BLINK – an algorithm to decide if a skin lesion requires biopsy to exclude skin cancer
- a preliminary trial to compare the 3-Point Checklist, the Menzies Method, clinical assessment alone with the BLINK algorithm
Peter Bourne M.B., B.S, FSCCANZ, Cliff Rosendahl M.B., B.S., FSCCANZ
Summary
Background: BLINK is an algorithm to help the primary care clinician decide whether to biopsy a suspicious skin lesion to exclude skin cancer. This algorithm is unique in that dermatoscopic assessment is pooled with other clinical information such as history of change, “ugly duckling” appearance or patient concern in a standardised scoring system.
Objectives: To test the accuracy of the Blink algorithm and compare this with the 3-Point Checklist, the Menzies Method and clinical assessment alone, in assessing 50 consecutive suspicious skin lesions encountered in an Australian primary care skin cancer practice.
Methods: Clinical and dermatoscopic images of 50 consecutively excised skin lesions, including 9 melanomas, were collected and examined by 4 participants: 2 primary care skin cancer doctors, one GP and a practice nurse undergoing dermoscopy training. The cases were assessed using the following methods: 3-Point Checklist, Menzies method, (only dermatoscopic images presented), clinical assessment, (only clinical images presented,) and Blink, (both clinical and dermatoscopic images presented). The participants were asked if the standard criteria for each of the 3 algorithms were present or absent, and to give a diagnosis of benign or malignant for the clinical assessment method. The overall sensitivity and specificity for each of the four methods were calculated as well as the mean number of melanomas excised.
Settings: Australian primary care skin cancer practice
Results: When comparing the four methods, the mean sensitivity for Blink was significantly greater than any of the other methods (Blink 90.8%; others 55.6%; P<.01). Of the 9 melanomas present in the series, the Blink method diagnosed 8.25 (mean), compared to means of 4.74 for 3-Point, 4.0 for Menzies and 2.25 for clinical. Differences between participants were not statistically significant in both cases.
Conclusions: This study indicates that the BLINK algorithm is a sensitive and reproducible skin cancer screening tool. The pooling of dermatoscopic assessment with clinical features in the history and examination appears to give a better overall result than just dermatoscopic or clinical assessment alone. The surprisingly poor performance of the 3-point checklist in this trial may relate to the prevalence of melanoma on sun damaged skin commonly seen in Australia.
Background:
In Australia, where the incidence of skin cancer is higher than any country in the world and where specialist dermatologist referral is often not practical, primary care doctors are required to manage skin cancer. The decision to excise or biopsy a suspicious skin lesion is often challenging for these generalist family physicians who traditionally receive minimal skin cancer training at an undergraduate or postgraduate level.
Dermoscopy has been shown to assist in the diagnosis of melanoma and non-melanoma skin cancer, most particularly if performed at an expert level. It would seem logical, therefore, that all primary care physicians in Australia undertake advanced training in this valuable diagnostic technique. Unfortunately, despite a significant increase in the availability of dermoscopy training courses in recent years, most generalist family doctors find it hard to allocate the necessary time to master this essential clinical skill. Indeed, even if mastered, dermoscopy alone does not always guarantee diagnostic accuracy. This may occur with lesions such as lentiginous melanoma, amelanotic or hypomelanotic melanoma and Merkel cell carcinoma.
In the authors’ experience, consideration of other features in the patient’s history and examination often assist in the diagnosis of difficult skin cancers. Surprisingly, however, no formal diagnostic algorithm appears to exist that combines these broader clinical aspects with dermatoscopic findings. Indeed, it could be said that the fervent pursuit of dermatoscopic excellence may have overshadowed a more traditional holistic diagnostic approach.
In an attempt to combine this broader clinical approach with a simplified, easily learned dermatoscopic analysis, the authors have devised a new algorithm for the assessment of difficult skin lesions. Of note is the fact that this new algorithm is applicable to all skin lesions, pigmented and non-pigmented, in contrast to most previous dermatoscopic algorithms.
