Malignant melanoma of unknown primary site. To make the long story short. A systematic review of the literature

1. Introduction 

Malignant melanoma is a potentially lethal neoplasm and a major public health concern. The incidence of cutaneous melanoma has increased significantly in the last three decades [1] while the lifetime risk for cutaneous melanoma development is one in 40 males and one in 60 females [2]. In 2008 more than 64,000 new cases and more than 8000 deaths from malignant melanoma were estimated to occur in the US [2].

Although more than 90% of melanomas have a cutaneous origin, occasionally it is discovered first as a secondary deposit, lymph node or visceral, without evident primary site. This entity of melanoma of unknown primary (MUP) was initially characterized by Das Gupta in 1963 who was the first to describe the criteria for MUP [3]. Indeed, Das Gupta excluded patients from his study fulfilling any of the followed described criteria: (1) patients with metastatic melanoma with evidence of previous orbital exenteration or enucleation; (2) patients with metastatic melanoma with history of excision, electrodesiccation or cauterisation of a mole, birthmark, freckle, chronic paronychia or skin blemish; (3) patients with metastatic melanoma in one of the node-bearing areas who presented a scar of previous local treatment in the skin area drained by this lymphatic basin; and finally (4) patients with metastatic melanoma who did not have a thorough physical examination including ophthalmoscopy and adequate examination of the anus and the genitalia. Nevertheless, the first reports of MUP come from registries dated as far back as 1917 [4]. To explain the phenomenon of MUP various theories have been developed such as the spontaneous regression of the primary site due to involvement of immunological mechanisms [5], [6].

The incidence of MUP varies considerably between the different studies, ranging from 1.2% [7] to 18% [8] or even 31% in special settings [9]. Moreover, there are conflicting results concerning survival differences between MUP and melanoma of known primary (MKP), therapeutic interventions and prognostic issues.

Although various reports exist concerning MUP, a comprehensive review on this field is missing. Herein we describe the characteristics of patients with melanoma of unknown primary as was revealed from the search of literature expanding in a period of more than 80 years. The demographics, therapeutic methods, survival differences and prognostic factors are presented.

2. Materials and methods 

We searched MEDLINE (last search, February 2010) using combinations of terms such as Melanom*, unknown origin, unknown primary, indolent, occult. We set no geographical restrictions. Reference lists in these trials were checked to identify any other published or unpublished data. We hand searched the references of review articles and evaluated symposia proceedings, poster presentations, and abstracts from major cancer meetings (including American Society of Clinical Oncology Annual Meetings, International Melanoma Congress, European Society of Medical Oncology/European CanCer Organisation Congress and Annual Cancer Symposium of society of surgical oncology). We also hand searched [10] the last 4 years of Annals of Oncology, Journal of Clinical Oncology and Melanoma Research. Cross-searches were performed in MEDLINE using the names of investigators who were lead authors in at least one eligible trial.

3. Results 

The PubMed/Medline search revealed 1183 hits. The eligible articles were 133 from which 9 were not from English literature. The search from the congresses revealed three eligible abstracts. Altogether, 41 peer reviewed articles [3], [4], [7], [8], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46] and one abstract were analyzed [47]. Twelve studies contained information from MUP confined exclusively to lymph nodes [19], [25], [27], [31], [32], [37], [42], [44], [46], [47]; parotid gland included [9], [19]. Three reports included only patients with metastatic visceral disease [8], [12], [37] while two studies included only teenagers and young adults [33], [43]. A distinct group included articles referring to MUP of the breast [41] and the lung [39]. Three studies were from Australia [15], [37], [39], one form Germany [34], one from Israel [22], one from Ireland [46], one from Poland [47] 3 from The Netherlands [26], [31], [38], one from USA and Canada [9] and 31 from the USA. The study period extended from 1917 to 2006, i.e. 89 years. Das Gupta criteria for the characterisation of MUP were used in seven studies [3], [13], [16], [18], [21], [24], [34], while in the rest of the studies various combinations of clinical and imaging techniques were used (Table 1). Indeed, history of a removed lesion without histological examination, thorough physical examination (including cutaneous, oral, ocular, otolaryngologic proctoscopic, urogenital evaluation), endoscopy studies (bronchoscopy, laryngoscopy), imaging techniques (sinus films, chest X-ray, CTs and PET/CT) and cytology were variously combined in for the assessment of MUP.

Table 1. Studies’ characteristics and Patient demographics.

