Dermatology: Practical and Conceptual


Research  |  Dermatol Pract Concept 2017;7(2):9 39

DERMATOLOGY PRACTICAL & CONCEPTUAL
www.derm101.com

Triage amalgamated dermoscopic algorithm (TADA) 
for skin cancer screening

Tova Rogers1, Maria Marino1, Stephen W. Dusza1, Shirin Bajaj1, Michael A. Marchetti1, 
Ashfaq Marghoob1

1 Dermatology Service, Department of Medicine, Memorial Sloan Kettering Cancer Center, New York, New York

Key words: dermoscopy, dermoscopy algorithm, melanoma, basal cell carcinoma, squamous cell carcinoma

Citation: Rogers T, Marino M, Dusza SW, Bajaj S, Marchetti MA, Marghoob A. Triage amalgamated dermoscopic algorithm (TADA) for 
skin cancer screening. Dermatol Pract Concept. 2017;7(2):9. DOI: https://doi.org/10.5826/dpc.0702a09

Received: January 30, 2016; Accepted: February 19, 2017; Published: April 30, 2017

Copyright: ©2017 Rogers et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, 
which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Funding: This research was funded in part through the NIH/NCI Cancer Center Support Grant P30 CA008748.

Competing interests: The authors have no conflicts of interest to disclose.

All authors have contributed significantly to this publication.

Corresponding author: Ashfaq A. Marghoob, MD, Dermatology Service, Department of Medicine, Memorial Sloan Kettering Cancer 
Center, 16 E 60th Street, New York 10065, USA. Email: marghooa@mskcc.org

Importance: Dermoscopic triage algorithms have been shown to improve beginners’ abilities for iden-
tifying pigmented skin lesions requiring biopsy.

Objective: To estimate the diagnostic accuracy of the Triage Amalgamated Dermoscopic Algorithm 
(TADA) for pigmented and nonpigmented skin cancers. Secondarily, to compare TADAs performance 
to those of existing triage algorithms for the identification of pigmented skin cancers.

Design: Cross-sectional, observational, reader study that took place at a beginner and intermediate 
level dermoscopy course.

Participants: Two hundred medical professionals of various specialties attended the course and 120 
voluntarily joined the study (60% participation rate).

Exposures: After receiving basic dermoscopy training, participants evaluated 50 polarized, dermo-
scopic images of pigmented (22 benign, 18 malignant) and nonpigmented (1 benign, 9 malignant) skin 
lesions using TADA. Pigmented lesions were also evaluated using the Three-Point Checklist and AC 
Rule. With TADA, participants first determined if a lesion was an unequivocal angioma, dermatofi-
broma, or seborrheic keratosis, which would exclude it from further evaluation. All other lesions were 
assessed for architectural disorder, starburst pattern, blue-black or gray color, shiny white structures, 
negative network, ulcer/erosion, or vessels. Any one feature indicated suspicion for malignancy.

Results: Most participants were dermatologists (n=64, 53.3%) or primary care physicians (n=41, 
34.2%), and many lacked previous dermoscopy training (n=52, 43.3%). TADA’s sensitivity and spec-
ificity for all skin cancers was 94.6% (95% CI=93.4-95.7%) and 72.5% (95% CI=70.1-74.7%), 
respectively. For pigmented skin cancers, the sensitivity and specificity were 94.0% (95% CI=92.9-
95.0%) and 75.5% (95% CI=73.8-77.2%). This compared to 71.9% (95%CI=69.8-73.9%) and 
81.4% (95%CI=79.7-83.0%) for the Three-Point Checklist and 88.6% (95%CI=87.1-89.9%) and 
78.7% (95%CI=76.9-80.3%) for the AC Rule.

Conclusions: These results suggest that TADA compares favorably to existing triage algorithms and 
might be a useful triage tool with high sensitivity and specificity for pigmented and nonpigmented skin 
cancers. Further studies are needed to validate these preliminary observations.

ABSTRACT

mailto:marghooa@mskcc.org


40 Research  |  Dermatol Pract Concept 2017;7(2):9

white color) and the AC Rule (asymmetry, color variation), 

have demonstrated the feasibility of this approach by quickly 

improving novices’ abilities to recognize pigmented lesions 

requiring biopsy [11,12]. A limitation of both methods is that 

they were designed for subsets of pigmented skin cancers.

