Machine learning to predict overall short-term mortality in cutaneous melanoma

BACKGROUND

Cutaneous malignant melanoma (CMM) ranks among the ten most frequent malignancies, clinicopathological staging being of key importance to predict prognosis. Artificial intelligence (AI) has been recently applied to develop prognostically reliable staging systems for CMM. This study aims to provide a useful machine learning based tool to predict the overall CMM short-term survival.

METHODS

CMM records as collected at the Veneto Cancer Registry (RTV) and at the Veneto regional health service were considered. A univariate Cox regression validated the strength and direction of each independent variable with overall mortality. A range of machine learning models (Logistic Regression classifier, Support-Vector Machine, Random Forest, Gradient Boosting, and k-Nearest Neighbors) and a Deep Neural Network were then trained to predict the 3-years mortality probability. Five-fold cross-validation and Grid Search were performed to test the best data preprocessing procedures, features selection, and to optimize models hyperparameters. A final evaluation was carried out on a separate test set in terms of balanced accuracy, precision, recall and F1 score. The best model was deployed as online tool.

RESULTS

The univariate analysis confirmed the significant prognostic value of TNM staging. Adjunctive clinicopathological variables not included in the AJCC 8th melanoma staging system, i.e., sex, tumor site, histotype, growth phase, and age, were significantly linked to overall survival. Among the models, the Neural Network and the Random Forest models featured the best prognostic performance, achieving a balanced accuracy of 91% and 88%, respectively. According to the Gini importance score, age, T and M stages, mitotic count, and ulceration appeared to be the variables with the greatest impact on survival prediction.

CONCLUSIONS

Using data from patients with CMM, we developed an AI algorithm with high staging reliability, on top of which a web tool was implemented ( unipd.link/melanomaprediction ). Being essentially based on routinely recorded clinicopathological variables, it can already be implemented with minimal effort and further tested in the current clinical practice, an essential phase for validating the model's accuracy beyond the original research context.

Is pizza sutable to type 1 diabetes? A real life identification of best compromise between taste and low glycemic index in patients on insulin pump

Opposed to whole wheat (WWP), traditional pizza (TP) is loved by patients with type 1 diabetes mellitus (T1DM) despite causing hyperglycemia. 50 well-trained T1DM patients had higher glucose levels after TP than after WWP or mixed flour pizza, which however was tasty, digestible and metabolically appropriate to break diet monotony.

Differences in telomere length between sporadic and familial cutaneous melanoma

BACKGROUND

Several pieces of evidence indicate that a complex relationship exists between constitutional telomere length (TL) and the risk of cutaneous melanoma. Although the general perception is that longer telomeres increase melanoma risk, some studies do not support this association. We hypothesize that discordant data are due to the characteristics of the studied populations.

OBJECTIVES

To evaluate the association of TL with familial and sporadic melanoma.

MATERIALS AND METHODS

TL was measured by multiplex quantitative polymerase chain reaction in leukocytes from 310 patients with melanoma according to familial/sporadic and single/multiple cancers and 216 age-matched controls.

RESULTS

Patients with sporadic melanoma were found to have shorter telomeres compared with those with familial melanoma. In addition, shorter telomeres, while tending to reduce the risk of familial melanoma regardless of single or multiple tumours, nearly trebled the risk of single sporadic melanoma.

CONCLUSIONS

This is the first time that TL has been correlated to opposite effects on melanoma risk according to the presence or absence of familial predisposition. Individual susceptibility to melanoma should be taken into account when assessing the role of TL as a risk factor.

Adherence to treatment guidelines for primary sarcomas affects patient survival: a side study of the European CONnective TIssue CAncer NETwork (CONTICANET)

BACKGROUND

The impact of adherence to clinical practice guidelines (CPGs) for loco-regional treatment (i.e. surgery and radiotherapy) and chemotherapy on local disease control and survival in sarcoma patients was investigated in a European study conducted in an Italian region (Veneto).