	Author	Year	Country	Study period	Setting	MUP criteria	MUP/MM pts	% of MM	Males/females	Median age (range)
	Pack [4]	1952	USA	1917–1950	Any MM	NA	29/1190	2.4
	Das Gupta [3]	1963	USA	1935–1957	Any MUP	Das Gupta”	37/992	3.7	23/14	Peak 40-49
	McSwain [11]	1964	USA	1925–1962	Any MM	NA	8/203	4
	Einhorn [12]	1974	USA	1967–1973	Disseminated MM	NA	65/426	15
	Baab [13]	1975	USA	1944–1969	Any MUP	“Das Gupta”	98 /2466	4	71/27	Peak 5th decade
	Conley [14]	1976	USA	1935–1972	Head and neck MM	NA	21/660	3.2
	Milton [15]	1977	Australia	1950–1977	Any MUP	Physical Examination	76/1833	4.1	52/24
	Giuliano [16]	1980	USA	1971–1978	Any MUP	“Das Gupta”	55/980	5.6	38/17	43 (20–70)
	Lopez [17]	1982	USA	1955–1976	Any MUP	NA	129/1045	8.1	87/42	46.4 (9–86)
	Chang [18]	1982	USA	1949–1975	Any MUP	“Das Gupta”	166/3805	4.4	108/58	48–47
	Spiro [19]	1983	USA	1965–1976	Cervical LNs of occult origin	Physical Examination Endoscopies; Imaging sputum cytology	11/132	8	8/3	57 (41–64)
	Reintgen [20]	1983	USA	1972–1982	Any MUP	Physical Examination Imaging	124/2612	4.8	97/37	47.9 (16–85)
	Panagopoulos [21]	1983	USA	1940–1975	Any MUP	“Das Gupta”	30/670	4.4	22/8	45 (22–79)
	Klausner [22]	1983	Israel	1973–1981	Any MUP	Physical Examination. History	12/530	5.2	6/6	10 pts over 50 years
	Santini [23]	1985	USA	1944–1980	MUP of parotid, cervical LN	Physical Examination. History	96		78/18	48 (15–81)
	Muchmore [24]	1986	USA	1958–1983	MUP of extremities	“Das Gupta”	29/822	3.5	16/13	4th decade peak
							29/296 stage III	9.8
	Wong [25]	1987	USA	1971–1986	MUP of LNs	Physical examination. History	188/4011	4.6	119/69	43 (18–73)
	Jonk [26]	1990	The Netherlands	1961–1986	MUP of LNs	NA	26		15/11	45 (21–81)
	Coit [27]	1991	USA	1974–1984	MM of axillary/inguinal LNs	NA	51/449	11.3
	Velez [28]	1991	USA	1977–1988	Any MUP	NA	64/1045	6	39/25	44.5 (2–73)
	Norman [29]	1992	USA	1986–1989	Any MUP	Physical examination. Imaging	18/580	3.1	10/8	38.4 (17–71)
	Sirott [30]	1993	USA	1984–1991	Any MM	NA	36/284	13
	Balm [31]	1994	Netherlands	1976–1992	MUP of neck/parotid LNs	Physical examination. History	17/300	5.7	12/5	42 (25–81)
	Nasri [32]	1994	USA	1964–1991	MUP of neck/parotid LNs	Physical examination. history; imaging	46	na	39/7	43 (15–85)
	Pappo [33]	1995	USA	1967–1988	MUP in children and adolescents	NA	3/33	9	2/1	12
	Sutherland [7]	1996	USA	1981–1987	Any MUP	NA	58/5004	1.2	36/22
							87/6900	1.3	58/29
							145/11,904	1.2	94/51
	Schlagenhauff [34]	1997	Germany	1976–1996	Any MUP	Das Gupta	75/3258	2.3	35/40	53
	Anbari [35]	1997	USA	1978–1996	Any MUP	History	40 (na)	na	25/15	49.5 (29–74)
	Chang [36]	1998	USA	1985–1994	Any MM	NA	1893/84836	2.2	1245/646	56
	Vijuk [37]	1998	Australia	1983–1996	MUP with visceral metastases	NA	146	4	106/40	55 (17–84)
	Strobbe [38]	1999	Netherlands	1961–1995	MM of obturator and iliac LNs	NA	7//71	9.8	na
	Wang [9]	1999	USA/Canada	1975–1997	MM of the parotid	NA	6 /19	31.5	6/0	69
	de Wilt [39]	2005	Australia	1947–2000	MUP of lung	Physical examination. History	15		12/3	59 (49–73)
	Katz [40]	2005	USA	1986–1996	Any MUP	NA	65/2485	2.6	41/24	54 (19–82)
	Ravdel [41]	2006	USA	1976–2005	MM of the breast	NA	4/27	14.8
	Cormier [42]	2006	USA	1990–2001	MUP of LNs	Physical examination	71 /804	8.8	51/20	51 (21–71)
	Lange [43]	2007	USA	1985–2003	Any MM of children/teenagers	NA	20/3158	0.7
	Lee [44]	2008	USA	1971–2005	MUP of LNs	Physical examination	262/1571	16.6	187/75	46 (16–86)
	Bedikian [45]	2008	USA	1987–2001	MM stage III/IV	NA	83/616	13.5
	Lee [8]	2009	USA	1971–2005	MUP disseminated	Physical examination. Imaging	398/2247	17.7	274/124	50 (18–99)
	Kelly [46]	2009	Ireland	2002–2007	Any MUP	Physical examination. Imaging; endoscopies	5/478	1.04	4/1	51.2 (39–63)
	Nowecki [47]	2009	Poland	1994–2006	MUP of LNs	NA	47/449	10.5	25/22	50

MM: malignant melanoma; LN: lymph nodes; pts: patients; Any: any site involved; NA:Not Assessed.

