PowerPoint Presentation


BACKGROUND

• Patients with cutaneous melanoma (CM) have an individual recurrence

risk determined by clinical, pathological, and genetic features.

• The 31-gene expression profile (31-GEP) test is an independent

significant predictor of 5-year risk of recurrence and distant metastasis.1-7

• 31-GEP results classify tumor biology as lowest-risk (Class 1A), low-risk

(Class 1B), high-risk (Class 2A), and highest-risk (Class 2B):

Development and validation of a nomogram incorporating the 31-GEP test and clinicopathologic 

factors for accurate prediction of recurrence risk in patients with cutaneous melanoma
Thorpe, R.1; Caruso, H.G.2; Covington, K.R.2; Brodland, D.3; Zitelli, J.3

1. Ada West Dermatology, 2. Castle Biosciences, Inc., 3. Zitelli and Brodland, P.C.

All patients

n=685

Class 1A

n=557

Class 1B

n=41

Class 2A

n=33

Class 2B

n=54

Age median (range), years
67 

(22-90)

65
(22-90)

71
(33-90)

71
(52-91)

74 
(26-90)

Breslow thickness (range), mm
0.5 

(0.1-13)

0.5 
(0.1-5)

0.8
(0.2-6)

1.2
(0.2-7.5)

2.6
(0.2-13)

Male 
60%

(411/685)

59%
(330/557)

51%
(21/41)

79%
(26/33)

63%
(34/54)

Ulceration present
7% 

(50/685)

3% 
(17/557)

7%
(3/41)

6% 
(2/33)

52% 
(28/54)

Mitotic rate ≥ 2 mm2
18%

(121/685)

9%
(52/557)

32%
(13/41)

52%
(17/33)

72%
(39/54)

OBJECTIVE: To develop a nomogram tool combining 31-GEP class 

and clinicopathologic risk features for predicting CM recurrence.

1. Gerami et al. Clin Cancer Res. 2015 Jan 1;21(1):175-83.

2. Gerami et al. JAAD. 2015 May;72(5):780-5.e3.

3. Zager et al. BMC Cancer. 2018 Feb 5;18(1):130.

4. Gastman et al. JAAD. 2019 Jan;80(1):149-157.e4.

5. Hsueh et al. J Hematol Oncol. 2017 Aug 29;10(1):152.

6. Greenhaw et al. Dermatol Surg. 2018 Dec;44(12):1494-1500.

7. Keller et al. Cancer Med. 2019 May;8(5):2205-2212.

REFERENCES DISCLOSURES
Castle Biosciences, Inc (CBI) provided statistical

analysis support for nomogram development. KRC and

HGC are employees and option holders of CBI.

Prospective cohort registry is independently managed

by the Cutaneous Oncology Research Consortium

(CORC). RT, DB, JZ have no relevant disclosures.

METHODS:  NOMOGRAM DEVELOPMENT

Table 1. Patient clinical and pathologic features per 31-GEP Class

RESULTS

• A prospective cohort of 685 patients from 9 dermatology centers with

minimum 1yr follow-up or a recurrence event at any time was included in

nomogram development.

• A logistic regression model was fitted on clinical and pathological data to

determine relative predictive value for recurrence risk. Covariate inclusion

for the model was selected by lowest Bayesian information criteria (BIC)

value with fewest clinical features.

• The nomogram was validated on a retrospective cohort of 901 Stage I-III

CM patients with > 5 years follow-up or a recurrence event, and goodness

of fit was determined by linear regression.

GEP Class N Events 1.5-yr RFS

Class 1A 557 14 (2.5%) 98.9% (98.0-99.8%)

Class 1B 41 6 (14.6%) 97.6% (93.0-100%)

Class 2A 33 4 (12.1%) 93.9% (86.1-100%)

Class 2B 54 24 (44.4%) 70.3% (59.1-83.6%)

p<0.0001

Class 1A

Class 1B

Class 2A

Class 2B

R
F

S

Time (years)

Figure 1. Multivariate Cox regression 

analysis of 31-GEP and clinicopathologic 

features

Figure 2. Kaplan-Meier estimation of 

Recurrence-Free Survival (RFS)

• This nomogram combines the 31-GEP test result with clinical features to

create a clinically useful, accurate tool for determining an individual’s risk

of recurrence to optimize patient care.

• Because Sentinel Lymph Node (SLN) status is not a feature in this

nomogram, this tool can be used to provide patient risk of recurrence prior

to or in the absence of a SLN biopsy.

• A future aim of this study is to generate a mobile application for

conversion of clinical and molecular data to a patient’s recurrence risk.

CONCLUSIONS

Figure 3:  Optimum Model selected by BIC Figure 4. Validation of the nomogram in a 

retrospective cohort of 901 patients with 

Stage I-III cutaneous melanoma

 
 
s
e
rv
e
d
  
e
c
u
rr
e
n
c
e
  
a
te

 redicted  ecurrence  ate

   

  2

   

   

   

              2

Predicted Recurrence Rate

O
b
s
e
rv

e
d
 R

e
c
u
rr

e
n
c
e
 R

a
te

p < 0.0001

Patients with 

Stage I-III 

melanoma

Class 1:

Low risk of SLN positivity 

(eligible T1-T2)

Low risk of melanoma

recurrence (T1-T4)

Class 2:

Higher risk of SLN positivity 

(eligible T1-T2)

High risk of melanoma 

recurrence (T1-T4)

•Quantifies expression of 31 

genes from primary tumor 

using RT-PCR

•Applies a validated algorithm

•Accurately classifies patients 

as low or high risk 

1A
Lowest risk

2B 
Highest risk

2A 
Increased risk

1B
Low risk

Figure 5. Impact of T stage and 31-GEP on risk of recurrence

AJCC v8
GEP 

Class 1A

GEP 

Class 1B/2A

GEP 

Class 2B

T1a 1.9% 2% 5% 9%

T1b 6.9% 4% 9% 19%

T2a 12.1% 7% 19% 24%

T2b 25.0% 15% 29% 45%

T3a 25.0% 11% 27% 40%

T3b 45.0% 19% 35% 53%

T4a 37.5% 15% 29% 45%

T4b 85.7% 48% 71% 83%

RESULTS CONTINUED

RFS Hazard Ratio