The acronym, “BLINK”, refers to a logical progression of clinical questions that should be asked when presented with a suspicious skin lesion. This approach requires that the user have a fundamental knowledge of dermoscopy, specifically the ability to recognise:
• an asymmetrical pigmentation pattern
• the presence of more than one colour
• (for amelanotic lesions), the presence of an irregular vascular pattern
• eight simplified dermatoscopic criteria commonly seen in malignant lesions (see below)
Unlike the classical 2-step approach, the Blink algorithm does not require the clinician to decide whether the lesion is melanocytic or not. As well, an in depth knowledge of the standard algorithms such as the 3 point check list, 7 point checklist, Menzies method, ABCD rule of dermoscopy, Kittler variation of pattern analysis, classical pattern analysis or the CASH algorithm is not necessary.
B L I N K:
These six questions are asked in order:
B – Benign
o Is the lesion immediately recognisable as a common benign tumour, on clinical and dermatoscopic examination, with other similar lesions being present on that part of the body e.g. typical solar lentigo, seborrheic keratosis, haemangioma or dermatofibroma?
If the answer is ‘yes’, no further action is required.
If the answer is ‘no’, then proceed to the following:
L – Lonely
o Is this lesion, clinically and dermatoscopically, the only one of its type on that region of the body, i.e. an “ugly duckling”?
‘Yes’ scores 1
I – Irregular
o For pigmented lesions, is the lesion dermatoscopically irregular, that is, does it have an asymmetrical pigmentation pattern and more than one colour?
o For non-pigmented lesions, is there an irregular vascular pattern?
‘Yes’ scores 1
N – Nervous and/or Change
o Is the patient nervous that the lesion may be a skin cancer?
o Does the patient, or another observer, feel that the lesion is changing?
N.B. Only a total score of 1 can be given if either or both of these last two questions are answered ‘yes’
K – Known clues
o Does the lesion definitely have any one of the following 7 dermatoscopic “clues” to malignancy?
1) Network – distinct variation in thickness of network line
2) Pseudopods or streaks – segmental
3) Black dots, globules or clods – peripheral
4) Eccentric structureless zone
5) Blue or grey colour – irregular distribution
6) Vessels: polymorphous, finely focused arborising or glomerular
7) Acral lesions – parallel ridge pattern or diffuse irregular brown pigmentation
‘Yes’ scores 1
From clinical experience,
the authors feel that if the lesion scores 2 or more out of 4, then it should be biopsied. If there is a score of 1 or 0, the lesion does not require biopsy and may be either reviewed at a later time, digitally monitored, or the patient simply reassured that the lesion is benign. This will depend upon the clinical scenario.
Patient concern for a lesion, (“Nervous”), has been included for two reasons. Firstly, lesions that appear at first glance to be completely innocent are sometimes given only a cursory inspection by the examining doctor during a skin examination. The patient, however, may have additional information regarding this lesion that they do not volunteer to the doctor, such as a history of bleeding, itching or change. Hence, a patient’s concern over a lesion may represent more than just a so-called “sixth sense”.
Secondly, there is an medico-legal aspect to “patient concern”. In the unfortunate circumstance of a lesion being initially dismissed by the clinician as being benign, and then subsequently being excised and shown to be melanoma, the patient may prove to be less forgiving of the clinician if concern was specifically expressed over this lesion at the initial presentation.
The two questions labeled “Nervous” and “Change” have been grouped together in an attempt to improve the specificity of this method. One point is scored if either is present, but only 1 point is scored if both are present. This is because one of the common causes of patient concern for a lesion is, in fact, a subjective history of change. If two points were allocated in this situation, despite there being no other positive features, then all such lesions would require excision. In the authors’ opinion, this would result in too many benign lesions being removed.