The triage amalgamated dermoscopic algorithm (TADA) 

was designed to identify common pigmented and nonpig-

mented skin cancers (Figure 1). Although TADA does not 

ask users to preselect and apply the algorithm to exclusively 

pigmented lesions, it does require that users first determine 

if a lesion is a dermoscopically unequivocal example of one 

of three commonly encountered benign neoplasms (angioma, 

dermatofibroma, seborrheic keratosis). If a lesion is deter-

mined to be one of these three neoplasms, it is excluded from 

further evaluation. If not, it is then evaluated for architectural 

disorder (i.e., disorganized or asymmetric distribution of 

colors and/or structures)—a robust criterion that is strongly 

associated with malignancy and has good interobserver agree-

ment, making it easier to teach and learn and lending itself to 

the inclusion in a triage algorithm [13,14]. To improve sen-

sitivity for organized and symmetric skin cancers, including 

certain melanomas (e.g., spitzoid, desmoplastic, nodular, and 

amelanotic) as well as non-melanoma skin cancers, TADA 

includes six additional criteria (starburst, blue-black or gray 

color, shiny white structures, negative network, ulcer/erosion, 

Introduction

Dermoscopy allows skilled observers to more accurately iden-

tify pigmented skin cancers compared to clinical exam alone 

[1-3]. In some cases, it can also help identify nonpigmented 

malignancies [4]. Despite the potential for improved skin 

cancer detection, a number of barriers are preventing many 

dermatologists, dermatology residents, and other medical 

professionals interested in skin cancer management from 

adopting dermoscopy. Lack of training has been cited as a 

major hindrance [5,6]. Nonetheless, the use of dermoscopy 

is increasing [7], and with it, interest in educational materi-

als that provide novices an entry point into dermoscopy [8].

Teaching beginners the numerous and often nuanced 

dermoscopic patterns and structures required for diagnosis 

can be daunting. This has led some authors to suggest that 

triage and not diagnosis be the goal of the dermoscopic 

evaluation when performed by non-experts [9,10]. Triage in 

the context of skin lesion evaluations requires the examiner 

to determine if a lesion is suspicious for malignancy, thus 

requiring a biopsy or specialist referral; it does not require 

that a specific diagnosis be made. Triage algorithms may be 

easier to teach, learn, and implement by allowing for the 

nonspecific identification of concerning lesions using limited 

dermoscopic criteria. The validation of two triage algorithms, 

the Three-Point Checklist (asymmetry, atypical network, blue-

Figure 1. The Triage Amalgamated Dermoscopic Algorithm (TADA)—Illustrated diagram outlining TADAs step-wise 

 approach to evaluating and managing pigmented and nonpigmented skin lesions. [Copyright: ©2017 Rogers et al.]



Research  |  Dermatol Pract Concept 2017;7(2):9 41

Powershot G16; Canon Inc., Tokyo, Japan) and a dermoscopy 

lens attachment (DermLite FOTO system; 3Gen Inc., San 

Juan Capistrano, CA, USA).

Dermoscopy Training: The study was conducted during 

the latter half of the second day of a three-day dermoscopy 

course. On day one of the course, participants were lec-

tured on basic dermoscopic criteria of common benign and 

malignant skin lesions. Participants were also introduced to 

the idea of dermoscopy algorithms as part of a lecture on 

dermoscopic teaching methodologies. On the morning of 

day two, participants reviewed the material covered on day 

one via unknown lesion identification sessions with feedback. 

Instruction on how to apply the Three-Point Checklist, AC 

Rule, and TADA algorithms occurred during a 30-minute 

training session immediately prior to the study.

