Multiple myeloma survival in New South Wales, Australia, by treatment era to 2020

Statistical analysis

The baseline characteristics in the 3 treatment eras were compared with chi-squared tests.

Survival was measured from the date of diagnosis to the date of death, 5 years after diagnosis, or the study end date (December 31, 2020), whichever came first. Those alive at the end of follow-up were censored. Relative survival, the ratio of the observed proportion surviving in a group of individuals with MM to the expected proportion that would have survived in a comparable group of individuals from the general population¹⁹, was calculated because this method is preferred for measuring cancer survival at a population level²⁰ and is robust to inaccuracies in causes of death in population-based data collection²¹. The cohort method was used, because this study was designed to assess temporal trends in survival²². Observed survival was estimated with the life table method²³, and expected survival was calculated with the Ederer II method²⁴, on the basis of NSW life tables stratified by gender, age, and calendar year.

To assess differences in survival over time, we conducted 2 types of analysis. First, we fitted a relative survival regression model for excess deaths from MM^25,26. In this analysis, numbers of death were modeled as a function of year of follow-up, gender, age group, and treatment era through Poisson regression with the logarithm of the person-years at risk as the offset. This model quantified the extent to which the excess risk of death in each treatment era differed from that in the reference era (1985–1995), after controlling for the factors included in the model. Relative excess risks (RERs) and their 95% confidence intervals (CIs) were calculated with the estimated coefficients and standard errors from the Poisson model.

Second, we estimated cause-specific survival with Fine and Gray’s competing risks cumulative incidence function and subdistribution hazard model²⁷ to adjust for other factors, as recommend by Lau et al.²⁸. In this analysis, MM, as the underlying cause of death, was treated as the primary event of interest, and death due to other causes was treated as a competing risk. Cox proportional hazards regression was used for fitting the subdistribution hazard model^28,29. All significance tests with P-value < 0.01 were considered to indicate statistical significance. All analyses were conducted in SAS version 9.4.

This analysis received ethical approval for the Cancer Institute NSW’s Enduring Cancer Data Linkage (also known as CanDLe) initiative from the NSW Population and Health Services Research Ethics Committee (2019/ETH12584). Data were stored in the Secure Unified Research Environment facility, a remote access computing environment to which authorized researchers were given encrypted access with strong authentication.

Relative survival

Relative survival improved over the 36-year study period; the largest increase was observed in the third treatment era. The 5-year relative survival over the treatment eras increased from 31.0% in 1985–1995, to 41.9% in 1996–2007, and 56.1% in 2008–2015 (Table 1 and Figure 1). The improvement in survival over time was observed for all age groups (Figure 1). For individuals younger than 70 years at diagnosis, the 5-year relative survival nearly doubled, from 36.5% in 1985–1995 to 68.5% in 2008–2015 (Figure 1). Less improvement was observed for those 70–79 years of age, whereas little improvement was found for those 80–89 years of age between the first 2 treatment eras. In contrast, significant improvements in relative survival were observed for individuals > 70 years of age in 2008–2015. Similar trends in relative survival were observed between men and women (data not shown).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Five-year relative survival among individuals diagnosed with multiple myeloma, by age at diagnosis in different treatment eras, New South Wales, Australia.

In the multivariable analysis, RERs due to diagnosis of MM were significantly lower (P < 0.0001) for individuals diagnosed in 1996–2007 (RER = 0.73, 95% CI: 0.69–0.78) and 2008–2015 (RER = 0.48, 95% CI: 0.45–0.52) than in 1985–1995 (Table 2). Age at diagnosis was a significant prognostic factor: older individuals had higher excess death (RER = 1.71 and 2.76 for those 70–79 and 80–89 years of age, respectively) than younger individuals (< 70 years of age). RERs decreased with year of follow-up, thereby indicating a progressively lower excess risk of death for people with prevalent MM.

Cause-specific survival

Cumulative incidence curves of death due to MM by treatment era are presented in Figure 2. The cumulative incidence significantly decreased over time (Gray’s test P < 0.0001), and the 5-year cumulative incidence of death due to MM was much higher among those diagnosed in 1985–1995 (0.58, 95% CI: 0.57–0.60) than in 1996–2007 (0.47, 95% CI: 0.46–0.49) or 2008–2015 (0.37, 95% CI: 0.36–0.39) (Table 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Cumulative incidence of death due to multiple myeloma among individuals diagnosed in different treatment eras, New South Wales, Australia.