With regards the situation where a skin lesion is assessed as showing one or more definite dermatoscopic features of malignancy, yet the lesion is judged to have a symmetrical pigmentation pattern, one colour, no “ugly duckling” features and no history of change, then the BLINK approach would lead the clinician to not biopsy this lesion, (score of 1). The authors feel that this is still the correct course of action as it is more likely that the assessed “criteria” for malignancy represents a false positive appraisal, rather than the lesion is in fact a cancer.
The eight listed “known” dermatoscopic clues to malignancy have been distilled from existing recognized criteria. Where possible, these criteria have been simplified to assist with ease of recognition by novice dermoscopists, e.g. “any grey colour in a facial lesion”. Only criteria judged to be of high sensitivity and specificity as well as demonstrating good interobserver agreement among student dermoscopists was included in this list.
Below are five examples from the authors’ own practices demonstrating the application of the Blink algorithm. In each case, the patient presented for a routine skin examination and was not aware of the lesion in question.
Case 1: An 82 year old male underwent a full skin examination as part of a routine annual checkup. A slightly raised, solitary, pink lesion was noted on his back. There were no symptoms and the patient was not aware of the lesion. Dermoscopy revealed an unpigmented lesion with mainly dotted vessels and a small area suggestive of regression at the inferior margin. There were no classical criteria for melanoma apart from subtle vessel irregularity. As lichen planus-like keratoses often have a similar appearance, the authors feel that this lesion may have been dismissed by many clinicians. Using the BLINK method, the lesion was assessed as not being classically benign, (B), as there were no other similar lesions present on that part of the body. The next five questions were considered and it was felt that the lesion was an “ugly duckling”, (L) and dermatoscopically irregular, with an irregular vascular pattern (I). The patient was not nervous about this lesion being a cancer, (N), and there was no history of change, (C). None of the eight dermatoscopic features of malignancy were unequivocally present, (K). This resulted in a score of 2 and the lesion was excised. Histology revealed an amelanotic level 2 superficial spreading malignant melanoma with active regression. It had a Breslow thickness of 0.44mm and was arising in a dysplastic compound naevus.
Case 1: Clinical view
Case 1: dermoscopy view
Case 1: Histology
Low power view shows a transition from dysplastic (Clark) naevus on the right to invasive melanoma on the left.
Invasive (level 2) malignant melanoma with a dermal mitosis evident centre field. There is a brisk lymphoid infiltrate and elsewhere there were changes of regression.
Case 2: A 41 year old woman was noticed to have a light brown macular lesion on the left upper arm, which was noticeably larger than any other lesion on this limb. She had multiple solar lentigines on the back but none appeared to be as large as this lesion. Hence, it was unable to be classed as definitely “benign”, (B), as per the above definition. Even though it was an “ugly duckling”, (L), the dermoscopic appearance was again remarkably bland. However, it was felt that the pigmentation pattern was irregular and that there were 2 colours, tan and dark brown, (I). There was no patient concern, (N), no history of change, (C), and no obvious dermoscopic features of malignancy, (K). This resulted in a score of 2 and the lesion was excised. Histology revealed a melanoma in situ arising in a lentiginous junctional dysplastic naevus. The author feels that the dermoscopic appearance of this lesion alone may not have been sufficient to lead to a diagnosis of melanoma and that the inclusion of other clinical aspects was helpful in deciding upon the correct course of action.
Case 2: Clinical view
Case 2: Dermoscopy view
Case 2: Histology
The low power view shows a broad junctional melanocytic proliferation that is predominantly nested to the left and lentiginous on the right.
High power view showing only a disordered lentiginous proliferation of large and somewhat pleomorphic melanocytes with Pagetoid epidermal invasion indicative of early melanoma-in-situ.