Evaluation of Study Lesions: Dermoscopic images were 

displayed in PowerPoint® and projected onto two large 

screens. Participants used worksheets to evaluate the study 

lesions. The worksheets separately listed the dermoscopic 

criteria included in the three algorithms. For the Three-Point 

Checklist, the criteria evaluated were asymmetry (monoaxial 

or biaxial), atypical network, and blue-white color, with 

two of the three being required for biopsy. For the AC Rule, 

the criteria evaluated were asymmetry and color variation, 

which were ranked on a scale of 1 to 10. Based on the evalu-

ation of these two criteria, users then determined if a lesion 

was suspicious for malignancy (yes or no) [22]. For TADA, 

participants were first asked to determine if a lesion was 

an unequivocal angioma, dermatofibroma, or seborrheic 

keratosis. If the lesion was determined to be one of these 

three, they were instructed to stop filling out the worksheet 

and wait for the next case. Otherwise, participants assessed 

the lesion for architectural disorder. Lesions demonstrating 

this feature were considered to be suspicious for malignancy 

without need for further evaluation for the remaining TADA 

criteria. Lesions lacking architectural disorder (i.e., organized, 

symmetric lesions) were evaluated further for the presence 

of starburst pattern, blue-black or gray color, shiny white 

structures, negative network, ulcer/erosion, or vessels, with 

the presence of any one feature indicating suspicion for 

malignancy. Lesions lacking all TADA criteria were consid-

ered equivocal and required monitoring for morphological 

changes or symptoms (i.e., itching, bleeding). Clinical images 

were not provided. However, information regarding textural 

features (i.e., firm, keratotic, smooth, dimpling) was given. 

The lesions were displayed in random order.

Statistical Analysis: Descriptive statistics were used to 

describe the study participants, study lesions, and participant 

evaluations. Three separate dichotomous outcome measures 

were created with the data to reflect the participants’ lesion 

evaluations for TADA, the Three-Point Checklist, and AC 

Rule. The primary independent variable for these analyses 

vessels) previously validated to be associated with different 

subtypes of melanoma and non-melanoma skin cancers and 

are independent predictors of malignancy [15-21]. The pres-

ence of any one feature included in TADA warrants a biopsy 

or specialist referral. Since shiny white structures can only be 

seen with polarized light, TADA requires the use of polarized 

dermoscopy. TADA was not tested on facial, mucosal, volar, 

or nail lesions.

The aim of this study was to determine the sensitivity and 

specificity of TADA for the detection of common skin cancers 

(melanoma, basal cell carcinoma, squamous cell carcinoma). 

A secondary aim was to compare the performance of TADA, 

the Three-Point Checklist, and AC Rule when identifying 

pigmented study lesions.

Materials and Methods

Study Design: This was a cross-sectional, observational study 

performed in an experimental setting.

Participants: This study was approved by the Memorial Sloan 

Kettering Cancer Center Institutional Review Board without 

requirement of written informed consent in accordance 

with the Helsinki Declaration. The study was performed on 

August 14, 2015, at a dermoscopy course for beginner and 

intermediate level dermoscopy users. All registered attendees 

were invited to participate. Participation was voluntary. There 

was no compensation or inducement. Participant character-

istics (age, sex, medical specialty, prior dermoscopy training, 

cumulative dermoscopy experience) were recorded on data 

collection forms.

Image Selection: The image records of AAM were retrospec-

tively and sequentially reviewed, starting from the most recent 

dermoscopic image on file, to identify an approximately 

equal proportion of representative examples of common 

benign and malignant skin lesions. Facial, mucosal, volar, and 

nail lesions were excluded. Sixty-two skin neoplasms were 

selected, of which twelve were excluded due to image quality 

or lack of polarized dermoscopic images. The resulting 50 

lesions included 27 malignant (16 melanomas, 7 basal cell 

carcinomas, and 4 squamous cell carcinomas) and 23 benign 

neoplasms (8 nevi, 5 angiomas, 5 seborrheic keratoses, 4 

dermatofibromas, and 1 clear cell acanthoma). All but one of 

the 16 melanomas measured less than 0.5 mm thick (nodular 

melanoma >1mm). Ten of the 50 lesions (20%) were clinically 

and dermoscopically nonpigmented (2 melanomas, 5 basal 

cell carcinomas, 2 squamous cell carcinomas, and 1 clear cell 

acanthoma). All malignant lesions were pathologically veri-

fied. Benign lesions were either evaluated pathologically or 

were required to be unchanged compared to baseline images. 

Images were captured with contact polarized dermoscopy 

(x10 magnification factor) using a digital camera (Canon 



42 Research  |  Dermatol Pract Concept 2017;7(2):9

96% of which were correctly classi-

fied (n=411). Angioma had a false posi-

tive rate (malignant lesions erroneously 

identified as angioma) of 0.2% (n=13). 