The majority of the studies were retrospective reviews of patients’ records treated in the referral center. In the Sutherland study there were data from mailed forms to 681 hospitals in 1981 and to 844 hospitals in 1987. Hence, this study is represented two times in data analysis except for survival where only the primary data were mature enough for the analysis. National Cancer Data Base was used in Chang et al. study [36] and in the report of Melanoma in children and teenagers [43] while the data from phase I/II [30] and phase II/III [45] were used in two other studies.

4. Therapeutic interventions (Table 2) 

Data were retrieved from 32 studies on the therapeutic modalities used for the treatment of MUP. Lymph node dissection either radical or modified was the gold standard of treatment for nodal disease. This was combined with parotidectomy whenever the parotid gland was involved [9], [22], [23], [32]. Further medical treatment is depicted in Table 2, according to the treatment setting (adjuvant or palliative). 474 patients were managed with an immunotherapeutic or chemotherapeutic regimen. Drugs included were: interferon-a, interleukin 2, dacarbazine, procarbazine, cisplatin, cyclophosphamide, etoposide, lomustine, dactinomycin, vincristine, methotrexate, Thio Tepa, CCNU, methyl CCNU, BCNU, phenylalanine mustard and estramustine. 131 patients received both chemotherapy and immunotherapy, 332 received involved field radiotherapy, while 385 patients were treated with a combination of radiation therapy and systemic modality. 454 patients were administered adjuvant treatment while 205 were treated with palliative intent.

Table 2. Therapeutic interventions.

		Surgical management (number of patients)	Further treatment modality
			Adjuvant	Palliative
	Das Gupta [3]	Radical LND		RT and /or chemotherapy
	Einhorn [12]			Chemotherapy
	Baab [13]	Radical regional LND	NA	NA
	Milton [15]	Radical LND	NA	NA
	Giuliano [16]	LND (patients with regional LN involvement only)	Immuno	Combination of resection, chemotherapy, radiation and immunotherapy
	Lopez [17]	Surgery for “cure”(29%)	NS: Immuno, Chemo. RT
	Chang [18]	Radical LND (pts with LN disease only);		Chemo, Immuno, RT
	Spiro [19]	Radical LND	RT
	Reintgen [20]	Surgery for isolated skin nodules and LNs, isolated pulmonary and CNS metastases	NS: Immuno, Chemo or both
	Panagopoulos [21]	Radical LND (1719). Biopsy in pts with lung and subcutaneous disease		Chemo or RT (for lung metastases or subcutaneous nodules)
	Klausner [22]	Radical LN dissection and/or parotidectomy	Chemo+Immuno
	Santini [23]	Neck dissection with or without parotidectomy in all but 14 (excisional biopsy)	Chemo+Immuno
	Mutcmore [24]	Regional perfusion+regional LND (pts with regional LN involvement)	NA
	Wong [25]	Regional LND	Immuno
	Jonk [26]	Radical LND	RT
	Velez [28]	Surgical resection for localized LN disease (32); surgical resection of gross tumour or for palliation (18)	Chemo±RT	Chemo±other combination
	Norman [29]	LN biopsy and LND	NA	NA
	Balm [31]	Surgical removal of neck and/or parotid LNs	RT	DTIC and/or Immuno, vaccine, RT
	Nasri [32]	Parotidectomy and/or LND in all but 6 (excisional biopsy)	Chemo+Immuno; RT
	Pappo [33]	LND		Chemo; Chemo+RT
	Sutherland [7] 1981 data	Surgical resection		Combinations of Chemo, RT, surgery
	Sutherland [28] 1987 data	Surgical resection		Combinations of Chemo, RT, surgery
	Schlagenhauff [34]	Removal cutaneous subcutaneous metastases; radical LND	NA	NA
	Anbari [35]	LND (26); surgical excision of subcutaneous nodules (3); excisional surgery for metastatic disease (9)	Immuno or chemo	Chemo; RT
	Chang [36]	Surgery only	NS: Chemo, RT, combination of RT and medical therapy
	Vijuk [37]	Complete metastatic resection (29); partial resection (26)		Immuno; Chemo; RT
	Strobbe [38]	Superficial and deep LND	NA	NA
	Wang [9]	Parotidectomy (6) combined with neck dissection (1)	NA	NA
	de Wilt [39]	Total lobectomies (10). Partial lobectomies (2), wedge resections (6)	NA	NA
	Cornier [42]	Complete LN dissection	RT; systemic therapya; vaccine
	Lee [44]	Regional LN dissection	Vaccine; BCG; GMCSF
	Kelly [46]	Regional LN dissection	BCG, INFα2b
	Nowecki [47]	Radical LN dissection

LND: lymph node dissection; Immuno: immunotherapy; Chemo: chemotherapy; RT: radiotherapy; NA: not assess; NS: not specified; BCG: bacille Calmette–Guerin; GMCSF: granulocyte macrophage colony-stimulating factor.

a INF-a, interleukin 2, chemotherapy (dacarbazine, cisplatin, vincristine) or combination.