Case 3: An 80 year old woman presented for a routine skin check. An irregular, pigmented, macular lesion was noted on the right jaw area that was larger than her surrounding solar lentigines. This lesion was considered to be an “ugly duckling”, (L), as well as having an irregular pigmentation pattern and two colours, (I). The patient was not aware of the lesion, (N,C), and the author was unable to confidently describe any known dermoscopy features of malignancy, (K). As this lesion scored 2, it was excised and a diagnosis of in situ lentigo maligna melanoma with regression was made on histology.
Case 3: Clinical view
Case 3: Dermoscopy view
Case 3: Histology
There are clear cut changes of melanoma-in-situ of lentigo maligna type at both low and high power. At high power extension of melanoma-in-situ into a follicular infundibulum
is seen.
Case 4: A 71 year old man presented complaining of a clinically obvious keratoacanthoma on the neck. On full skin examination, a flat, pigmented, slightly scaly lesion was noted on the left lower leg. The patient had not noticed this before. Dermoscopically the lesion was not obviously benign (B) but was lonely (L) and irregular (I). The patient was not nervous (N) about the lesion and it was not reported to have changed (C). There were no obvious known dermatoscopic criteria (K) for melanoma. Hence the lesion scored 2 and was excised. Histology revealed a pigmented intra-epidermal carcinoma (Bowen’s disease) and level 1 (in-situ) malignant melanoma arising in a dysplastic naevus. This lesion could have easily been mistaken for a seborrheic keratosis or pigmented actinic keratosis but was biopsied after applying the Blink algorithm.
Case 4: Clinical view
Case 4: Dermoscopy view
Case 4: Histology
High power view showing the transition from the nested proliferation on the left to the disordered lentiginous proliferation indicative of melanoma-in-situ on the right.
Intraepidermal carcinoma (to the left) is colonized by lentiginous melanoma-in-situ (from the right). Pigmentation is visible within the area of the intraepidermal carcinoma colonized by melanoma-in-situ.
Case 5: A 34 year old man was noted to have a pigmented lesion on the right lower leg which he felt had not changed. His girlfriend however thought it may have slightly increased in size. Dermatoscopically, the darker area appeared to have “fat fingers” and the lighter area features of a solar lentigo, perhaps leading to a diagnosis of seborrheic keratosis. Applying the Blink algorithm, the lesion was not really a stereotypical benign (B) seborrheic keratosis and there were no similar lesions in the vicinity. It was an ugly duckling (L) and irregular (I). The patient was not concerned (N) and the history of change (C) was uncertain. There were no obvious known features of melanoma (K). Hence this lesion scores 2, or perhaps 3 if the history of change is included. The histology report stated that this was an usual lesion which had a very nested junctional pattern with some atypia but no true dysplastic naevus features. Towards one edge there were some pagetoid features indicative of an evolving level 1 melanoma.
Case 5: Clinical (image courtesy of Dr Alan Cameron)
Case 5: Dermoscopy (image courtesy of Dr Alan Cameron)
Case 5: Histology
Low power view shows a predominantly nested atypical melanocytic proliferation.
High power view shows a disordered junctional melanocytic proliferation with Pagetoid melanocytic invasion indicative of early melanoma-in-situ.
METHODS:
To assess the validity of this new algorithm, a preliminary trial was conducted comparing two commonly used traditional dermatoscopic algorithms, (the 3-Point Checklist and the Menzies Method), clinical assessment alone and the BLINK algorithm. Each method was used to decide whether suspicious skin lesions should be biopsied or not to exclude skin cancer. This was considered to be a more practical end point for primary care physicians than a specific diagnosis.
From 1st June to 6th July 2009, all skin lesions excised to exclude skin cancer were recorded by an experienced primary care doctor, (A.C.), working in a dedicated Australian skin cancer practice. Lesions that were easily recognisable as non-melanoma skin cancer on clinical grounds were not included in this series as it was judged that these tumours did not require the use of an algorithm to make a confident diagnosis. As well, lesions that were excised at the patient’s request for cosmetic reasons were not included.