Of the 458 dermatofibroma diagnoses 

made (31.7% of the 1,443 lesion evalu-

ations in step one), 94% (n=431) were 

correct. The false positive rate for der-

matofibroma was 1% (n=29). The diag-

nosis of seborrheic keratosis was made 

on 558 occasions (38.7% of the 1,443 

is the benign or malignant nature of 

the lesion based on histologic evalua-

tion. Separate cross-classifications of 

the benign/malignant nature of a lesion 

by participant algorithm outcome were 

created and used to calculate overall 

estimates of diagnostic accuracy. Since 

study participants evaluated multiple 

study lesions, a general estimating equa-

tions approach was used to estimate 

model-based diagnostic accuracy mea-

sures while evaluating the effect of par-

ticipant characteristics, such as previous 

dermoscopy training and/or years prac-

ticing dermatology. Separate models 

were independently used to estimate 

sensitivity and specificity. Algorithm 

performance comparisons of sensitivity, 

specificity and area under the receiver 

operating characteristic (ROC) curve 

were made. For the Three-Point Check-

list and AC Rule, only participants’ 

responses for pigmented study lesions 

were recorded and used for statistical 

analysis. All statistical analyses were 

performed with Stata v14.1 (Stata Cor-

poration, College Station, TX).

Results

Two hundred individuals attended the 

dermoscopy course and 120 (60%) par-

ticipated in the study. Participant char-

acteristics, including age, sex, medical 

specialty, previous dermoscopy training, 

and years of dermoscopy experience, are 

indicated in Table 1.

In total, 5,646 lesion evaluations 

were performed, 3,036 malignant 

(53.8%) and 2,610 benign (46.2%), 

with a mean of 47 evaluations per par-

ticipant (out of a possible 50). In the first 

step of TADA, 25.6% (n=1,443/5,646) 

lesion evaluations resulted in the diag-

nosis of angioma, dermatofibroma, or 

seborrheic keratosis (Figure 2). Of these 

lesions, 94% (n=1,357) were histologi-

cally benign and 90% (n=1,301) were 

one of the three, above-named, benign 

lesions. The diagnosis of angioma was 

made on 427 evaluations (29.6% of the 

1,443 lesion evaluations in step one), 

TABLE 1. Characteristics of study participants (n=120). 
[Copyright: ©2017 Rogers et al.]

Variable Coding n (%)

Age <20 7 (5.8)

21-30 28 (23.3)

31-40 25 (20.8)

41-50 37 (30.8)

51-60 15 (12.5)

61-70 4 (3.3)

71-80 0 (0.0)

>81 0 (0.0)

Did Not Respond 4 (3.3)

Sex Male 52 (43.3)

Female 64 (53.3)

Did Not Respond 4 (3.3)

Specialty Dermatology 64 (53.3)

Internal Medicine 22 (18.3)

Family Medicine 19 (15.8)

Emergency Medicine 2 (1.7)

General Surgery 2 (1.7)

Pathology 1 (0.8)

Dentistry 1 (0.8)

Instrumental Physics 1 (0.8)

Medical Student 1 (0.8)

Did Not Respond 7 (5.8)

Previous Dermoscopy Training Yes 63 (52.5)

No 52 (43.3)

Did Not Respond 5 (4.2)

Years of Previous Dermoscopy 
Experience

0 28 (23.3)

≤1 24 (20.0)

2-5 35 (29.2)

6-10 18 (15.0)

>10 9 (7.5)

Did Not Respond 6 (5.0)

lesion evaluations in step one). Of these, 

82% (n=459) were correctly classified. 

Seborrheic keratosis had a false positive 

rate of 1.4% (n=44). Among the 3,036 

evaluations of histologically malignant 

lesions, 2,950 were performed in the 

second step of TADA, denoting that 

97% of malignant study lesions were 

correctly triaged in step one.

Participants performed 4,203 lesion 

evaluations (74.4%) in step two of 



Research  |  Dermatol Pract Concept 2017;7(2):9 43

ing, are listed in Table 2. In order to compare the diagnostic 

performance of TADA to the Three-Point Checklist and AC 

Rule, nonpigmented study lesions (n=10) were excluded 

from the analysis. In this evaluation, TADA performed with 

the highest sensitivity (94.0%, 95% CI: 92.9%-95.0%), fol-

lowed by the AC Rule (88.6%, 95% CI: 87.1%-89.9%) and 

the Three-Point Checklist (71.9%, 95% CI: 69.8%-73.9%). 

