Cutaneous melanoma is a clinically heterogeneous disease, with both mild and life-threatening cases. A range of clinical, histologic, and molecular factors that help predict survival have been identified.1 Many patients, however, have tumors of uncertain behavior, hence the importance of investigating other factors that could help determine prognosis.2

The 2 main forms of vitamin D that are important to humans are vitamin D2 (ergocalciferol) and vitamin D3 (cholecalciferol). While they can both be obtained from diet, 90% of vitamin D3 is obtained via skin exposure to UV radiation from the sun.3 Excessive exposure, however, contributes to skin aging and favors the development of skin cancer. It is not surprising thus that patients with melanoma are often found to have low vitamin D levels during follow-up, as they tend to take extreme measures to protect themselves from the sun.4 Vitamin D may be involved in the natural history of cancer because it supports the proper function of the immune system and also has a role in cell division, where it induces differentiation and apoptosis and inhibits cell proliferation and angiogenesis in both normal and cancer cells.5

Vitamin D levels and germline vitamin D receptor polymorphisms could have a pathogenic role in cutaneous melanoma.5–9 Previous studies have suggested that vitamin D might be of prognostic value in melanoma, with low levels indicating potentially shorter survival, independently of well-known poor prognosis factors.10–16 The studies to date, however, have used highly heterogeneous methods, especially in terms of how vitamin D serum levels are measured.

The aim of this study was to evaluate the prognostic value of vitamin D serum levels in a retrospective cohort of patients with cutaneous melanoma, with adjustment for key prognostic markers and use of a specific statistical methodology to select an optimal vitamin D serum cutoff associated with survival.

The patients’ characteristics according to vitamin D level are shown in Table 1. We studied 286 patients, mostly men (57%), with a median age of 57.5 years (IQR, 45.75–70 years). The most common tumor was an SMM (186, 65%) located on the trunk (129, 45.1%) with a Breslow thickness ≤2mm (234, 81.8%), a mitotic rate ≤5/mm2 (243, 86.2%), and no ulceration (85.6%)

Thirteen patients (4.5%) had a vitamin D serum level <9.25ng/mL while 273 (95.5%) had a level above this cutoff.

Tumor location varied significantly according to vitamin D level. Patients in the lower cutoff group were significantly more likely than those with a level ≥9.25ng/mL to have LMM or ALM and significantly less likely to have SSM or NM (P=.005) (Table 1). Ulceration was also more common in these patients (38.5% vs. 13.2%, P=.026) (Table 1).

No significant between-group differences were found for sex, age, location, Breslow thickness, or mitotic rate.

Overall survival

Twenty-four patients (8.4%) died over a median follow-up of 39.4 months. Of these, 5 had a vitamin D level <9.25ng/mL (5/13, 38.5%) and 19 had a level ≥9.25ng/mL (19/273, 7.0%). Overall survival was significantly different between the 2 groups (P=.0024) (Fig. 1).

Figure 1.

Overall survival was significantly worse in patients with a vitamin D serum level <9.25ng/mL versus ≥9.25mg/mL.

(0,17MB).

The univariate Cox regression model confirmed that patients with a vitamin D level <9.25ng/mL were more likely to have worse overall survival than those with a level above the cutoff (hazard ratio [HR]=0.2; 95% CI, 0.1–0.7; P=.005). Worse survival was also observed in patients aged 65 years or older (HR=4.6; 95% CI, 2.0–10.9; P<.001) and patients with a Breslow thickness >2mm (HR=6.1; 95% CI, 2.7–13.6; P<.001), a mitotic rate >5mitoses/mm2 (HR=6.3; 95% CI, 2.8–4.4; P<.001), and ulceration (HR=4.6; 95% CI, 2.0–10.7; P<.001). Location on the upper extremities or trunk was associated with better overall survival ([upper extremity vs. head/neck: HR=0.2; 95% CI, 0.0–0.9; P=.032]; [trunk vs. head/neck: HR=0.3; 95% CI, 0.1–0.8; P=.013]). Better survival was also observed in patients with SSM versus LMM (HR=0.3; 95% CI, 0.1–0.8; P=.020) (Table 2).

