and surgical treatment differences and the impact that these have on Area Health Service survival need to be investigated further using linkage with inpatient data.

A further contentious area is whether using cause specific survival instead of relative survival adequately reflects survival differentials. Previous analysis comparing survival by cancer type and stage using cause specific and relative survival in NSW shows that there is very little difference in the outcome. Gamel  et al ⁵ used both methods to determine survival in a population of 119,502 breast cancer patients from the Surveillance, Epidemiology and End Results (SEER) programme data set, with 20 years of follow-up.

In all strata, there was only minimal deviation between the two survival methods. Cox proportional hazards regression modeling has been found to be a reliable method for modeling survival and has been used to show  variation in survival outcomes in NSW for cancers of the breast,⁶ ovary and bladder,^{7, 8} WA for breast cancer ⁹ and on SEER data for bowel cancer,¹⁰ Florida and Bulgaria on cervical cancer patients using National cancer registry data,^{11, 12} Alabama cancer registry on pancreatic cancer patients,^{13, 14} SEER data on lung cancer patients.^{15 13}

Conclusion

Cancer survival is most influenced by extent of disease at diagnosis and least affected by AHS of residence. Factors that influence the odds of presenting with localised disease are the inverse of those that influence survival. The hazard or likelihood of dying of cancer and the odds of presenting with localised cancer varied by AHS and cancer type while controlling for all other factors. Future studies will need to consider the impact of co-morbidities and surgical treatment on survival. With good cause of death data proportional hazards regression modeling is an acceptable and practical method to demonstrate factors that most explain the variance in survival outcomes so that targeted strategies can be employed.

References

1. Yu, X.Q., et al., A population-based study from New South Wales, Australia 1996-2001: area variation in survival from colorectal cancer. Eur J Cancer, 2005. 41(17): p. 2715-21.
2. Yu, X.Q., et al., Assessing the impact of socio-economic status on cancer survival in New South Wales, Australia 1996-2001. Cancer Causes Control, 2008. 19(10): p. 1383-90.
3. Yu, X.Q., et al., Trends in survival and excess risk of death after diagnosis of cancer in 1980-1996 in New South Wales, Australia. Int J Cancer, 2006. 119(4): p. 894-900.
4. Woods, L.M., et al., Large differences in patterns of breast cancer survival between Australia and England: a comparative study using cancer registry data. Int J Cancer, 2009. 124(10): p. 2391-9.
5. Gamel, J.W. and R.L. Vogel, Non-parametric comparison of relative versus cause-specific survival in Surveillance, Epidemiology and End Results (SEER) programme breast cancer patients. Stat Methods Med Res, 2001. 10(5): p. 339-52.
6. Tracey, E., et al., Survival and degree of spread for female breast cancers in New South Wales from 1980 to 2003: implications for cancer control. Cancer Causes Control, 2008. 19(10): p. 1121-30.
7. Tracey, E., et al., Bladder cancer survivals in New South Wales, Australia: why do women have poorer survival than men? BJU Int, 2009. 104(4): p. 498-504.
8. Tracey, E.A., et al., Reasons for improved survival from ovarian cancer in New South Wales, Australia, between 1980 and 2003: implications for cancer control. Int J Gynecol Cancer, 2009. 19(4): p. 591-9.
9. Clayforth, C., et al., Five-year survival from breast cancer in Western Australia over a decade. Breast, 2007. 16(4): p. 375-81.
10. Redaniel, M.T., et al., Inter-country and ethnic variation in colorectal cancer survival: comparisons between a Philippine population, Filipino-Americans and Caucasians. BMC Cancer, 2010. 10: p. 100.
11. Brookfield, K.F., et al., Disparities in survival among women with invasive cervical cancer: a problem of access to care. Cancer, 2009. 115(1): p. 166-78.
12. Kostova, P., V. Zlatkov, and S. Danon, Five-year overall survival and prognostic factors in patients with cervical cancer in Bulgaria. J BUON, 2008. 13(3): p. 363-8.
13. Eloubeidi, M.A., et al., Impact of staging transesophageal EUS on treatment and survival in patients with non-small-cell lung cancer. Gastrointest Endosc, 2008. 67(2): p. 193-8.
14. Eloubeidi, M.A., et al., Prognostic factors for survival in pancreatic cancer: a population-based study. Am J Surg, 2006. 192(3): p. 322-9.
15. Bach, P.B., et al., Racial differences in the treatment of early-stage lung cancer. N Engl J Med, 1999. 341(16): p. 1198-205.