The Three-Point Checklist had the highest specificity (81.4% 

(95% CI: 79.7%-83.0%), followed by the AC Rule (78.7%, 

95% CI: 76.9%-80.3%) and TADA (75.5%, 95% CI: 73.8%-

77.2%). ROC curves for the three algorithms highlight these 

results (Figure 3).

Discussion

In this pilot study, we tested a novel triage algorithm to 

determine its sensitivity and specificity for common skin 

cancers. A significant proportion of study participants lacked 

TADA, of which 3,590 (85.4%) were identified as suspicious 

for malignancy based on the presence of any single TADA 

criterion. Eighty percent of these lesions were true malignan-

cies. The features that most strongly discriminated benign 

and malignant study lesions were disorganized architecture, 

ulcers/erosions, and shiny white structures. Disorganized 

architecture was identified in 57% of histologically malignant 

lesions versus 21% of benign lesions (p<0.001). Ulcers/ero-

sions were identified in 29% versus 11% of malignant and 

benign lesions, respectively (p<0.001). Shiny white structures 

were identified in 25% versus 11% of malignant and benign 

lesions, respectively (p<0.001). Of the 613 lesion evaluations 

lacking any TADA criteria (14.6% of the 4,203 lesion evalu-

ations in step two), 87% were truly benign.

TADA had a sensitivity of 94.6% and a specificity of 

72.5% for all malignant study lesions. Sensitivity and speci-

ficity estimates for the individual study lesions, as well as for 

participants with and without previous dermoscopy train-

Figure 2. Breakdown of participants’ responses for TADA Step 1 and Step 2—Results of the 5,646 lesion evaluations per-

formed by study participants as a function of the true diagnoses of study lesions. Benign lesions correctly identified in TADA 

step 1 are classified as “true benign” and are further broken down into either “correctly classified,” for lesions with true diag-

noses of angioma, dermatofibroma, or seborrheic keratosis, or “incorrectly classified,” for all other benign lesions. Malignant 

study lesions incorrectly identified as benign in TADA step 1 are classified as “false benign.” Lesions identified as malignant in 

TADA step 2 (positive for any one criteria) are classified as either “true malignant,” for lesions with true malignant diagnoses, 

or “false malignant,” for lesions with true benign diagnoses. Lesions identified as not malignant, or equivocal, in TADA step 

2 (negative for all criteria) are either classified as “true benign,” for lesions with true benign diagnoses, or “false benign,” for 

lesions with true malignant diagnoses.

Abbreviations: DF, dermatofibroma; SK, seborrheic keratosis; CCA, clear cell acanthoma; AMM, amelanotic melanoma; 

BCC, basal cell carcinoma; MM, malignant melanoma; NM, nodular melanoma; SCC, squamous cell carcinoma

[Copyright: ©2017 Rogers et al.]



44 Research  |  Dermatol Pract Concept 2017;7(2):9

laypersons, respectively [11,12]. Our study population also 

compares favorably to the participant profile in a study that 

reevaluated the Three-Point Checklist, of which 24 indi-

viduals (14%) lacked previous dermoscopy experience.9 The 

Three-Point Checklist was also evaluated in a prospective 

trial with 73 primary care physicians; however, the previous 

dermoscopic experience or training of participants was not 

previous dermoscopy training (43%, n=52) and/or experi-

ence (23%, n=28). This allowed us to evaluate the potential 

utility of TADA as a skin cancer detection aid for inexpe-

rienced dermoscopists, who comprise our target audience. 

Our study population had a greater sample of beginners 

than the pilot studies for the Three-Point Checklist and AC 

Rule, which included 6 inexperienced dermoscopists and 17 

TABLE 2. Model-based estimates of sensitivity and specificity for TADA for all study lesions. 
[Copyright: ©2017 Rogers et al.]

