Predictive Biomarkers for LNG-IUS Treatment Response in Endometrial Hyperplasia

Prospective cohort study (August 1998 – December 2007, tertiary centre, Birmingham UK).
Investigates whether baseline immunohistochemical (IHC) expression of five biomarkers in endometrial hyperplasia (EH) predicts:
- Histological regression after Levonorgestrel-releasing intra-uterine system (LNG-IUS) therapy.
- Subsequent relapse after initial regression.
Biomarkers tested: Estrogen Receptor (ER), Progesterone Receptor (PR), Cyclo-oxygenase-2 (COX-2), DNA mismatch-repair protein Mlh1, anti-apoptotic protein Bcl-2.

Weak/absent ER & PR expression might signal progesterone resistance → poor response to LNG-IUS.
COX-2 over-expression correlates with aromatase activity → heightened local estrogen biosynthesis → potential treatment failure.
Loss of Mlh1 (methylation / MSI) & high Bcl-2 (apoptosis inhibition) have been linked to EH progression; could predict persistence/relapse.

Included: All women with complex hyperplasia (CH) or atypical complex hyperplasia (ACH) opting for LNG-IUS.
Excluded: No follow-up histology, insufficient tissue for IHC, inadequate IHC staining.
Final analysed N = 174 (155 CH, 19 ACH).

LNG-IUS (Mirena®) inserted after counselling; intended duration 5 y.
Histological surveillance:
- Office Pipelle every 6 mo × 2 y, then yearly to 5 y.
- Hysteroscopic curettage if sampling inadequate or clinical concern.
Median follow-up: $72.1\,\text{months}$ (IQR 59.1–89.8).

6 × 5 µm FFPE sections per case; 5 for IHC, 1 for H&E.
Automated Dako staining; antigen retrieval pH 8, 98 °C.
Primary antibodies & dilutions:
- ER-α (1D5) $1{:}150$
- PR-A/B (1A6) $1{:}50$
- COX-2 (4H12) $1{:}50$
- Mlh1 (G168-728) $1{:}100$
- Bcl-2 (124) $1{:}100$
Semi-quantitative Q-score (0-8) = intensity (0-3) + proportion (0-5).
Two blinded assessors; inter-observer agreement $\kappa = 0.801 \pm 0.036$ (82.6 %).

Descriptive stats; $\chi^2$ for categorical variables.
Time-to-event (regression, relapse) analysed with Cox proportional hazards; results as Hazard Ratios (HR).
Explored all centile cut-offs of Q-score; reported two lowest p-value thresholds (5th/10th vs 25th/50th/75th/90th).
Missing data handled by complete-case analysis.
Software: STATA v12.1.
Rule-of-ten events/variable acknowledged; study may be under-powered (Type II error risk).

Initial outcomes: 164 regressed (94.3 %), 10 persisted (5.7 %).
ER expression
- Cut-off Q ≥ 2 (5th centile): $HR = 0.09\,(95\%\,CI\,0.01–0.39,\,p = 0.001)$
- Cut-off Q ≥ 4 (10th centile): $HR = 0.19\,(95\%\,CI\,0.05–0.47,\,p = 0.008)$
  → Weak/negative ER raises risk of persistence.
PR expression
- Q ≥ 2 (5th): $HR = 0.32\,(95\%\,CI\,0.09–0.88,\,p = 0.02)$
- Q ≥ 4 (10th): $HR = 0.46\,(95\%\,CI\,0.17–0.91,\,p = 0.02)$
COX-2, Mlh1, Bcl-2: No significant association (all p > 0.14).

Analysable: 152 women (11 lost to follow-up, 1 hysterectomy).
Relapse events: 18 (11.8 %); median time to relapse $32.2\,\text{mo}$ (IQR 11.3–57.7).
None of the five biomarkers predicted relapse at either optimal or secondary cut-off (all p > 0.11).
- Example: ER Q ≥ 2 (5th centile) $HR = 0.38\,(95\%\,CI\,0.05–2.92,\,p = 0.354)$ .

EH is estrogen-dependent; local estrogen synthesis promoted via aromatase & COX-2 pathway.
COX-2 links inflammation → PGE₂ → upregulates aromatase → estrogen excess → proliferation.
Mlh1 methylation causes microsatellite instability; Bcl-2 inhibits apoptosis.
Obesity contributes via peripheral aromatisation of androgens & hyperinsulinaemia.

Weak ER/PR expression modestly increases risk of persistence, but predictive strength is low; LNG-IUS still induces regression in most.
No biomarker reliably predicts relapse; surveillance strategy cannot yet be individualised on IHC alone.
Obesity (BMI > 35 kg/m²) emerged in broader research by authors as an independent relapse predictor → weight-management interventions may enhance long-term success.
For ACH where surgery declined/unfit, LNG-IUS remains viable but demands vigilant follow-up regardless of biomarker status.

Few adverse events (10 persistence, 18 relapse) → limited power; potential Type II errors (false-negatives).
Office Pipelle sampling under LNG-IUS can be histologically challenging (stromal decidualisation, glandular atrophy) → risk of misclassification.
Single-centre; external validity needs confirmation.

External validation cohorts to confirm weak ER/PR signal.
Composite prognostic model integrating:
- BMI (modifiable),
- Histological class (CH vs ACH),
- Molecular markers (existing + novel genomics/proteomics),
- Clinical factors (age, parity, diabetes, tamoxifen/HRT exposure).
Investigate aromatase & COX-2 inhibitors as adjuncts where over-expression documented.

Inter-observer agreement $\kappa = 0.801$ (excellent).
Persisted hyperplasia rate with LNG-IUS ≈ $5.7\%$ .
Relapse rate after regression ≈ $11.8\%$ over 6 y.
Rule-of-ten events per variable in Cox models can be relaxed but study acknowledges limitation $[Vittinghoff\,\&\,McCulloch\,2007]$ .

LNG-IUS superior to oral progestogens for EH regression $[Gallos\,2010\,meta-analysis]$ .
ACH progression to carcinoma $\approx 29\%$ ; concomitant cancer present in $\le 43\%$ at diagnosis $[Trimble\,2006]$ —hence treated surgically unless contraindicated.
Combination aromatase + COX-2 inhibition under trial in breast cancer; potential translational avenue for EH.
Excess body weight a well-established risk factor for EH/cancer progression $[Anderson\,1996]$ .