Natural History of DCIS of the Breast
TL;DR14–53% of DCIS misdiagnosed as benign and left after biopsy alone progressed to invasive cancer over 10+ yr; precise progression rates not estimable.
The RT-relevant read is the recurrence gap: pooled recurrence 22.5% after CS alone vs 8.9% with CS+RT, widest for comedo type (34.4% vs 8.3%), with about half of recurrences invasive. Low-grade DCIS recurs with longer latency, so short follow-up flatters RT omission in exactly the lesions where omission is proposed.
CS+RT cut pooled recurrence to 8.9% from 22.5% with CS alone, with the largest gaps in comedo type, necrosis, and high and intermediate nuclear grade. These are cross-series pooled rates, not randomized subgroup contrasts, so they inform but do not settle RT omission by grade, especially for slow-recurring low-grade DCIS.
Recurrence depends strongly on excision margins, and pooled recurrence was 1.4% after mastectomy vs 22.5% after CS alone. DCIS left after biopsy alone progressed in 14–53% over 10+ yr, so incomplete excision is not a benign default, though most of those lesions were low grade.
9 details 4 trials watching
Review drawing on MEDLINE and citation tracking of key reviews and editorials. Evidence streams: misdiagnosed-DCIS follow-up, recurrence after BCS, autopsy, risk factors, animal and mathematical models.
Untreated misdiagnosed DCIS progressed in 14–53% over 10+ yr. Pooled recurrence 22.5% CS alone, 8.9% CS+RT, 1.4% mastectomy. Autopsy median DCIS prevalence 8.9%.
| Group | CS alone | CS+RT | Mastectomy |
|---|---|---|---|
| Overall | 22.5% (95% CI 16.9–28.2) | 8.9% (95% CI 6.8–11.0) | 1.4% (95% CI 0.7–2.1) |
| Comedo type | 34.4% | 8.3% | NR |
| Necrosis present | 24.7% | 8.2% | NR |
| High nuclear grade | 23% | 9.5% | NR |
| Intermediate nuclear grade | 22.7% | 7.5% | NR |
Predates the low-risk DCIS active-surveillance trials (LORIS, COMET, LORD), which test directly the progression question this review could only estimate indirectly from misdiagnosed biopsies.
Recurrence after surgery may not reflect the untreated course and depends strongly on excision margins. Boyages subgroup rates are pooled across series, not randomized contrasts.
Authors conclude DCIS is a precursor of most invasive breast cancer but not all DCIS progresses in the medium term. The autopsy reservoir argument for overdiagnosis is weakened because modelling predicts that reservoir even if all DCIS progresses.
- Grade-specific progression rate of untreated DCIS n=2500 · primary completion 2034-02 · LORD: surveillance vs standard tx, grade I/II DCIS
- Which low-risk DCIS can safely omit surgery or RT active Comparing an Operation to Monitoring, With or Without Endocrine Therapy (COMET) Trial For Low Risk DCIS Phase NAn=997 · primary completion 2024-06 · COMET: active monitoring vs surgery, low-risk DCISnot yet Assessment of Biosignature Classification of DCIS for RadioTherapy Benefit Post Lumpectomy (ABCD RT) Phase 3n=5270 · primary completion 2039-07 · ph3 RT omission if DCISionRT score <2.8
- Biomarkers predicting invasive recurrence of DCIS n=1500 · primary completion 2025-07 · DCISionRT registry linking test to outcomes
📚 Sources · 📄 1 paper
The longer read
The useful move in this review is that it separates two questions a DCIS discussion tends to blur: how often DCIS becomes invasive if left in place, and how often it returns after excision with or without RT. The misdiagnosed-biopsy series speak to the first, the Boyages pooled recurrence data to the second, and the authors are explicit that neither can stand in for the other. Recurrence after surgery describes residual disease shaped by margins, not the natural course of an untouched lesion.
The 14–53% progression range deserves less confidence than its frequent citation suggests. It comes from a handful of small retrospective series, the largest identifying 80 misdiagnosed cases, and the biases run in both directions. Most lesions were low grade, which pushes the estimate down for high-grade disease. Follow-up was incomplete in several series, half of Rosen's cohort for example, and women later diagnosed with cancer were easier to trace, which pushes it up. Latency matters too: in the Page series, two invasive cancers appeared at 23 and 31 years, so a 10 to 15 year window truncates the outcome for exactly the indolent lesions now proposed for de-escalation.
For an RT reader, the Boyages figures carry the decision content. Pooled recurrence after CS alone was 22.5% against 8.9% with CS+RT, and the absolute difference was largest for comedo type, necrosis, and high and intermediate nuclear grade. These are pooled rates across series with different margins and follow-up, not randomized subgroup contrasts, so they describe where RT appeared to do the most work rather than proving where it can be dropped. The review's own observation that low-grade DCIS recurs with longer latency is the more important point for omission: a low-grade cohort looks safe at short follow-up partly because its events have not happened yet.
The autopsy argument is handled more carefully than it usually is. A median undiagnosed DCIS prevalence of 8.9% is often read as proof of overdiagnosis, but the Kopans model shows a reservoir of that kind would exist even if every lesion eventually progressed, simply because slow growers sit below the detection threshold for years. Classification drift between ADH and DCIS further inflates the autopsy numbers. That leaves the overdiagnosis case resting mainly on the misdiagnosed-biopsy series, which are the weakest-sampled data in the review.
What this does not settle is grade-specific and age-specific progression, and which markers identify the lesion that will invade. The later active-surveillance trials (LORIS, COMET, LORD) were designed to answer the progression question prospectively in low-risk disease; this review explains why the retrospective literature could not, and why their follow-up will need to be long before a reassuring early result means much.