Variable

Sensitivity

Variable

Specificity

Coding Estimate (95% CI) P-value Coding Estimate (95% CI) P-value

Overall 94.6 (93.4—95.7) — Overall 72.5 (70.1—74.7) —

Diagnosis AMM 95.6 (91.5 -99.8) 0.942 Diagnosis Angioma 76.4 (72.6—80.4) <0.001

BCC 95.2 (92.9 -97.6) 0.651 CCA 39.1 (37.0—41.4) <0.001

MM 94.4 (92.3 -96.6) 0.251 DF 93.6 (90.0—98.1) <0.001

NM 91.6 (88.2 -95.1) 0.020 Nevus 69.4 (67.0—71.9) —

SCC 95.7 (93.8 -97.7) — SK 82.9 (79.2—86.7) <0.001

Previous 
Dermoscopy 
Training

No 93.6 (91.8—95.5) — Previous 
Dermoscopy 
Training

No 69.0 (64.6—73.7) —

Yes 95.4 (94.8—99.6) 0.14 Yes 73.2 (71.4—84.5) 0.450

Abbreviations: AMM, amelanotic melanoma; BCC, basal cell carcinoma; MM, malignant melanoma; NM, nodular melanoma; SCC, 
squamous cell carcinoma; CCA, clear cell acanthoma; DF, dermatofibroma; SK, seborrheic keratosis

Figure 3. Receiver operating characteristic (ROC) curves for TADA, the Three-Point Checklist, and AC Rule—ROC curves 

demonstrate the diagnostic performance of the three algorithms compared for the identification of pigmented skin lesions. 

[Copyright: ©2017 Rogers et al.]



Research  |  Dermatol Pract Concept 2017;7(2):9 45

ficities ranging from 91.0% to 96.3% and 32.8% to 71.9%, 

respectively, for the identification of pigmented melanoma 

and basal cell carcinoma [9,11]. The AC Rule has reported 

sensitivities and specificities of 94% and 62%, respectively, 

for pigmented melanoma.12 Two of the three criteria used 

in the Three-Point Checklist are for pigmented lesions and 

at least two of these criteria must be present for a lesion 

to warrant a biopsy. This greatly decreases the likelihood 

that the algorithm will identify nonpigmented malignancies. 

Regarding the AC Rule, the final determinant of whether or 

not a lesion requires a biopsy is the user’s level of suspicion, 

which reflects the presence of one or both of the algorithms 

criteria. One of these criteria, asymmetry, is not specific to 

pigmented lesions. The AC Rule thus might be applicable for 

nonpigmented lesions, however, only those with disorganized 

architecture. Similarly, the algorithm might miss organized, 

homogenously pigmented skin cancers, such as some nodular 

melanomas.

When comparing the results of TADA to that of the Three-

Point Checklist and AC Rule for the detection of pigmented 

lesions, we found that TADA performed with the highest 

sensitivity by as much as 22%. The fact that the sensitivity 

estimates for TADA were the same on the entire data set and 

on the subset of pigmented lesions (94.6% vs. 94.0%) might 

reflect the inclusion of sufficient dermoscopic criteria for the 

identification of both pigmented and nonpigmented skin can-

cers. Regarding specificity, we found that all three algorithms 

performed well. The specificity seen with TADA was slightly 

lower than for the other two algorithms. Certain features 

included in TADA, such as vessels of any morphology, will 

invariably lead to biopsies of some benign lesions, like clear 

cell acanthomas or some intradermal nevi. TADA knowingly 

sacrifices this specificity for simplicity and sensitivity. Addi-

tional training in the identification of certain benign lesions 

would likely increase TADAs specificity.

The results of this study suggest that TADA might be 

a useful triage tool by providing a simplified dermoscopic 

method with high sensitivitity for common skin cancers. 

However, our study population consisted of individuals 

attending a dermoscopy course and were thus a motivated 

group with an interest in learning dermoscopy. As such, our 

results might not be generalizable to all novices. To make 

definitive statements about the efficacy of TADA in clinical 

practice, our findings would need confirmation with a greater 

number of inexperienced dermoscopists evaluating a larger 

and more diverse sample of neoplasms with a more balanced 

proportion of pigmented and nonpigmented lesions. This 

would ideally be achieved with a prospective study. Since 

clinical-dermoscopic correlation can be crucial for certain 

diagnoses, the addition of gross images would more closely 

correlate to the clincal scenerio. Our use of clinical descrip-

tives (i.e., firm, keratotic) could have introduced an informa-

indicated [23]. A limitation of our study is that non-partici-

pant characteristics were not recorded and we are unable to 

report on any differences between participants (n=120) and 

non-participants (n=80).

The overall sensitivity of TADA for pigmented and non-

pigmented skin cancers was 94.6%. This value was margin-

ally influenced by participants’ previous dermoscopy training 

(95.4% vs. 93.6%). The first criterion included in TADA is 

architectural disorder, which is not an objective criterion in 

that it cannot be defined by any given shape or color. It is 

rather the result of the overall impression, or gestalt, of an 

asymmetric or chaotic lesion. The subjective interpretation 

of disorganization within a lesion has been shown to have 

better interobserver agreement than most objectively defined 

criteria [13]. It has also been shown to be one of the dermo-

scopic criteria with the highest discriminatory power [13-15]. 

