Skip to main content

Table 1

From: Molecular and gene expression-based predictors of response to preoperative chemotherapy

Reference

Patient characteristics

Treatment

Results

Chang et al. [1]

24 (discovery) and 6 (validation) patients. with LABC

4 cycles docetaxel every 3 weeks

92 differentially expressed genes (P < 0.001) LOOCV results in training: predictive accuracy of 88% (95% CI, 68–97%), PPV of 92%, NPV of 83%, sensitivity of 85% (95% CI, 55–98%), specificity of 91% (95% CI, 59–100%)

   

Independent validation: all six patients with sensitive tumors correctly identified

Ayers et al. [2]

24 (discovery) and 12 (validation) patients with breast cancer

3 or 12 cycles paclitaxel and 4 cycles of FAC (T/FAC)

No single gene sufficiently associated with pCR to serve as single valid marker 74-gene predictor for response to chemotherapy

   

- Predictive accuracy of 78% (95% CI, 52–94%), PPV of 100% (95% CI, 29–100%), NPV of 73% (95% CI, 45–92%), sensitivity of 43% (95% CI, 10–82%), specificity of 100% (95% CI, 72–100%)

Iwao-Koizumi et al. [3]

44 (discovery) and 26 (validation) patients. with stage II/IV breast cancer

4 cycles of docetaxel

85-gene-predictor of pCR with:

   

- Predictive accuracy of 80.7% (95% CI, 63.5–92.5%), PPV of 73.3% (95% CI, 49.5–90.3%), NPV of 90.9% (95% CI, 65.9–99.4%), sensitivity of 91.7% (95% CI, 68.1–99.5%), specificity of 71.4% (95% CI, 46.7–89.5%)

Hannemann et al. [4]

48 patients with LABC

6 cycles of doxorubicin/docetaxel (AD, n = 24) or doxorubicin/cyclophosphamide (AC, n = 24)

No differentially expressed genes identified, no possible multigene predictor developed

   

Patient samples did not cluster distinctly for pCR/near pCR compared with RD in hierarchical clustering

Gianni et al. [5]

89 (INT-Milan) and 82 (MDACC) patients with LABC

3 cycles of doxorubicin/paclitaxel, 12 cycles of weekly paclitaxel

86 genes correlating with pCR (P < 0.05) forming three clusters

   

- ER gene cluster

   

- Proliferation gene cluster

   

- Immune-related gene cluster

   

RS correlated with pCR in INT-Milan patients 86-gene model developed on INT-Milan patients validated on MDACC samples

Folgueira et al. [6]

(1) 38 (discovery), 13 (validation)

 

(1) 25 (3.8%) differentially expressed transcripts between responders and non-responders, three-gene classifier could not be validated

 

(2) 31 (discovery), 13 (validation)

 

(2) Three-gene classifier successfully developed and validated both by LOOCV and in an independent validation set

Dressman et al. [7]

37 patients with stage IIB/III breast cancer

Liposomal doxorubicin × 4/paclitaxel combined with local whole breast hyperthermia

(a) 22-gene signature characterizing IBC identified and validated by LOOVC

   

(b) 18-gene signature characterizing IBC identified and validated by LOOVC

   

(c) No gene signature predicting clinical response could be identified

Thuerigen et al. [8]

52 (discovery/GEsDoc), 48 (validation/GEDoc) patients with T2-4N0-2M0 breast cancer

GEsDoc (gemcitabine, epirubicin every 2 weeks × 5 followed by docetaxel every 2 weeks × 2), GEDoc (gemcitabine, epirubicin, docetaxel every 3 weeks)

512-gene-signature predictive of pCR: predictive accuracy of 88% (95% CI, 75–95%), PPV of 64% (95% CI, 39–81%), NPV of 95% (95% CI, 82–99%), sensitivity of 78% (95% CI, 40–97%), specificity of 90% (95% CI, 76–97%)

Park et al. [9]

21 patients with stage II/III breast cancer

4 cycles of FEC every 3 weeks followed by 12 cycles of weekly paclitaxel

11 differentially expressed ABC transporters Multigene predictor model with the ABC transporters differentially expressed between the two classes (P < 0.003) with predictive of pCR with:

   

- Predictive accuracy of 92.8% (95% CI, 88.0–97.4%), PPV of 93.2% (95% CI, 85.2–100%), NPV of 93.6% (95% CI, 87.8–99.4%), sensitivity of 88.1% (95% CI, 76.8–99.4%), specificity of 95.9% (95% CI, 87.8–100%)

Cleator et al. [10]

40 patients with primary breast cancer

6 cycles doxorubicin/cyclophosphamide (AC)

253 differentially expressed genes;

   

- 75 genes overexpressed in resistant tumors (that is, transcription, differentiation, signal transduction, amino acid metabolism)

   

- 178 genes overexpressed in sensitive tumors (that is, cell cycle, survival, stress response, and estrogen-regulated genes)

Hess et al. [11]

82 (discovery), 51 (validation)

3 or 12 cycles paclitaxel and 4 cycles of FAC (T/FAC)

30-gene predictor identified and validated by fivefold cross-validation, permutation testing and application to an independent data set:

   

- Predictive accuracy of 76% (95% CI, 62–87%), PPV of 52% (95% CI, 31–73%), NPV of 96% (95% CI, 82–100%), sensitivity of 92% (95% CI, 64–100%), specificity of 71% (95% CI, 54–85%)

Mina et al. [12]

45 patients with stage II/III breast cancer

3 cycles doxorubicin every 2 weeks and 6 cycles docetaxel weekly × 6

22 of 274 candidate genes correlated with pCR (P < 0.05) forming three large clusters: angiogenesis-related genes, proliferation-related genes, invasion-related genes, no correlation between RS and pCR, 24/274 genes correlated with inflammatory phenotype

  1. ABC transporter, ATP binding cassette transporter; discovery, development/training set; FAC, 5-fluorouracil, doxorubicine, cyclophosphamide; FEC, 5-fluorouracil, epirubicin, cyclophosphamide; IBC, inflammatory breast cancer; LABC, locally advanced breast cancer; LOOCV, leave-one-out cross-validation; NPV, negative predictive value; pCR, pathological complete response; PPV, positive predictive value; RD, residual disease; RS, recurrence score (Oncotype DX©); validation, validation set.