PATIENTS AND METHODS

The completeness of the adherence to the Italian CPGs for sarcomas treatment was assessed by comparing the patient's charts and the CPGs. Propensity score-adjusted multivariate survival analysis was used to assess the impact of CPGs adherence on patient clinical outcomes.

RESULTS

A total of 151 patients were included. Adherence to CPGs for loco-regional therapy and chemotherapy was observed in 106 out of 147 (70.2%) and 129 out of 139 (85.4%) patients, respectively. Non-adherence to CPGs for loco-regional treatment was independently associated with AJCC stage III disease [odds ratio (OR) 1.77, P = 0.011] and tumor-positive excision margin (OR 3.55, P = 0.003). Patients not treated according to the CPGs were at a higher risk of local recurrence [hazard ratio (HR) 5.4, P < 0.001] and had a shorter sarcoma-specific survival (HR 4.05, P < 0.001), independently of tumor stage.

CONCLUSIONS

Incomplete adherence to CPGs for loco-regional treatment of sarcomas was associated with worse prognosis in patients with non-metastatic tumors.

Histopathological characteristics of subsequent melanomas in patients with multiple primary melanomas

BACKGROUND

Multiple primary melanomas (MPM) occur in up to 20% of melanoma patients, and subsequent tumours seem to have a favourable histopathological pattern.

OBJECTIVE

A prospectively collected cohort of 194 patients with MPM was retrospectively reviewed to investigate clinical and histopathological features of first and subsequent melanomas.

METHODS

Patients with MPM who were diagnosed at our Department (1985-2011) and who attended at least a follow-up control yearly were identified.

RESULTS

The number of nevi was <10, 10-50 and >50 in 8.7%, 41% and 50.3% of patients respectively. Histopathological dysplastic nevi have been diagnosed in 105 patients. During a median follow-up of 58 months, 159 (81.9%), 24 (12.3%), 7 (3.6%) and 4 (2%) patients developed 2, 3, 4 and ≥ 5 melanomas, respectively. The median time to second primary melanoma was 45 months. The second primary melanoma was diagnosed within 1-year and after 5-year from the first melanoma in 36.6% and 17.3% of patients respectively. First and second primary melanomas were in situ in 41 (21%) and 104 (54%) patients respectively (P < 0.001). Among patients with ≥ 2 invasive melanomas (N = 80), median tumour thickness and ulceration of first and second primaries were 0.91 and 0.44 mm (P <0.001), and 32% and 7.7% (P = 0.001) respectively.

CONCLUSIONS

Subsequent melanomas occurred within 1-year from the appearance of the first melanoma in 36% of patients with MPM, while a late melanoma diagnosis was detected in 17% of cases. Second primary melanoma had favourable histopathological features. Our findings support long-term skin surveillance to detect subsequent melanomas at an early stage.

Sarcoma: concordance between initial diagnosis and centralized expert review in a population-based study within three European regions

BACKGROUND

Sarcomas represent a heterogeneous group of tumors. Accurate determination of histological diagnosis and prognostic factors is critical for the delineation of treatment strategies. The contribution of second opinion (SO) to improve diagnostic accuracy has been suggested for sarcoma but has never been established in population-based studies.

METHODS

Histological data of patients diagnosed with sarcoma in Rhone-Alpes (France), Veneto (Italy) and Aquitaine (France) over a 2-year period were collected. Initial diagnoses were systematically compared with SO from regional and national experts.

RESULTS

Of 2016 selected patients, 1463 (73%) matched the inclusion criteria and were analyzed. Full concordance between primary diagnosis and SO (the first pathologist and the expert reached identical conclusions) was observed in 824 (56%) cases, partial concordance (identical diagnosis of connective tumor but different grade or histological subtype) in 518 (35%) cases and complete discordance (benign versus malignant, different histological type or invalidation of the diagnosis of sarcoma) in 121 (8%) cases. The major discrepancies were related to histological grade (n = 274, 43%), histological type (n = 144, 24%), subtype (n = 18, 3%) and grade plus subtype or grade plus histological type (n = 178, 29%).