The remaining 50 consecutive cases underwent photographic documentation with clinical and dermatoscopic views, (non-polarised dermatoscopic images taken with a non-polarised Dermlite Foto, or a Dermlite Fluid dermatoscope, and Canon D40 digital camera). Written patient consent was obtained in every case and any history of lesion “change” or “patient concern” was documented.
All 50 lesions were subjected to histological examination and 19 ultimately proved to be skin cancer:
• 9 melanoma (7 in situ and 2 invasive)
• 6 basal cell carcinoma
• 3 Bowen’s disease
• 1 keratoacanthoma
For the purposes of this trial, the keratoacanthoma was regarded as a squamous cell carcinoma and included in the skin cancer group.
Four participants, all with various degrees of dermatoscopic skill, were asked to review the cases using the four different diagnostic methods. Two participants were primary care doctors working full-time in skin cancer clinics, (P.B. and D.B.), one was a family GP with specific interest in skin cancer, (C.R.), and one a skin cancer practice nurse undergoing a dermoscopy training course, (H.C.). All were familiar with the 3 point checklist and the Menzies method and had read the explanation of the Blink algorithm presented earlier in this article. All 50 cases were assessed four times by the participants, each time using a different method. With the 3-Point Checklist and Menzies method, only dermatoscopic images were presented; with the clinical assessment alone, only clinical images were supplied; with the Blink algorithm, both clinical and dermatoscopic images were presented. The participants were asked to decide if the criteria listed for each method were present or absent, and enter this information into a formatted Excel spreadsheet. The diagnosis was then calculated automatically. A diagnosis of malignancy, or suspicion of malignancy, equated to a need for biopsy. For the purposes of the trial, all histologically proven skin cancers were deemed to require biopsy.
As the cases were assessed four times by each of the participants using four different methods, 16 contingency tables were obtained. Sensitivity and specificity were calculated for each. As two of the algorithms, (3-point and Menzies), related only to pigmented lesions, the five non-pigmented samples were excluded from the contingency table for these methods.
Analysis of variance was used to detect differences between the two factors, (methods and participants). The variables analysed were sensitivity, specificity, melanomas detected, total skin cancers detected, and number of biopsies indicated. When a factor was found to be significant, the least significant difference (LSD)1 was calculated. Any two means whose difference exceeds this value were declared significantly different.
RESULTS:
There were significant differences between the four methods for sensitivity (F3,9= 9.43, P< .01), specificity (F3,9= 4.03, P<0 .05), melanomas detected (F3,9= 17.50, P<.001), total cancers detected (F3,9= 16.64, P<.001), and number of biopsies indicated (F3,9= 5.18, P< .05). However, differences between participants were not statistically significant for any of the variables, (P>.05).
Table 1: Mean values of methods (with LSD) for sensitivity, specificity, melanomas found, total cancers and number of biopsies - means followed by the same letter in each column are not significantly different
Sensitivity
(%) Specificity
(%) Melanomas
found Total cancers found Number biopsies
3-point 59.4a 42.2a 4.75b 9.5a 26.3ab
Menzies 54.7a 69.0ab 4.00ab 8.8a 17.8a
Clinical 52.6a 74.8b 2.25a 10.0a 18.3a
Blink 90.8b 50.0ab 8.25c 17.3b 32.8b
LSD 21.5 28.4 2.23 3.5 11.6
As can be seen in Table 1, the overall sensitivity of the Blink approach was much higher than the other three methods – Blink achieving 90.8% and the others 55.6% (mean). The Blink approach resulted in a significantly higher number of biopsies than the Menzies or clinical methods but was able to correctly identify significantly more of the 19 skin cancers (mean 17.3), than the other three methods (mean 9.4). Of the 9 melanomas present, more were found using the Blink algorithm (mean 8.25) than the two dermoscopy algorithms (mean 4.37), with clinical alone method finding the least (2.25).