5. Survival/recurrence 

Outcome and relapse data are summarized in Table 3a, Table 3b. For optimal comparison of the survival data two big groups were created; one with nodal involvement and one with visceral disease. Twenty-three studies gave information about survival from MUP with nodal disease and 16 from visceral disease. Considering nodal disease (Table 5a) the median overall survival varied between 24 and 127 months reaching up to 165 months in the matched control population of the John Wayne cancer institute [44]. The 5-year survival rate was between 28.6% and 75.6% while the 10-year survival rate ranged between 18.8% and 62.9%. In the case of visceral disease the median survival ranged between 3 and 13.2 months and the 5-year survival rate between 5.9% and 18%. There were no 10-year survivors.

Table 3a. Survival of patients with lymph node disease and sites of relapse.

	Author/year	Metastatic site	MUP/MKP	OS (months)	5 ySR (%)	p	10 ySR (%)	Site of relapse
	Das Gupta [3]	LNs	24	NA	30a	NA	NA
	Baab [13]	LNs	49	NA	33	NA	22	57% LN, 20% skin, 20% subcut. 3% LN+skin
	Lopez [17]b	Any	129	10	13	<0.001	NA
		LN only	29	70.2
		Regional	15	7.5
		Disseminated	85	6.1
	Chang [18]	LNs	75	NA	46	NR	41	21% regional; 39% distant
			NA		39		21
	Reintgen [20]b	Subcut (I)	19	28.3 (I)	NA	I vs. III 0.10	NA
		LNs (II)	79	31.4 (II)		II vs. III 0.0003
		Disseminated (III)	26	7.3 (III)
	Panagopoulos [21]	LNs	19	NA	29	NA	29
			22 MKP with		29		29
			metachronous LNs
			55 MKP with		10		0
			Synchronous LNs
	Santini [23]	LNs	96	NA	39	NA	19	Local (neck recurrence)
			274		23		14
	Muchmore [24]b	LNs	15	NA	63	NA	52
		Subcut	9		58		42
	Wong [25]	LNs	188	37	46	0,10	41
			387	NA	37		NA
	Jonk [26]	LNs	26	35	49	NA	49
	Coit [27]	LNs	37	35	46	0.0004c	39
			145(trunk)	27	28		23
			43 (arm)	64	54		46
			125 (leg)	48	46		37
	Velez [28]	Localized	34	53	45	NR	NR
	Norman [29]	LNs	18	NA	NA	0.96	NR	50% regional LN; 25% distant LN; 12.5% CNS;
			38d	NA	NA			12.5% distant skin
			14d	NA	NA
	Balm [31]	LNs	17	36	48	NA	32	3 regional 5 distant
	Nasri [32]	LNs/parotid	46	24	56	NA	41	11 regional 10 distant
	Schlagenhauff [34]	LNs	37	NA	50	0.14	NA
			NA		36
	Anbari [35]	LNs	20e	NR	57e	0.008	NA
			46	20	19
	Chang [36]	LNs	160 (LD)	NA	55	NA	NA
			40 (no LD)		27
	Strobbe [38]	LNs	7	NA	76	NA	63
			71		24		20
	Katz [40]b	LNs	28	48.2	39	<0.01	NA
		Other	37	12	14
	Cornier [42]	LN	71	76	55	0.04	44	25% regional; 70% distant
			446	35.5	42		33	5% synchronous
	Lee [44]	LNs	262	127	55	0.0021	50
			1309	42	44		38
	Lee [44]	After matching	221	165	58	0.0006	52
			221	34	40		36
	Nowecki [47]	LNs	47	35	40	NS	NA	15% LNs, 66.5% visceral
			402	24	32			18.5% subcutaneous

NA: not assessed; LN: lymph node; Cut: cutaneous; subcut: subcutaneous; OS: overall survival; ySR: year survival rate; NS: not significant; NR: not reached; LD: lymphadenectomy; CNS: central nervous system; BOLD: pts with MKP.

a 10 (out of 24) patients were treated and achieved this survival. b Comparison of MUP subgroups. c In comparison with MKP of the trunk. d 38 patients were with MKP, no palpable LNs, 14 patients with MKP and palpable LNs. e 20 patients with LN disease without adjuvant chemotherapy; 4 year survival is represented.