CONCLUSION

More than 40% of first histological diagnoses were modified at second reading, possibly resulting in different treatment decisions.

Sentinel node biopsy and ultrasound scanning in cutaneous melanoma: clinical and technical considerations

1.5 mm and in all cases with two metastatic SNs, further positive additional nodes were found. The mean counts per 10 s (CP10S) ratio for SN and non-SN values was 5.62 (1.29-23.51) and 3.09 (1.03-10.99) in the intra-operative and extra-operative phases, respectively. US scanning and preoperative lymphoscintigraphy associated with PBD allows preoperative patient selection and accurate SN(s) identification. Breslow thickness and the number of metastatic SN(s), but not their type, are correlated with disease spread; CP10S contributed to the differentiation amongst the nodes and the determining of procedure's completion.

Early detection of melanoma: an educational campaign in Padova, Italy

To evaluate a public campaign for the early referral and treatment of cutaneous melanoma, an educational programme based on self-selection by subjects was organized in Padova, Italy in 1991. In the period from 1991 to 1996, 90,000 leaflets containing information on naevi, melanoma and skin self-examination were mailed to each household, reaching a population of 243,000 subjects. A total of 2050 individuals requested a skin check as a result of the leaflet. Most were at low risk, the majority being female (68%) and aged under 40 years (51.6%), with no risk factors (58.3%). One hundred and ninety subjects were referred for surgery for pigmented and non-pigmented suspect lesions. Histological diagnoses, obtained for all lesions, comprised 13 melanomas, 17 dysplastic naevi, 17 basocellular carcinomas, 140 pigmented benign lesions and three lesions of other types. The percentage of thin melanomas (< 1.50 mm) was 92.3%. Three hundred and fifty patients considered at risk at the first skin examination attended regular follow-up examinations. The sensitivity and predictive positive value of the visual examination were 92.8% and 6.8%, respectively. The impact of this campaign was evaluated in the Local Health District of Padova, comparing data from the pre-campaign period (1987-1990) with those from the campaign period (1991-1996); a trend towards a lower stage was observed (mean thickness 2.0 mm versus 1.50 mm; P < 0.02).

Hyperthermic antiblastic perfusion with alpha tumor necrosis factor and doxorubicin for the treatment of soft tissue limb sarcoma in candidates for amputation: results of a phase I study

To improve the therapeutic effectiveness of hyperthermic antiblastic perfusion (HAP), the association of recombinant tumor necrosis factor alpha (rTNF alpha), doxorubicin, and true hyperthermia (41 degrees C) was employed for the treatment of soft tissue limb sarcoma. A dose-escalation study according to Fibonacci's modified scheme was conducted, starting with a rTNF alpha dose of 0.5-3.3 mg. The doxorubicin doses (0.7 and 1.4 mg for the upper and lower limbs, respectively) and temperature level (41 degrees C) remained unchanged. Eighteen patients have been treated thus far: 9 males and 9 females of a mean age of 33 years (range: 24-71 years). The tumor was located in the upper limb in one patient and in the lower limbs in seventeen. Only 16 patients were evaluable, as 2 refused further treatment after the perfusion. In terms of local toxicity, a grade I limb reaction was observed in 3 patients, a grade II or III in 10 patients, and a grade IV in 5 patients, showing a strict correlation between the TNF dose and the grade of limb reaction. In fact, a grade III-IV limb reaction was observed in 66.6% of the patients treated with > 1 mg of rTNF alpha. The maximum tolerable dose in association with doxorubicin and true hyperthermia (41 degrees C) was 2.4 mg. Eleven patients showed a good pathological response (> 75%) and five patients showed a partial response (> 25%-< 75%). In no case was stable or progressive disease observed. The postperfusional tumor shrinkage permitted limb-sparing surgery in 75% of the patients, all of whom were candidates for amputation before HAP. No recurrences have been observed thus far. Two patients developed regional disease: one presented with a skip femur metastasis that disappeared after radiotherapy and systemic chemotherapy; the second developed regional node involvement, requiring a radical node dissection. Another patient had pulmonary metastases, 2 months after the HAP, which were resected. At a median follow-up of 12 months, all the patients are living without disease. The results of this phase I study suggest that the association of rTNF alpha, doxorubicin, and true HAP (41 degrees C) by regional perfusion is feasible and safe at a maximum tolerable rTNF alpha dose of 2.4 mg. However, because no correlation was found between the amount of rTNF alpha and the tumor response, 1 mg is recommended as the dose able to provide a high tumor necrosis rate and low local and systemic toxicity. This association appears to play an important role in the neoadjuvant treatment of soft tissue limb sarcoma.