DISCUSSION:
The incidence of melanoma worldwide is increasing faster than any other cancer, with an annualized rate increase of six percent. In Australia, the melanoma incidence rates are around four times as high as those found in Canada, the US or the UK, with men having a 1 in 15 lifetime risk of getting a melanoma by age 85 and women 1 in 25.
More skin cancers are diagnosed and treated in Australia at a primary care level than by referral to specialist colleagues or public hospitals. These generalist primary care practitioners require a simple yet reliable skin cancer screening tool that will enable them to detect skin cancer at an early stage when complete cure is possible. Currently, most introductory skin cancer courses in Australia advocate dermatoscopic assessment of suspicious skin lesions using either the 3-point checklist or the Menzies method.
This preliminary trial indicates that the Blink approach, where clinical aspects are combined with simplified dermoscopy features, may be significantly better at detecting skin cancer than these other two ‘dermoscopy only’ methods. It is noteworthy that Blink found almost all the 9 melanomas in the series (mean 8.25), whereas 3-point checklist, which was designed to be a melanoma screening tool for non-experts, missed almost half (4.75). Interestingly, the sensitivity of the 3-point checklist (59.4%) in this small trial was well below that obtained in the published open internet study (91.0%).
As well, the specificity of the 3-point checklist obtained in the open internet study (71.9%), also dropped noticeably in this trial (42.2%). For a screening tool intended for use by non-experts, this is acceptable, as long as any sacrifice of specificity is rewarded by a high detection rate of melanoma. Unfortunately, the 3-point was unable to distinguish almost half the melanomas in this series, and hence its value in Australian conditions becomes questionable. In contrast, despite much the same specificity and number of biopsies indicated, the Blink approach was able to pick almost all the melanomas.
These results are somewhat surprising and beg the question: are Australian melanomas different to those seen in Italy, where the cases were collected for the two 3-point checklist trials? Perhaps the dermatoscopic criteria selected for the 3-point checklist are less relevant to the type of melanoma seen in Australia.
There is little data available regarding the prevalence of melanoma sub-types in Australia. However, in the authors’ experience, non-facial lentigo maligna-type melanoma, or lentiginous melanoma, presents quite frequently and can be difficult to diagnose on purely dermatoscopic grounds. In non-indigenous Australians, extensive solar damage on exposed skin is a common finding, no doubt due to both an Anglo-Celtic heritage and excessive UV exposure as a consequence of high latitude and atmospheric ozone depletion. As severely sun-damaged skin is often quite lentiginous, non facial lentigo maligna melanoma can be difficult to recognise dermatoscopically as the “network” may be impossible to distinguish from “finger-printing” seen solar lentigines. Consequently, the authors feel that clinicians may not always score the presence of the criterion, “atypical network”, when using the 3-point checklist. Despite most of these melanomas having the criterion “asymmetry”, not all will have the third criterion, “blue-white structures”, and hence a score of 1 may result, suggesting benignancy.
However, in the authors’ experience, these melanomas are usually “ugly ducklings” and will slowly change over time. An algorithm such as Blink, which takes these factors into account, may well show a higher sensitivity in these cases.
It is worth noting that the clinical assessment only method appeared to perform poorly with regard the number of melanomas found (2.25), and overall sensitivity (52.6%). The authors are not certain as to the significance of these figures as the assessment of skin lesions using two clinical photographs is unlikely to be as satisfactory as real life examination.
Finally, as the vast majority of skin cancers in Australia are non-pigmented, any skin cancer screening tool should be applicable to non-pigmented as well as pigmented skin lesions. Hence, while useful for the diagnosis of pigmented basal cell carcinoma and pigmented melanoma, the commonly taught diagnostic methods may not be ideal for a country like Australia, and newer approaches, such as Blink, should be explored.
Overall, the results of this small trial indicate that further studies are justified to determine whether this new algorithm, combining clinical and dermatoscopic criteria, should be recommended to all primary care doctors as a screening tool for skin cancer in Australia.