Table 3b. Survival data of patients with non-nodal disease.

	Author/year	Metastatic site	MUP/MKP	OS (months)	5 ySR (%)	p
	Pack [4]	Visceral	29	NA	5.9	NA
			1161		21.4
	Das Gupta [3]	Visceral	13	3 (max 12)	0	NA
	Einhorn [12]	Visceral	65	18/7a	NA	NA
			426	31
	Milton [15]	LNs	50	NA	25.5	NA
		Visceral	16		26.1
		Any MKP	1757		70,3
	Chang [18]	Visceral	91		8	NA
	Reintgen [20]	Subcut	19	28.3	NA	0.006
		Subcut	NA	80.9
	Coit [27]	Visceral	10	10	7	0,15b
			38 (trunk)	9	11
			4 (arm)	5	0
			43 (leg)	13	8
	Velez [28]	Disseminated	30	7	10	NA
	Sutherland [7]	Any	58	NR	19.9	NA
	Schlagenhauff [34]	Visceral	8	6	NA	0.14
			NA	5
	Anbari [35]	Visceral	11	13.2		NA
	Chang [36]	Any MUP	518	NA	29.1	NA
		Cutaneous	23696		80.8
		Mucous m	306		25.0
		Ocular	1275		74.6
	Vijuk [37]	Visceral	146	7.7	NA	0.024
			146	5.8
	De Wilt [39]	Resected pulmonary MUP	15	32	42	NA
	Wang [8]	Visceral disease	146	16	18	0.0002
			146	11	10
	Lee [8]	Visceral disease	398	16	18	0.488
			1849	15	16
	Lee [8]	After matching	392	16	18	<0.001
			392	11	10

NA: not assessed; LN: lymph node; OS: overall survival; ySR: year survival rate; BOLD: pts with MKP.

a 18 months for 34 pts with complete removal of metastatic melanoma; 7 months for 31 pts with non-resectable disease. b In comparison with MKP of the trunk.

A comparison of outcome of MUP patients with LN involvement with MKP patients was found in eleven studies, historical controls in two and matched patients in seven [18], [21], [23], [25], [27], [29], [35], [38], [42], [44], [47]. Indeed, although MUP patients had better survival in all studies statistical significance was achieved in only four of them [27], [35], [42], [44]. There were four studies comparing MUP patients of nodal involvement only with subcutaneous or visceral disease [17], [20], [24], [40]. In all cases, the first subgroup fared significantly better. Indeed, two studies showed that MUP patients with LN disease at presentation had significantly better overall survival from patients with distant metastases; subcutaneous and visceral (Reintgen et al.: 31.4 months vs. 11.3 months; p=0.02 and Katz et al. p=0.01).

Considering relapse data, the majority of recurrence appeared in the first 2 years after diagnosis while the relapse rate varied mainly between 42% and 62% [13], [18], [20], [29], [31], [32], [42]. In the Baab study 80% of those who recurred developed local relapse, 10% in the contralateral site and 10% in both sites [13]. The average time of recurrence was 7.4 months for men and 8.3 months for women. In the Chang et al. study [18], 65% of males and 25% of females relapsed, the majority within 1 year (71%) while distant metastases were twice as frequent as regional ones (39% vs. 21% respectively). In the Giuliano et al. study fewer early relapses occurred in MUP patients compared to MKP, though at 24 months these rates were identical [16]. The patients with MUP who did not receive adjuvant immunotherapy had disease recurrence more often than those who received immunotherapy, though the differences were modest and patients were not randomized. The majority of studies describe high rate of relapse (usually over 45%), in contrast to the Sutherland study which reported a recurrence rate of only 5.2%. Nevertheless in this study there was a substantial high rate of residual disease and unknown data (72% and 12% respectively). Finally in the distinct cohort of MUP cases of the lung [39] there was a wide range of recurrence sites (skin, lymph nodes, lung, brain, bowel, bone, adrenal glands and liver).

6. Prognostic factors 

The various prognostic factors of significant importance are tabulated in Table 4. Tumour regression is a well recognised factor with the involvement of immune system to describe this correlation. A low number of involved lymph nodes was a favourable factor [15], [19], [26], [42], [44], [45], [47] but could not be confirmed in all studies [25], [32]. Females seem to have better survival in various studies [26], [42], [44], [45] but the opposite result could also be noticed [28], [40]. There was no consensus on the impact of age on prognosis. Hence, although Santini, Nasri, Cornier and Lee agreed that younger patients had better survival [26], [32], [42], [44] they reported so with different age cut offs while in other studies age was not a prognosticator [31], [34]. In general, surgical treatment offered survival advantage as compared with those MUP patients who received no treatment [36]. Nevertheless, in case of neck dissection the appropriate approach, modified versus radical, could not be determined [23], [32]. Interestingly enough those who received adjuvant chemotherapy or radiotherapy seemed to have worse survival, though not of statistical significance [23], [26], [32]. Moreover, in the Lee study the use of adjuvant therapy after lymphadenectomy did not affect the outcome in both MUP and MKP patients [44]. The fact that survival differences remained in favour of MUP irrespective of treatment may suggest an intrinsic biological difference between these two entities.