Management of cutaneous melanoma M0: state of the art and trends

This article reviews the epidemiology, diagnosis and treatment of cutaneous melanoma, including the most recent developments. The combination of positive family history, fair complexion, number of nevi, exposure to sun and/or chromosomal alterations seem to be implicated in the pathogenesis of cutaneous melanoma. Melanomas can be classified according to their growth patterns, and tumour microstaging is of straightforward predictive value for survival and risk of metastasis, although new factors are also being investigated. As yet, surgical excision is the only effective treatment available for primary tumours, resection margins varying according to tumour thickness. Elective node dissection is, however, no longer advocated for melanomas thinner than 1.5 mm, and there is disagreement as to its role for thicker lesions. In contrast, selective node dissection at the time of definitive surgery is becoming more widely accepted, with regional node dissection being restricted to positive cases. Therapeutic dissection is required for lymph node involvement, the most common pattern of recurrence from melanoma, which affects nearly 30% of all patients. Complete remission rates from isolated limb perfusion, which has been employed in patients with multiple recurrences or in-transit metastases, range from 40 to 90%, depending on drugs and techniques used in different series; the best responses so far have been obtained with tumour necrosis factor in combination with melphalan. Patients with thick lesions (> 4 mm) or lymph node metastases have a high risk of micrometastases that would warrant adjuvant therapy. The only agent found to affect survival is interferon alpha-2.

The impact of ultrasound scanning in the staging and follow-up of patients with clinical stage I cutaneous melanoma

In 85 patients with intermediate to high risk cutaneous melanoma, concomitant palpation and ultrasound scanning (US) of the axillary and inguinal sites were performed in order to detect any locoregional lymph node metastases at pre-operative staging and postoperative monitoring. At 12 months follow-up, US identified 12 out of 13 (92%) histologically proven metastases, while palpation indicated metastatic disease in only 3 (23%) patients. Metastases were intranodal in 6 out of 12 cases detected with US, and extranodal in all the 3 cases identified by palpation, thus confirming that US is more effective than palpation in the early detection of lymph node metastases from melanoma. US was also more effective in discriminating all non-neoplastic lymph nodes: its overall specificity was 100% versus 85% for palpation. Thus, when carried out by well-trained radiologists, US is a very useful diagnostic tool for the surgical oncologist dealing with melanoma patients.

Proliferating cell nuclear antigen (PCNA) and recurrence in patients with cutaneous melanoma

A positive correlation between PCNA and the most important histoprognostic factors of cutaneous melanoma has been demonstrated. The aim of our work was to evaluate the efficacy of PCNA in predicting melanoma recurrence and to compare it with that of Breslow thickness. One-hundred and fifteen patients (75 women, 40 men; mean age 50 years) with primary cutaneous melanoma were retrieved. pTNM stages were as follows: stage I, 54 patients; stage II, 31 patients; stage III, 26 patients; and stage IV, four patients. The mean follow-up period was 55 months (range 2-260). Six patients developed lymph node metastases and 28 developed distant metastases; 27 patients died within 2-202 months from diagnosis. Tumour thickness was re-evaluated for each case. PCNA immunostaining was performed using the avidin-biotin complex method and the percentage of PCNA-positive tumour cells was indicated as the PCNA index. In order to evaluate and compare the PCNA index and Breslow thickness as predictors of recurrence, the receiver-operating characteristic (ROC) curve method, based on true-positive and false-positive rates was used. The PCNA index showed the highest true-positive rates and the lowest false-positive rates in the 5-30 interval. The PCNA index optimal cut-off is 20, characterized by 70% sensitivity and 80% specificity; Breslow thickness optimal cut-off is 3.5 mm, with 40% sensitivity and 90% specificity. Our results indicate that the PCNA index has a higher efficacy in predicting locoregional and distant recurrences in patients presenting primary cutaneous melanoma.

