Overdiagnosis in the population-based organized breast cancer screening program estimated by a non-homogeneous multi-state model: a cohort study using individual data with long-term follow-up

Table 3 Maximum-likelihood estimates and 95% confidence intervals based on the non-homogeneous and the homogeneous multi-state model

Description	Parameter	MLE (95% CI)
		Non-homogeneous model			Homogeneous model
		40–49 years	50–59 years	60–69 years^a	Homogeneous model
Transition rate from free of BC to progressive PCDP	λ₁₂(t)	–	0.00276 (0.00269, 0.00283)	0.00381 (0.00371, 0.00390)	0.00306 (0.00301, 0.00311)
Transition rate from progressive PCDP to CP	λ₂₃(t)	0.385 (0.342, 0.428)	0.464 (0.436, 0.492)	0.284 (0.267, 0.301)	0.418 (0.400, 0.436)
Ratio of λ₁₄(t) to λ₁₂(t)	r	0.00182 (0, 0.00523)			9.999×10⁻⁴ (0, 3.432×10⁻³)
Sensitivity	S	0.880 (0.852, 0.909)			0.924 (0.901, 0.947)
-2*log(likelihood)	NA	198,024			198,878
Mean sojourn time (year)	\( \frac{1}{\lambda_{23}(t)} \)	2.60 (2.31, 2.89)	2.16 (2.03, 2.29)	3.52 (3.31, 3.73)	2.39 (2.29, 2.49)

Abbreviations: BC breast cancer, CI confidence interval, CP clinical phase, MLE maximum likelihood estimate, MST mean sojourn time, NA not applicable, PCDP preclinical screen-detectable phase
Likelihood ratio test of the non-homogeneous versus the homogeneous model: \( {\chi}_{(3)}^2 \) = 854, P <0.001
^a9.65% of the women were invited to screening at age 70–74

ISSN: 1465-542X