Table 4. Significant prognostic factors.

		Prognostic parameters	p [HR (CI)]
	Milton [15]	1 LN≫multiple LNs	<0.05
	Lopez [17]	Pts with lymphadenectomy≫no curative resection	<0.001
	Chang [18]	Early≫late surgery	<0.001
	Spiro [19]	N1≫N2, 3	<0.0001
	Santini [23]	Modified≫radical ND (5 yrs SR)	0.05
		Age<30≫>50 (2 yrs SR)	0.025
	Muchmore [24]	Spontaneous regression (5 yrs SR)	Significant better
	Jonk [26]	Females≫males (5 yrs SR)	0.0095
		Groin≫axillary≫cervical (5 yrs SR)	0.008
		Spontaneous regression≫no regression	0.039
		1 LN≫multiple LN (5 yrs SR)	0.048
	Nasri [32]	Age <30 yrs≫>60 yrs (5, 10 yrs SR)	<0.05
	Anbari [35]	MUP≫MKP	0.008
	Chang [36]	Regional≫metastatic disease	<0.0001
		Surgery≫no treatment	<0.0001
	Vijuk [37]	MUP≫MKP (without/with treatment)	0.01/0.026
		Complete surgical resection	<0.0001
		Partial surgical resection	0.005
		Metastases to bone and other visceral organs was adverse prognostic factor	>0.0001–0.077
	Katz [40]	LNs≫others	<0.01
		Men≫women	0.09
	Cornier [42]	MUP vs. MKP OS	0.06 [0.61 (0.42–0.86)]
		Male vs. females OS	0.001 [1.48 (1.17–1.89)]
		<50 yrs vs. <50 yrs OS	0.020 [0.76 (0.61–0.96)]
		N3 vs. N1 OS	<0.001 [2.21 (1.67–2.91)]
	Lee [44] (multivariate analysis)	Females vs. males	0.0004 [1.335 (1.137–1.568)]
		<60 vs. >60 yrs	0.0017 [1.294 (1.102–1.520)]
		1 vs. 2–3 vs. 3 LNs	0.0001 [1.256 (1.141–1.384]
		Decade of diagnosis 1970–79 vs. 80–89 vs. 90–99 vs. 2000–2005	0.0131 [0.989 (0.981–0.998]
		MUP vs. MKP
			0.0001 [1.507 (1.220–1.862)]
	Lee [8] (multivariate analysis)	MUP vs. MKP	0.032 [1.141 (1.012–1.287)]
		Age (<60 vs. >60)	0.007 [1.148 (1.038–1.268)]
		Sex (female vs. male)	0.0012 [1.170 (1.064–1.278)]
		Site of metastasis (M1a vs. M1b vs. M1c)	<0.0001 [1.336 (1.252–1.425)]
		Number of metastatic sites (1 vs. 2 vs. >3)	<0.0001 [1.303 (1.235–1.275)]
		Decade of diagnosis (1970–’79 vs. ‘80–’89 vs. ‘90–’99 vs. 2000-2005)	<0.0001 [0.713 (0.676–1.287)]
	Nowecki [47]	Number of metastatic LNs	0.004
		Exrtracapsular extension of metastases	0.05

≫: better than; M: metastasis; MM: malignant melanoma; LN: lymph node; RND: radical neck dissection; SR: survival rate; yrs: years; adj: adjuvant; DFS: disease free survival; OS: overall survival.

Non-surgical management of clinically palpable stage III MUP or MKP was associated with worse survival than surgical management possibly because cytoreduction decreased not only tumour load but also the level of immunosuppression. Other favourable prognostic factors were prompt surgery, 3 months regarded as the cut off [18]. Early treated patients had a 57% 10 years rate compared with 18% respectively in late treated patients. In the Chang study although patients with no residual disease had better survival rate this was not stastistically significant [18]. Vijuk and Coates did not find the number of metastatic sites of independent prognostic significance for survival although the presence of metastases was an unfavourable factor [37].