Phase II study on neoadjuvant hyperthermic-antiblastic perfusion with doxorubicin in patients with intermediate or high grade limb sarcomas

BACKGROUND

Locoregional control of soft tissue sarcomas of the limbs is achieved generally using a multidisciplinary approach consisting of conservative surgery combined with radiation therapy, intraarterial chemotherapy, or hyperthermic antiblastic perfusion (HAP). Before surgery, HAP seems to be the more suitable tool in decreasing tumor mass and allowing limb-sparing surgery. The authors' aim was to ascertain the activity of HAP with doxorubicin against intermediate or high grade limb tumors.

METHODS

In 23 patients with limb sarcomas (2 patients International Union Against Cancer Stage IIA, 4 stage IIB, 1 stage IIIA, 11 stage IIIB, and 5 stage IVB) doxorubicin was administered via HAP 4-6 weeks before surgery. The drug (bolus, 0.7-1.4 mg/kg) was perfused for 60 minutes with a tumor temperature of at least 40.5 degrees C (range, 40.5-42.6 degrees). Tumor necrosis was then assessed radiologically and pathologically.

RESULTS

Systemic toxicity was hematologic grade (G) 2 in 2 patients, gastrointestinal (hepatic) in 6, G1 in 2, G2 in 3, and G3 in 1; 2 patients had alopecia; locoregional toxicity (graded according to Wieberdink) was G1 or G2 in 18, G3 in 4, and G4 in 1. Tumor necrosis was more than 50% in 17 patients (74%). Limb-sparing surgery was feasible in 20 patients (91%). At present, 14 patients are alive. Six had local recurrences, and eight had distant metastases.

CONCLUSIONS

Our findings show that HAP with doxorubicin is an active and well-tolerated procedure within a multidisciplinary approach to treatment of limb sarcomas.

[A comparison between echography and computed tomography in assessing neoplastic recurrences in superficial soft tissues]

Thirty-seven superficial soft-tissue recurrences were evaluated with ultrasonography (US) and computed tomography (CT) to assess the correct diagnostic approach. US and CT examinations were performed at the same time. High-frequency US probes and a third-generation CT scanner were employed; all the lesions underwent also histology or cytology. US correctly identified as recurrences or fibrous tissue all the 37 lesions, whereas CT diagnosed 30 lesions only. Seven of the 14 recurrences < 2 cm diameter were not demonstrated. In conclusion, US provides more reliable information than CT relative to small lesions, which suggests that US must be performed just after therapy. Nevertheless, when bone involvement is suspected, CT is required and its use is also suggested to monitor distant metastases.

Adriamycin in hyperthermic perfusion for advanced limb sarcomas

From February 1989 to April 1991, 12 hyperthermic limb perfusions (HLP) with adriamycin (ADR) were performed in 12 patients with high grade soft tissue sarcomas (9 in the leg and 3 in the arm); two were at A.J.C. stage IIb, eight stage IIIb and two stage IV. ADR (0.7-1.5 mg/kg) was administered in bolus at a mean temperature of 41.5 degrees C and perfused for 60 min. No systemic toxicity was reported; seven patients had grade II locoregional toxicity, another four grade III and one grade IV. Tumor necrosis (radiological evaluation) was less than 25% in two patients, 25% to 50% in two, 50% to 75% in five and greater than 75% in three. Limb sparing surgery was feasible in ten patients. At present eight patients are alive; three had local recurrence and four distant metastases.