Background: From 2003 to 2005, standardised 5-year cancer survival in China was much lower than in developed countries and varied substantially by geographical area. Monitoring population-level cancer survival is crucial to the understanding of the overall effectiveness of cancer care. We therefore aimed to investigate survival statistics for people with cancer in China between 2003 and 2015.

Methods: We used population-based data from 17 cancer registries in China. Data for the study population was submitted by the end of July 31, 2016, with follow-up data on vital status obtained on Dec 31, 2015. We used anonymised, individual cancer registration records of patients (aged 0–99 years) diagnosed with primary, invasive cancers from 2003 to 2013. Patients eligible for inclusion had data for demographic characteristics, date of diagnosis, anatomical site, morphology, behaviour code, vital status, and last date of contact. We analysed 5-year relative survival by sex, age, and geographical area, for all cancers combined and 26 different cancer types, between 2003 and 2015. We stratified survival estimates by calendar period (2003–05, 2006–08, 2009–11, and 2012–15).

Findings: There were 678 842 records of patients with invasive cancer who were diagnosed between 2003 and 2013. Of these records, 659 732 (97·2%) were eligible for inclusion in the final analyses. From 2003–05 to 2012–15, age-standardised 5-year relative survival increased substantially for all cancers combined, for both male and female patients, from 30·9% (95% CI 30·6–31·2) to 40·5% (40·3–40·7). Age-standardised 5-year relative survival also increased for most cancer types, including cancers of the uterus (average change per calendar period 5·5% [95% CI 2·5–8·5]), thyroid (5·4% [3·2–7·6]), cervix (4·5% [2·9–6·2]), and bone (3·2% [2·1–4·4]). In 2012–15, age-standardised 5-year survival for all patients with cancer was higher in urban areas (46·7%, 95% CI 46·5–47·0) than in rural areas (33·6%, 33·3–33·9), except for patients with oesophageal or cervical cancer; but improvements in survival were greater for patients residing in rural areas than in urban areas. Relative survival decreased with increasing age. The increasing trends in survival were consistent with the upward trends of medical expenditure of the country during the period studied.

Interpretation: There was a marked overall increase in cancer survival from 2003 to 2015 in the population covered by these cancer registries in China, possibly reflecting advances in the quality of cancer care in these areas. The survival gap between urban and rural areas narrowed over time, although geographical differences in cancer survival remained. Insight into these trends will help prioritise areas that need increased cancer care.