Акценти от 1-ви ден

Genetic susceptibility to melanoma, Role of computers in dermatology, Histopathology in the diagnosis of melanoma, 8th edition AJCC staging, Unresolved questions about melanoma


Dermatologist, Rennes University Hospital, France

This first day of the EADO Congress in Barcelona was very eventful. I really appreciated the overall atmosphere which was warm and friendly.

I must admit that this is my first EADO Congress. I'm more in the habit of attending congresses on dermatology or cancer in the broad sense. As a dermatologist specialising in skin cancer, this congress is fully in line with my centres of interest! With my co-reporter, Dr Oriol YÉLAMOS, an onco-dermatologist here in Barcelona, we divided up the topics to report on, and I'd planned to provide you with a summary of the first plenary session. Unfortunately, the first speaker, Dr Salvador AZNAR, had to cancel his presentation for today, which meant we were unable to attend his talk that had such an attractive title: "Metastatic-initiating cells and fatty acid metabolism."

He was replaced at the last minute by Pr Susana PUIG, Head of the Dermatology Department of Barcelona Hospital, who gave a presentation relating to one of her areas of expertise: genetic susceptibility to melanoma.

She started by reiterating that of all cancers, melanoma has the greatest genetic inheritability, as demonstrated by a prospective study of more than 200,000 identical and fraternal twins (Mucci et al., JAMA 2016.).

However, the majority (78%) of families predisposed to melanoma do not yet have a clear genetic explanation. Of the families in which at least one causal mutation has been identified (22%), the vast majority have a CDKN2A mutation (20% of families). A CDKN2A germline mutation, although very uncommon in the general population, very sharply increased the risk of developing several forms of melanoma, especially SSM, the first occurring at a relatively young age but with a high lifetime risk, on areas with a relatively lower degree of sun exposure (Ingvar et al., JID 2016.). No specific dermoscopic patterns were identified in familial melanomas with CDKN2A mutations. With metastatic melanoma, the response rate with immunotherapy seemed better in this genetic context than in the general population, but this result was achieved with a small study population (11 responding patients/19 patients with CDKN2A mutations, i.e. a 60% response rate, and 6/19 complete responses, i.e. 32%). These positive responses seemed correlated with a more frequently high mutational load than for general melanomas.

Lastly, families with CDKN2A mutations are predisposed not only to melanoma but also to pancreatic, breast and lung cancer (Potrony et al., JAAD 2014.). It is therefore essential that the doctors who monitor these families address the topics of smoking cessation and breast-cancer screening, in addition to providing standard recommendations on sun protection and ongoing dermatological monitoring.

Other genes highly predisposed to melanoma (BAP1, POT1, TERT, CDK4, etc.), while often cited with CDKN2A, are mutated much less frequently than CDKN2A. It is nonetheless important to think of the BAP1 gene with combinations of skin melanoma/uveal melanoma/mesothelioma/kidney cancer. Families with BAP1 mutations are also predisposed to breast cancer, basal cell carcinoma, and especially "BAP-omas", which are papular achromatic/pinkish naevi with irregular vessels within their borders in dermoscopy. The achromatic appearance of BAP-omas is due to the fact that the loss of BAP1 in melanocytes leads to a loss of pigmentation.

In Ashkenazi Jewish families, the main predisposing gene for melanoma is not CDKN2A but rather POT1, also associated with thyroid, breast and kidney cancer. MITF mutations are classified as "intermediate" in terms of the frequency, penetrance and occurrence of melanoma, half-way between the extreme rarity/dangerousness of CDKN2A mutations and the relative "banality"/safety of MC1R "red hair" alleles. These MITF mutations seem to promote a large number of reticulated naevi/melanomas in dermoscopy, sometimes with very rapid growth, and are not mutually exclusive with CDKN2A.

Lastly, Pr PUIG spoke about the MC1R gene and its "red hair" alleles, which are common in the general population and increase the risk of melanoma much more moderately than the aforementioned mutations. Different variants of this MC1R gene seem to not only provide the red-hair phenotype and freckles, but also slightly increase the risk of developing Parkinson's disease or Alzheimer's.