7. Discussion 

The aim of our study was to bring together the information of almost 90 years of clinical practice and to analyse all the potential information given. The first one to portray criteria for the definition of MUP was Das Gupta in 1963. The fact that in only seven studies these criteria were used and 13 used a different combination of clinical and imaging definition criteria suggests that the MUP population studied is heterogeneous. With this drawback in mind, an attractive information that came out was the incidence of MUP. Despite the variations seen in studies the incidence of 3.2% may represent the true rate of MUP. The high number of 4348 MUP patients analysed can assure us that this is the realistic rate and the inter study differences are diluted. It is obvious that in our study the different methodological approaches available at different time periods could have resulted in different efficacy discovering a “hidden” primary tumour. The invention of CT scan in 1972 and its wide availability by the beginning of 80s has changed dramatically the field of imaging in medicine. In an effort to eliminate the impact of diagnostic diversity on incidence, we divided the studies to those performed in pre-CT era (before 1980) and to those after that. Although it seems that MUP incidence may decrease in time we have to keep in mind that this result is based mainly on the data coming from National Cancer Data Base [36]. When we performed the analysis without this study the incidence in CT era was increased to 4.4% and the overall incidence to 4.7%. Our intention to find the incidence in the PET/CT era could not be accomplished since studies describing the exact number of patients that underwent this exam are missing.

In respect to gender characteristics, twice as many males were shown to develop MUP. The exact reason for this could not be assessed but may be related to the pathogenesis of MUP. Men may be more likely to ignore a primary melanoma until after it regresses and then presents as a metastasis; alternatively, as the distribution of known melanomas differs between men and women, males may develop melanomas in places in which regression is more likely. The peak age of 4th and 5th decade is comparable with cutaneous MKP [36], [44] but it is lower than mucosal and ocular MKP [36]. The prevalence of younger age of MUP may be attributed to stronger immune response.

MUP metastatic to regional lymph nodes is more common in men than in women. It is also more likely to involve the axillary rather than the inguinal or cervical basin. The different distribution of melanoma primary in men and women may be an explanation. It is noteworthy that in men the metastases were located predominantly in the axillae, the head and neck region and the trunk, whereas in women metastases were found more frequently in the extremities. This is similar to the gender-related patterns of primary melanomas. On the other hand the presence of involved inguinal nodes in females is related to the higher incidence of anogenital melanoma in women.

There is a long lasting debate regarding outcome differences between patients with MUP and patients with MKP. We divided the relative studies in two main groups; those with localized nodal disease and those with non-nodal disease. Although our initial aim was to perform regression analysis the heterogeneity of the studies and the lack of adequate number of matched controls precluded us from doing so. In the nodal group there are seven studies showing that MUP patients have better survival than MKP patents [23], [27], [35], [38], [42], [44], [47]; four of them significantly so [27], [35], [42], [44]. On the other hand, there are at least five studies in which survival difference could not be demonstrated [21], [25], [29], [34], [47]. Panagopoulos et al. [21] noted that MUP patients had equal survival with MKP patients with metachronous LN metastases while both groups fared better than MKP patients with LN metastases from the beginning. In the Schlagenhauff study [34] although there were no differences between MUP and MKP patients with nodal or visceral involvement, MUP patients with cutaneous metastases fared better than MKP patients with in transit metastases (83% vs. 50% 5-year survival rate; p=0.02). In spite of these, the information given from the larger study for nodal metastasis from unknown primary melanoma [44], where also matched control was performed, should be regarded as the baseline for further comparisons between MUP and MKP. The wide range in overall survival (24–127 months) is attributed mainly to the Lee et al. study [44] where in the matched cohort survival reached up to 165 months. The Lee trial data may be explained by the larger population, the better design and the longer follow-up of the study. It seems that the different time period the studies were performed, reflecting different diagnostic approach to MUP, the different study designs, the considerable heterogeneity and the small number of patients in certain cohorts may be further issues to consider for this large discrepancy in survival.

Published data depict a more complicated picture concerning visceral disease. Only in the Vijuk and Lee study could it be demonstrated that MUP patients had better survival [8], [37]. However considering that these were actually matched control studies, the impact of this difference is of significance. In the rest of the studies no significant differences could be obtained. Of notice is the Reintgen study where no differences between MUP and MKP in both patient groups with lymph node and visceral metastases could be noted [20]. However, MUP patients with subcutaneous disease fared much worse than the corresponding MKP patients (p=0.006). We should comment on the seemingly adverse effect of adjuvant chemotherapy and radiotherapy on survival [23], [26], [32]. Although not statistically significant this observation is not evidence-based, comes from retrospective studies of different setting and should not mislead us.