Indeed in the present study, architectural disorder allowed for 

the correct identification of greater than 50% of malignant 

study lesions. In order to identify malignancies with ordered 

and symmetric appearances, participants needed to be able to 

recognize six additional features, three of which (blue-black 

or gray color, ulcer/erosion, and vessels of any morphology) 

are colors and structures not specific to dermoscopy and, in 

our experience, beginners have been able to quickly recognize. 

Facial, acral, nail, and mucosal lesions were not evaluated and 

the algorithm states that TADA cannot be used for lesions on 

these sites. While our results for non-melanoma skin cancer 

can likely be generalized to facial lesions, more robust studies 

across multiple ages and skin color cohorts are needed to vali-

date the dermoscopic features of early special sites melanomas.

Notably, untrained participants achieved an overall speci-

ficity of 69% using TADA. Additionally, the specificities for 

the three types of benign lesions included in the algorithm 

ranged from 76% to 94%. This finding substantiates our 

view that beginners can be quickly trained to accurately 

identify classic examples of certain benign lesions. In many 

instances, these benign neoplasms can have dermoscopic 

characteristics attributable to malignant lesions, such as the 

blue, black, or gray colors commonly observed in seborrheic 

keratoses or the shiny white structures or scar-like areas seen 

in dermatofibromas [24,25]. However, when these features 

are viewed in the context of the global lesion pattern as a 

whole, the diagnosis can become apparent. Additionally, 

the frequency with which these lesions are encountered in 

clinical practice can allow one to rapidly gain experience in 

their identification. While requiring users to gain additional 

dermoscopic knowledge in order to identify these lesions is 

arguably a limitation of TADA, it also seemed to strengthen 

the algorithm, as indicated by the high specificities achieved 

for these lesions.

Pre-selection of lesions is not something unique to TADA. 

The Three-Point Checklist has reported sensitivities and speci-



46 Research  |  Dermatol Pract Concept 2017;7(2):9

12. Luttrell MJ, McClenahan P, Hofmann-Wellenhof R, Fink-Puches 

R, Soyer HP. Laypersons’ sensitivity for melanoma identification 

is higher with dermoscopy images than clinical photographs. Br 

J Dermatol. 2012;167:1037-1041.

13. Carrera C, Marchetti MA, Dusza SW, et al. Validity and Reliability 

of Dermoscopic Criteria Used to Differentiate Nevi From Mela-

noma: A Web-Based International Dermoscopy Society Study. 

JAMA Dermatol. 2016;152(7):798-806.

14. Rosendahl C, Tschandl P, Cameron A, Kittler H. Diagnostic ac-

curacy of dermatoscopy for melanocytic and nonmelanocytic 

pigmented lesions. J Am Acad Dermatol. 2011;64:1068-1073.

15. Henning JS, Dusza SW, Wang SQ, et al. The CASH (color, archi-

tecture, symmetry, and homogeneity) algorithm for dermoscopy. 

J Am Acad Dermatol. 2007;56:45-52.

16. Lallas A, Moscarella E, Longo C, et al. Likelihood of finding 

melanoma when removing a Spitzoid-looking lesion in patients 

aged 12 years or older. J Am Acad Dermatol. 2015;72:47-53.

17. Argenziano G, Longo C, Cameron A, et al. Blue-black rule: a 

simple dermoscopic clue to recognize pigmented nodular mela-

noma. Br J Dermatol. 2011;165:1251-1255.

18. Balagula Y, Braun RP, Rabinovitz HS, et al. The significance of 

crystalline/chrysalis structures in the diagnosis of melanocytic and 

nonmelanocytic lesions. J Am Acad Dermatol. 2012;67:194.e1-8.

19. Bassoli S, Ferrari C, Borsari S, et al. Negative pigment network 

identifies a peculiar melanoma subtype and represents a clue to 

melanoma diagnosis: a dermoscopic study of 401 melanomas. 

Acta Derm Venereol. 2013;93:650-655.