In conclusion, heredity plays a major role in the development of melanoma, even though most of the families explored in oncogenetics do not yet have a clearly identified predisposing gene. The increased availability of broad somatic and germline analyses is already enabling the identification of new candidate mutations, inherited from parents, in genes involved in cancer. The challenge for the coming years will be to formally and rapidly demonstrate the causal relationship between these new mutations and predisposition to melanoma.


Dermatologist, Hospital Clínic & Centro Médico Teknon, Barcelona, Spain

SYMPOSIUM: Controversies & new development in melanoma



Will the computers replace the dermatologist?


Computers will replace dermatologists at a certain point of time since now all the elements are in place to have this happening. Computers need to be powerful and need to get trained with large numbers of images to perform automatic diagnosis. However, current computers are very powerful, computer algorithms can diagnose correctly (results from Esteva et al in Nature 2017 showed that machines are at least equal to dermatologists in diagnosing skin lesions.), and many public databases exist so algorithms can learn how to diagnose and are fed with large databases.

It is important to highlight that some differences in algorithms of artificial intelligence exist: machine learning first needs someone to tell the machine what it is what, and then deep learning learns on its own. However, both methods have shown that they can do equally or better than dermatologists.

Another advantage of computers is that they can increase accessibility of specialists or to correct diagnosis to people who live in areas that are not much populated or areas that are do not have available dermatologists. For example, there are almost entire states in the USA where there are NO dermatologists, which is a problem. Hence, artificial intelligence will increase equity and access to correct care.


Yes: computers will replace some boring tasks done by dermatologists.

Lack of dermatologists will continue.

Better technology will come in the future.


The diagnosis of the melanoma is complex and sometimes computers cannot integrate all the information necessary for the diagnosis of skin cancer. Computers, in order to diagnose dermatologic conditions they need a lot of data, and for some skin diseases the data is limited, and also some of the data is not good (ie poor quality images). Also, humans can learn with very small amounts of data, and computers cannot. This is especially very important for rare conditions where not a lot of images or data are available. Also, the data that feeds databases or artificial intelligence algorithms can be preselected and therefore a bias can be introduced in these algorithms. Additionally, computers cannot take the responsibility of mistakes. So who is going to take the moral responsibility? Finally, computers cannot recognize emotions, and sometimes the diagnosis of a given condition depends on the gestalt of the physician, and therefore the gut feeling of dermatologists cannot be replaced by computers.


No: Today computers cannot replace computers. Deep learning is far from human brain capability Many legal and ethical considerations


Is the pathology the right gold standard for the melanoma?


The vast majority of melanocytic lesions are easily diagnosed with clinical skills. Then, surgery is generally restricted to atypical lesions which can be most of the time easily diagnosed with pathology.

Also, pathology seems the only way to diagnose which are complicated. Examples of lesions that can only be diagnosed with pathology include:

- Spitzoid lesions

- Desmoplastic melanoma

- Amelanotic melanoma

- Lentigo vs lentigo maligna melanoma

- Persistent melanoma in a scar of previous surgery

- Partially or totally regressed melanomas

Also, pathology has prognostic information such as establishing the Breslow depth. So yes, pathology is still the gold standard to diagnose melanoma.


Pathology diagnosis sometimes can be different among different dermatopathologists which can cause important problems to the physician and anxiety to the patient. Many aspects influence when a physician performs a diagnosis: knowledge, experience, medicolegal issues… Depending on this, dermatopathologists can render different diagnosis when evaluating the same exact case.

Also, sometimes some lesions are in the gray zone: what makes the difference between a severely dysplastic nevus vs a melanoma in situ? Sometimes the only way to know whether a lesion is benign or not is see what happens with the lesion, since lesion biology can be different even in lesions looking similar. Another typical example is the distinction between atypical spitz tumor vs spitzoid melanoma. Sometimes the only way to know is following the patient and identifying whether the patient will develop metastases or not.

To solve these issues there are new trends which will help melanoma diagnosis and will show that pathology is NOT the gold standard:

- Genomic classification of lesions

- Artificial intelligence: it will also be in pathology diagnosis.

So, to summarize, the gold standard is the clinical-pathologic correlation.


Is the new AJCC a progress?