The estimated hazard ratios (HRs) by treatment era are reported in Table 3. After adjustment for gender and age at diagnosis, treatment era was a significant prognostic factor (P < 0.0001): HRs were significantly diminished in the later 2 eras (HR = 0.87, 95% CI: 0.81–0.92 for 1996–2007 and HR = 0.69, 95% CI: 0.64–0.73 for 2008–2015).

Treatment era

Treatment options have been an important focus of MM management, given that MM is considered an incurable disease, and appropriate treatment can prolong life¹⁰. The treatment advancements in Australia from 1985 to 2020 have been previously outlined^12,30. Briefly, the progression from chemotherapy alone to SCT to novel agents has introduced more options for the increasing number of individuals diagnosed and living with MM^6,30. From the 1990s, SCT was heralded as a breakthrough in MM treatment, with evidence of improved 5-year overall survival in individuals with MM who received SCT compared with those who did not receive SCT in Australia¹⁵; however, improvements were generally seen in younger individuals (under 70 years of age)¹⁶. In our study, relative survival increased in the 1996–2007 era, aligning with the availability of SCT. The improvement in relative survival continued in the 2008–2015 treatment era, thus reflecting the introduction of novel agents in Australia. Evidence has indicated that SCT, compared with conventional therapies, has resulted in improvements in overall survival^31,32. Although not directly comparable, a Chinese hospital-based study³² has found that people with MM who received SCT had significantly higher survival than those who did not receive SCT (similarly to our comparison of the first 2 treatment eras). In the same study, novel-agent-based regimens were associated with improved survival for people with MM³², similarly to our observations in the first and third treatment eras in this study. The use of combinations of novel agents or younger age at treatment (< 70 years) have also demonstrated more pronounced improvements in overall survival^33,34. The improvements in 5-year relative survival across the treatment eras observed in our study were similar to those in a study by Harwood et al.¹⁶ reporting rates of 30% (1982–1995), 43% (1996–2007), and 53% (2008–2014).

Age group

Our findings showed improvements in relative survival across treatment eras by age group and gender. In older age groups (≥ 70 years), survival improved across the treatment eras, from 26% to 48% in people 70–79 years of age and, to a lesser extent, from 19% to 34% in people ≥ 80 years of age. Although many studies have reported improved survival outcomes over time for all age groups, the increase has commonly been less pronounced in older age groups compared to in younger age groups^17,18,35, as observed in those > 70 years of age between 1985–1995 and 1996–2007. An improvement was also seen in Harwood et al.¹⁶ but was less pronounced in people ≥ 80 years of age, in whom the relative survival was 23% in 2008–2014. More recently, improvements in all age groups have become more evident^3,16–18,35. The observation of less pronounced improvements in survival among older age groups until more recently is attributable to the lack of treatment options before the introduction of novel agents, given that Australian guidelines regarding SCT eligibility for people > 65 years of age have only recently changed, and that poorer disease biology and/or other comorbidities are associated with older age^12,15,36.

Strengths, limitations, and suggestions for future research

This study has several notable strengths. First the study included a substantially larger sample than previous Australian studies. Moreover, it captured all individuals with MM in the chosen population (NSW), unlike randomized controlled trials or other commonly used treatment efficacy study designs, which tend to involve strict recruitment selection criteria. Furthermore, we used triangulation within methods to increase study validity, and we observed a temporal survival improvement with 2 different analytical approaches: relative and cause-specific survival.

Although robust, our study is not without limitations. Given that comprehensive treatment information was not available, the exact treatment regimens in the cohort are unknown. The treatment eras were used as proxies for the treatment provided, in line with previous studies¹⁶. Furthermore, individuals were categorized into treatment eras according to their diagnosis date; however, although the date of diagnosis determined the number and choice of treatment options available to an individual, it did not indicate that the individual actually received the newest type of treatment available or even any treatment at all. For example, an individual diagnosed with MM between 2008 and 2014 (the “novel agent” treatment era) might have received SCT; however, the 5-year survival outcomes were calculated as part of the third treatment era. This approach might have overestimated the survival benefits attributed to treatment available at the time, because the survival improvements might potentially have been due to treatment regimens from earlier eras or other unknown lifestyle or biological factors. Although our relative survival analysis considered this possibility by using calendar-specific lifetables from the general population, such biases cannot be completely eliminated when long-term historical data are used.

As costly new MM therapies emerge, demonstrating the efficacy of MM treatments in terms of not only survival but also quality of life outcomes will become increasingly important, to ensure that treatments extend life without sacrificing quality of life. This aspect represents an important gap in the current MM literature: recent Australian research has reported that MM is associated with the highest levels of psychological distress, disability, and pain among all cancer types, as well as the lowest likelihood of continuing in the workforce, thus resulting in high indirect costs^37–39.