Various theories have been developed concerning the appearance of the MUP phenomenon [3], [35], [42]. The first one proposed by Smith and Stehlinin in 1965 support the disappearance of a primary lesion as a result of spontaneous regression after metastasis has occurred [48]. Partial or complete spontaneous resolution of melanoma cells from the primary site is fairly common and it was actually described in various studies [16], [17], [20], [21], [24], [26], [35]. A second assumption was that the origin of the melanoma is in lymph nodes or other subcutaneous tissues or in viscera. Ectopic benign nevus cells have sometimes been found in lymph nodes and other tissues [49], [50]. Third, a synchronous unrecognised melanoma and fourth possibility, a previously excised or traumatically removed melanoma have been proposed as explanations. The most likely explanation for MUP is immune induced regression of the primary tumour [51]. Melanoma accounts for 11% of all instances of spontaneous tumour regression. The incidence of spontaneous regression of metastases from malignant melanoma is approximately one per 400 patients, and possible mechanisms include immunologic, endocrine, inflammatory and tumour nutritional factors. Partial regression has been reported in 9–46% of primary melanomas. Complete spontaneous regression of metastatic melanoma is very rare, with an estimated incidence of 0.22–0.27%. Mauer et al. identified a circulating factor that potentiated lymphocytic cytotoxicity in regressing melanoma [52]; since then, increased lymphocytic infiltrates have been demonstrated in regressed melanomas [53]. Cellular immune recognition and response to melanoma-associated antigens have been thought to induce spontaneous regression mediated by cytotoxic lymphocytes [5], [53], [54]. Furthermore, a favourable prognosis has been associated with the presence of tumour-infiltrating lymphocytes in melanoma [6], [55], [56]. Humoral mechanisms for melanoma tumour destruction also have been proposed. Antibody attachment to melanoma cell membranes has been demonstrated by direct and indirect immunofluorescence [57]. Cytotoxic antibodies against cultured melanoma cells have been found in serum of patients with melanoma [58], [59], [60]. A study comparing the immunologic aspects of MUP and MKP discovered a higher prevalence of antimelanoma antibodies in patients with MUP but no difference in lymphoproliferative activity [61]. In keeping with this explanation, it has been suggested that the immunologic response that results in primary tumour regression may contribute to the more favourable outcomes seen in patients with MUP [62], [63].

There is increasing evidence that reduced immunosuppression might allow enhanced antitumour immune functions, including humoral and cellular cytotoxic responses that eliminate microscopic tumour load. Specific immune responses have been correlated with survival after resection of melanoma without adjuvant immunotherapy. Further characterisation of these immune responses should allow the identification of biomarkers to monitor a patient's response to adjuvant therapy. It is obvious that many prognostic factors have been proposed but few of them can be used definitely. However the existing evidence substantiates the potential benefit of surgical resection of patients with isolated lymph node or visceral disease from unknown primary melanomas. Hence, the relatively favourable long-term survival of patients with MUP supports the belief that, in the context of regional lymph node disease, MUP constitutes a manifestation of stage III disease rather than stage IV (M1a) distant lymph node disease. Therefore, patients who have metastatic melanoma in a regional node in the absence of a known primary site should undergo completion lymph node dissection. These patients also should be considered for adjuvant treatment trials that were designed for patients with stage III disease.

Metastatic melanoma should be considered in the differential diagnosis of all patients who present with a malignancy of unknown origin, particularly when lymph nodes or cutaneous metastases [64] are the primary presenting site. Das Gupta criteria were described 45 years ago and they were mainly clinical ones. But the end of the first decade of the 21st century new criteria should be evaluated with the introduction of imaging techniques as a mainstay for the diagnosis. Thorough clinical evaluation with total skin and subungual examination, detailed otorhinolaryngologic and eye evaluation with thorough oral and nasal mucosa assessment; proctoscopy, scrotal and gynecologic examinations for patients with inguinal lymph node metastases should be considered as appropriate. CT imaging of head and neck, brain imaging (CT or MRI preferably) and CT imaging of the chest/abdomen and pelvis should be escorted to rule out metastatic disease. Although PET/CT has a potential utility in the staging of melanoma [65], [66], its role has not been demarcated for the identification of the primary site of MUP and therefore should not be recommended in first intent. However, in cases of inconclusive CT/MRI when surgical resection is an option PET/CT should be applied [67]. A practical approach for patients with resectable lymph node involvement or potentially resectable visceral metastases is proposed in Table 5.

Table 5. Practical approach.

	Recommended examination	Comments
	Thorough clinical examination	Total skin examination; subungual otorhinolaryngologic, ophthalmologic, scrotal, gynecologic examination; proctoscopy
	IHC (S100, Vimentin, HMB 45)	Histological confirmation
	Electron microscopy (melanosomes or premelanosomes)	Equivocal cases
	CT	Thorax, abdomen, pelvis
	MRI	Brain
	PET/CT	Inconclusive cases of potentially resectable disease (mainly nodal involvement or oligometastatic visceral)

IHC: immunohistochemistry.

8. Conclusions 

Conflict of interest 

None declared.

Reviewers 

Maurice Matter, MD, Centre Hospitalier Universitaire Vaudois (CHUV), Department of Visceral Surgery, Rue du Bugnon 21, CH-1011 Lausanne, Switzerland.

Piotr Rutakowski, MD, PhD, M. Sklodowska-Curie Memorial Cancer Centre and Institute of Oncology, Dept of Soft Tissue/Bone Sarcoma and Melanoma, Warsaw, Poland.