AJCC 8th edition data reflects contemporary management of the melanoma. This is because AJCC should serve as a communication tools. Some changes have occurred in the last version of AJCC - For more about the changes check the paper by Gershenwald et al. CA Cancer J Clin. 2017 Nov;67(6):472-492 (PMID 29028110):

- The T changed:

o the thickness is recorded to nearest 0.1mm since it’s very difficult to measure less than 0.1 mm.

o Mitotic rate doesn’t change the T but should also be reported.

o T category changes at 0.8 mm

- Also, the new guidelines try to adapt to the advent of adjuvant therapy and the presence of clinical trials.

- Also, according to the results of the MSLT-II results, now the guidelines and the NCCN 2018 guidelines also include the option of node dissection but ALSO ultrasound follow-up in cases with positive SLNB.

- N3 expanded, whereas N1a/b and N2a/b have not changed.

- A grid has been created to facilitate the staging process.

- However, in the new guidelines the biomarkers are not mentioned since there is not enough data to suggest its use.

- In included LDH measurement in the M category, as well a new anatomic area which includes the central nervous system (M1d)


Grob The new AJCC is not a progress at all, especially for stage III. The problem is that AJCC is mostly an epidemiological tool, which by the way is already obsolete: the majority of people will be in the AJCC low risk group, but more people will die in this group because more people are in that category.

AJCC is not an improvement as a prognostic marker! AJCC may be useful in cases in stages I and II, as well as for stage IV, for stage III it is very messy. Also, AJCC 8th is counter-intuitive since IIC are doing worse than IIIA! However, this shows indirectly that SLNB may not have a very important prognostic value.

Also, if we allow adjuvant treatment in IIIA, then we need to try adjuvant therapy in stage IIA-IIB since they have a similar prognosis to IIIA. As a conclusion, it would be a good idea to keep AJCC 7th for stage III, until the revolution of biomarkers comes.

SYMPOSIUM: Controversies & open questions in melanoma



Shall all Spitz-like tumours be excised?


Since spitz nevi can be very irregular and spitzoid melanomas can be regular, at least from a clinical perspective let’s excise them all. Why? Because 13% of spitzoid symmetrical lesions after the age of 12 they are melanomas! Also although melanomas in children are rare, around 25% were spitzoid and symmetrical Also, more than 25% of benign Spitz nevi will behave in a weird way and will become irregular, leading to excision. Therefore, it seems reasonable to excise all spiztoid lesions.


Most guidelines do not recommend excising spitzoid lesions in children, especially in < 12 years, since it is known that the behavior of these lesions will be benign. These numbers are also more important if the lesions are symmetrical.


We have biomarkers for treatment decisions.

Both doctors more or less agreed that there are not a lot of single biomarkers to determine which treatment works best, but the future will be combining many clinical and biomarkers.


Biomarker may allow administering the right drug to the right patient, maximizing the benefit. For other cancers there are some biomarkers that are useful. However, nowadays there are not a lot of biomarkers for melanoma. Today the only one used in melanoma is BRAF.

Immunomarkers: PD1 and the rest are not yet validated in melanoma. PDL1 not useful, but it may be good when combined with other immunomarkers.

Clinical biomarkers can be useful also to predict response to treatment: for example, if brain or liver metastases some drugs may work better than others.

- If brain metastases, it is beneficial to administer combined ipi + antiPD1

So probably the future will be combining clinical and molecular markers, as well as gene expression profiling data.


Only one biomarker available: BRAF. That’s it.

Expression of PD1 or PDL1 don’t make much difference when checkpoint inhibitors are administered.

Also, high or low tumor mutation burden (TMB) does not seem to be related to response.

So it’s not here yet, but the future is combination of multiple markers to guide treatment.


Do we still need surgery in advanced melanoma?


Approximately 45% of patients who have macroscopic nodal disease are alive at 10 years (from AJCC database). Intralesional therapies have similar effect (around 50% response).

With stage 4 patients, some retrospective from MD Anderson have shown that surgical candidates did better, but there is obviously selection bias. When looking at randomized trials such as MSLT-1 for stage 4 patients the mea n survival for surgery was 17.6 months of medial survival vs no surgery that was 7.5 months.

Surgery also can be very important in residual tumor after targeted treatment.


More than surgery not being useful, what needs to be done is integrate all the treatments:

Initially treat them with targeted therapies and checkpoint inhibitors and then later do surgery on them.