20. Altamura D, Menzies SW, Argenziano G, et al. Dermatoscopy of 

basal cell carcinoma: morphologic variability of global and local 

features and accuracy of diagnosis. J Am Acad Dermatol. 2010; 

62:67-75.

21. Pizzichetta MA, Talamini R, Stanganelli I, et al. Amelanotic/

hypomelanotic melanoma: clinical and dermoscopic features. Br 

J Dermatol. 2004;150:1117-1124.

22. Luttrell MJ, Hofmann-Wellenhof R, Fink-Puches R, Soyer HP. The 

AC Rule for melanoma: a simpler tool for the wider community. 

J Am Acad Dermatol. 2011;65:1233-1234.

23. Argenziano G, Puig S, Zalaudek I, et al. Dermoscopy improves 

accuracy of primary care physicians to triage lesions suggestive 

of skin cancer. J Clin Oncol. 2006;24:1877-1882.

24. Braun RP, Rabinovitz HS, Krischer J, et al. Dermoscopy of 

pigmented seborrheic keratosis: a morphological study. Arch 

Dermatol. 2002;138:1556-1560.

25. Zaballos P, Puig S, Llambrich A, Malvehy J. Dermoscopy of der-

matofibromas: a prospective morphological study of 412 cases. 

Arch Dermatol. 2008;144:75-83.

26. Pagnanelli G, Soyer HP, Argenziano G, et al. Diagnosis of pig-

mented skin lesions by dermoscopy: web-based training im-

proves diagnostic performance of non-experts. Br J Dermatol. 

2003;148:698-702.

tion bias not reproducible in clinical settings, however, this 

limitiation would have more relevance to a study evaluating 

teladermoscopy. Further, in the present study, all participants 

received standardized training. Randomizing participants to 

various levels and durations of training would allow us to 

determine if the teaching modality for TADA can be stream-

lined. An upcoming study will address the latter limitation 

and also determine if the inclusion of step one (identifying 

common benign lesions) truly strengthens the algorithm.

References

 1. Kittler H, Pehamberger H, Wolff K, Binder M. Diagnostic accu-

racy of dermoscopy. Lancet Oncol. 2002;3:159-165.

 2. Bafounta ML, Beauchet A, Aegerter P, Saiag P. Is dermoscopy (epi-

luminescence microscopy) useful for the diagnosis of melanoma? 

Results of a meta-analysis using techniques adapted to the evalu-

ation of diagnostic tests. Arch Dermatol. 2001;137:1343-1350.

 3. Vestergaard ME, Macaskill P, Holt PE, Menzies SW. Dermoscopy 

compared with naked eye examination for the diagnosis of pri-

mary melanoma: a meta-analysis of studies performed in a clinical 

setting. Br J Dermatol. 2008;159:669-676.

 4. Argenziano G, Albertini G, Zalaudek I. Commentary: im-

proved detection of nonpigmented skin tumors. Dermatol Surg. 

2012;38:1445-1447.

 5. Nehal KS, Oliveria SA, Marghoob AA, et al. Use of and beliefs 

about dermoscopy in the management of patients with pigmented 

lesions: a survey of dermatology residency programmes in the 

United States. Melanoma Res. 2002;12:601-605.

 6. Engasser HC, Warshaw EM. Dermatoscopy use by US dermatolo-

gists: a cross-sectional survey. J Am Acad Dermatol. 2010;63:412-

419.

 7. Terushkin V, Oliveria SA, Marghoob AA, Halpern AC. Use of 

and beliefs about total body photography and dermatoscopy 

among US dermatology training programs: an update. J Am Acad 

Dermato.l 2010;62:794-803.

 8. Anderson RT, Dziak K, McBride J, Camacho F, Hege AC, Torti 

FM. Demand for continuing medical education programs on 

cancer care among primary care physicians in North Carolina. N 

C Med J. 2004;65:130-135.

 9. Zalaudek I, Argenziano G, Soyer HP, et al. Three-point check-

list of dermoscopy: an open internet study. Br J Dermatol. 

2006;154:431-437.

10. Robison S, Kljakovic M, Barry P. Choosing to biopsy or refer sus-

picious melanocytic lesions in general practice. BMC Fam Pract. 

2012;13:78.

11. Soyer HP, Argenziano G, Zalaudek I, et al. Three-point checklist 

of dermoscopy. A new screening method for early detection of 

melanoma. Dermatology. 2004;208:27-31.