GÖTEBORG-1 active surveillance, 25-year follow-up
ForScreen-detected very-low/low/intermediate-risk localized PC on AS
TL;DR25-yr PC-specific survival 94% (CI 91-98%) on AS, but 22-yr failure-free survival 68% and treatment-free survival 38%.
RT was the exit for 44 of 488 men, but the paper reports no outcomes by treatment, so the efficacy of delayed RT after AS can't be separated out. Failure kept accruing past 10 yr in every risk group. For a man choosing AS over upfront RT, the trade is decades of monitoring, not a 10-15 yr window.
In a screen-detected very-low or low-risk man, this supports AS as low risk for PC death out to 25 yr. In intermediate-risk disease (24-yr PCSS 85%, 19-yr FFS 55%), it supports discussing upfront definitive therapy as a real alternative. It does not extend to high-risk PC.
RT was the exit for 44 of 488 men, with no outcomes reported by treatment, so the efficacy of delayed RT can't be judged from this paper. Failure keeps accruing past 10 yr, which matters when counseling a man choosing between AS and upfront RT.
RP was the main exit from AS (141 of 232 discontinuations). Gleason score (HR 3.12) and PSA density (HR 1.78) at diagnosis predicted later failure, which helps identify men who might do better with upfront surgery than with prolonged surveillance.
12 details
Observational AS cohort inside the GÖTEBORG-1 PSA screening trial (1995 to 2014 diagnoses, follow-up to Dec 31, 2023). Data were collected prospectively, but AS was not randomized and followed no predefined protocol. Kaplan-Meier estimates plus a multivariable Cox model for failure.
Of 1052 screen-detected PCs, 494 (47%) were on AS and 488 were included: 251 very-low, 129 low, 108 intermediate risk. High-risk excluded. Median age 66, PSA 4.1 ng/ml, PSA density 0.12.
Treatment-free, failure-free (noncurative PSA relapse, starting HT, metastasis, or PC death), PC-specific, and overall survival, all measured from diagnosis.
RT was the active treatment for 44 men leaving AS, against 141 RP. Post-RT relapse was defined as nadir +2 ng/ml. No dose, technique, or RT-specific outcome is reported.
PCSS 94% at 25 yr against FFS 68% at 22 yr; 14 PC deaths and 81 failures. Gleason score (HR 3.12) and PSA density (HR 1.78) predicted failure.
| Endpoint | 15 yr | 20 yr | 22 yr | 25 yr |
|---|---|---|---|---|
| PC-specific survival | 97% (CI 95-99%) | 95% (CI 93-98%) | n/a | 94% (CI 91-98%) |
| Overall survival | 63% (CI 58-67%) | 46% (CI 41-51%) | n/a | 32% (CI 26-38%) |
| Treatment-free survival | 48% (CI 43-54%) | 43% (CI 37-50%) | 38% (CI 31-46%) | n/a |
| Failure-free survival | 81% (CI 77-85%) | 74% (CI 68-81%) | 68% (CI 60-78%) | n/a |
| Endpoint | Very low (n=251) | Low (n=129) | Intermediate (n=108) |
|---|---|---|---|
| TFS 19 yr | 55% (CI 48-63) | 35% (CI 26-47) | 30% (CI 18-48) |
| FFS 19 yr | 85% | 74% | 55% |
| PCSS 24 yr | 99% (CI 97-100) | 92% (CI 83-100) | 85% (CI 75-95) |
| OS 24 yr | 38% (CI 30-48) | 34% (CI 26-45) | 22% (CI 12-37) |
PC mortality falls in the range of ProtecT (3.4%) and the Toronto cohort (5.7% at 15 yr). It is roughly half of PIVOT (11.4%) and well below SPCG-4, a gap the authors attribute to screen detection versus clinical diagnosis.
Gleason reclassification after 2005 means some low-risk pts here would be intermediate risk today. Failure lumps PC death with untreated low-level PSA relapse. Risk-group differences in PCSS were not formally tested.
Failure hazard stayed roughly constant over time, and nearly half of failures came >10 yr after diagnosis, so there is no safe point to stop monitoring. Failures in Gleason 6 pts show that some Gleason 6 tumors do progress.
Nonrandomized AS cohort with no comparator arm and no predefined surveillance protocol. Pre-MRI sextant selection limits how far the result transfers to contemporary AS.
- Does MRI-based selection and monitoring lower long-term AS failure?
- Is there any safe point to stop AS monitoring?
- How does intermediate-risk AS perform under modern grading?
📚 Sources · 📄 1 paper
The longer read
The durable finding here is not the 94% PC-specific survival at 25 years. Earlier AS series at 10 to 15 years already predicted that. The new information is the pace of failure beyond 15 years. Failure-free survival fell from 81% at 15 years to 68% at 22 years, treatment-free survival kept falling, and nearly half of all failures came more than a decade after diagnosis. Cohorts that stopped at 10 or 15 years, including the PASS and PRIAS-type registries the authors cite, could not see this tail. Screen-detected tumors carry long lead times, so the events that matter show up late.
Set against ProtecT, the result fits rather than contradicts. ProtecT found that men randomized to monitoring carried a higher metastasis risk than men treated upfront, with low PC mortality in every arm. GÖTEBORG-1 reports PC mortality in a similar range (ProtecT 3.4%, Toronto 5.7% at 15 years) and a large burden of failure short of death. Both point the same way: surveillance protects against overtreatment at the cost of some missed cure, and that cost only becomes visible with long follow-up. The contrast with PIVOT and SPCG-4 is best read as a population effect, screen-detected versus clinically diagnosed disease, not as evidence that this AS approach was superior.
Methodology should lower a reader's confidence in how well these numbers transfer, and it cuts in two directions. The cohort was selected and followed with PSA and sextant biopsies, almost no MRI, and no fixed rebiopsy schedule. That likely undersampled disease at entry and caught progression late. Modern MRI-based selection and monitoring should reduce both problems, which would make contemporary outcomes look better. Grading drift pushes the other way: some tumors called low risk here would be graded intermediate risk today and treated upfront. Contemporary AS also increasingly includes favorable intermediate-risk disease, where this cohort already showed the worst figures: 19-year FFS 55% and 24-year PCSS 85%. Those intermediate-risk estimates rest on 108 men with wide intervals, and the paper did not formally test the risk-group differences. They inform counseling but should not be read as precise.
The failure definition also needs scrutiny. It combines PC death, metastasis, starting hormonal therapy, and PSA relapse after surgery or RT with no further treatment. The authors defend this as a failure to achieve the curative intent the men enrolled with. That is reasonable for counseling, but it inflates the event count compared with endpoints that change survival. The 18 men with untreated relapse are a different clinical problem from the 14 who died.
For radiation oncologists the paper is mostly context, not RT evidence. Forty-four men went to RT, but outcomes are not broken out by treatment, so this says nothing about whether delayed RT after long surveillance performs as well as upfront RT. What it does settle is the counseling frame for a man weighing surveillance against definitive therapy: the risk of death is low, but monitoring never ends, and exiting to treatment late is common. What would have to be true for this to overstate the risk is exactly what modern selection claims to fix. That claim still lacks 20-year data.