Background: Evidence of gastric cancer (GC) screening’s impact on GC incidence remains limited, hindering the development of health policies for screening programs. We aimed to assess the population impact of organized GC screening programs on GC incidence in Seoul, South Korea. Methods: South Korea initiated nationwide organized GC screening in 2002 for individuals aged 40 years or above. We estimated the effect on age-standardized GC incidence in the Seoul population aged 40 and above using a flexible synthetic control method (SCM). We considered a total of 32 registries without organized GC screening as candidate controls for Seoul. The SCM constructs a synthetic control by deriving a weighted average of control registries based on outcomes during the pre-intervention period. We compared post-intervention trends in GC incidence in Seoul with the trend of the synthetic control to estimate average post-intervention rate ratios (RRs). We used data from the Cancer Incidence in Five Continents (CI5) Plus database for 1993-2012, CI5 Volume XII for 2013-2017, and the Seoul Cancer Registry for 1993-2020 for outcomes. We conducted several sensitivity analyses, including the negative control outcome analysis, leave-one-out examination, predictability test by narrowing the pre-intervention period, and expanded analysis with the latest 5-year interval data from CI5 Volume XII. We employed various alternative methods, including the simplified difference-in-differences (DID) analysis and synthetic DID analysis. We also conducted the updated age-period-cohort (APC) analysis to evaluate screening’s impact on GC incidence by integrating a screening period function. GC incidence between 1993 and 2020 without the screening was then estimated by summing up the age, period, and cohort effects. Results: SCM included 14 registries as the synthetic control, with Shanghai serving as the main contributor. The pre-intervention fit between Seoul and the synthetic control was good. The base-case analysis indicated that the screening program was associated with an increase in GC incidence in Seoul, with an average post-intervention RR of 1.20 (95% CI 1.13-1.27). A progressively expanding gap was observed between Seoul and the synthetic control from 2002 to 2012 (Figure 1). Sensitivity analyses confirmed the increased effect on incidence. But the expanded analysis with latest data and APC analysis showed that the effect began to fall after 2012 (Figure 2). Conclusions: We found that organized GC screening increased GC incidence in Seoul, Korea, possibly due to early detection and program expansion. It may take some time for a compensatory fall in the subsequent GC incidence and any broader preventative effect of GC screening to become evident. Future studies are needed to explore potential overdiagnosis, examine the longer-term effect and underlying mechanisms.

<b>Figure 1 Trends in age-standardized gastric cancer incidence in Seoul, South Korea versus the synthetic control (A) as well as the intervention effects and placebo tests (B).</b> The dashed vertical line indicates the introduction of nationwide organized gastric cancer screening. In panel B, the red lines denote the estimated intervention effects in Seoul. Meanwhile, the grey lines correspond to the effects estimated in placebo tests, where each control registry was treated as the unit of interest. RR indicates the rate ratio. CI indicates confidence interval.

Figure 1 Trends in age-standardized gastric cancer incidence in Seoul, South Korea versus the synthetic control (A) as well as the intervention effects and placebo tests (B). The dashed vertical line indicates the introduction of nationwide organized gastric cancer screening. In panel B, the red lines denote the estimated intervention effects in Seoul. Meanwhile, the grey lines correspond to the effects estimated in placebo tests, where each control registry was treated as the unit of interest. RR indicates the rate ratio. CI indicates confidence interval.

<b>Figure 2 Expanded synthetic control analysis with latest 5-year interval data from CI5 XII (A & C).</b> <b>Estimated screening effect, observed and modelled incidence rates with and without the screening effect from the age-period-cohort model (B & D).</b> In panels A and C, the dashed lines connect the estimates of 2012 and those of 2013-2017. The dashed vertical line indicates the start of nationwide organized screening. In panel C, the red line denotes estimated intervention effects in Seoul. Grey lines correspond to effects estimated in placebo tests, where each control registry was treated as the unit of interest. APC means age-period-cohort analysis, which was based on Seoul Cancer Registry’s data for 1993-2020. In panel B, the red line indicates observed incidence rates in Seoul, and green line and blue line denote modelled incidence with and without screening effect. In panel D, black points indicate point estimates of rate ratios and vertical lines denote ranges of confidence intervals.

Figure 2 Expanded synthetic control analysis with latest 5-year interval data from CI5 XII (A & C). Estimated screening effect, observed and modelled incidence rates with and without the screening effect from the age-period-cohort model (B & D). In panels A and C, the dashed lines connect the estimates of 2012 and those of 2013-2017. The dashed vertical line indicates the start of nationwide organized screening. In panel C, the red line denotes estimated intervention effects in Seoul. Grey lines correspond to effects estimated in placebo tests, where each control registry was treated as the unit of interest. APC means age-period-cohort analysis, which was based on Seoul Cancer Registry’s data for 1993-2020. In panel B, the red line indicates observed incidence rates in Seoul, and green line and blue line denote modelled incidence with and without screening effect. In panel D, black points indicate point estimates of rate ratios and vertical lines denote ranges of confidence intervals.

Yige Li

Appearances