Table 2.

Univariate Cox proportional hazards model.

Variable	HR	95% CI	P
Sex
Male	Ref.	Ref.	–
Female	0.7	0.3–1.5	.320
Age
<65 y	Ref.	Ref.	–
≥65 y	4.6	2.0–10.9	<.001
Tumor location
Head/neck	Ref.	Ref.	–
Upper extremity	0.2	0.0–0.9	.032
Trunk	0.3	0.1–0.8	.013
Lower extremity	0.5	0.2–1.6	.229
Acral site	1.1	0.3–4.0	.909
Histologic subtype
LMM	Ref.	Ref.	–
SSM	0.3	0.1–0.8	.020
NM	1.2	0.4–4.0	.765
ALM	0.9	0.2–4.9	.938
Other/not specified	0.4	0.1–3.8	.451
Breslow thickness
≤2mm	Ref.	Ref.	–
>2mm	6.1	2.7–13.6	<.001
Mitotic rate (MD=4)
≤5mitoses/mm2	Ref.	Ref.	–
>5mitoses/mm2	6.3	2.8–14.4	<.001
Ulceration (MD=1)
No	Ref.	Ref.	–
Yes	4.6	2.0–10.7	<.001
Vitamin D
>9.25ng/mL	Ref.	Ref.	–
≥9.25ng/mL	0.2	0.1–0.7	.005

Abbreviations: HR, hazard ratio; MD, missing data.

The variables included in the final multivariate model were age (≥65 years vs. <65 years; HR=4.4; 95% CI, 1.8–10.4; P<.001), mitotic rate (>5mitoses/mm2 vs. ≤5mitoses/mm2; HR=6.3 95% CI, 2.6–15.1; P<.001), ulceration (yes vs. no; HR=3.7; 95% CI, 1.5–9.1; P=.005), and vitamin D level (≥9.25ng/mL vs. <9.25ng/mL; HR=0.3; 95% CI, 0.1–1.0; P=.050). Age, mitotic rate, ulceration, and vitamin D level were independent predictors of overall melanoma survival in the multivariate analysis (Table 3).

Table 3.

Final multivariate Cox proportional hazards model.a

Variable	HR	95% CI	P
Age
<65 y	Ref.	Ref.	–
≥65 y	4.4	1.8–10.4	<.001
Mitotic rate
≤5mitoses/mm2	Ref.	Ref.	–
>5mitoses/mm2	6.3	2.6–15.1	<.001
Ulceration
No	Ref.	Ref.	–
Yes	3.7	1.5–9.1	.005
Vitamin D
>9.25ng/mL	Ref.	Ref.	–
≥9.25ng/mL	0.3	0.1–1.0	.050

Abbreviation: HR, hazard ratio.

a

Five cases excluded because of missing data.

In this study of 286 patients with melanoma, a vitamin D serum level <9.25ng/mL was associated with worse overall survival, independently of Breslow thickness and ulceration(2 key prognostic factors) and other known factors such as mitotic rate and age.1

Our findings are consistent with previous reports, summarized in a recent meta-analysis.13 Studies to date, however, have used highly heterogeneous methods, particularly to measure serum vitamin D levels.

In their first study of the association between vitamin D and melanoma, Newton-Bishop et al.10 found an association between low levels and worse recurrence-free and overall survival, even after adjustment for Breslow thickness. They analyzed vitamin D levels as a continuous variable and a categorical variable based on tertiles calculated using values reported for the study population. In addition, blood sampling was performed 3–6 months after diagnosis.10

In a second study, the same group observed worse melanoma-specific survival in patients with a low vitamin D level at diagnosis (<20nmol/L, which is equivalent to <8.01ng/mL [conversion factor, 2.496]).15 They also observed an improvement in survival for each 10-nmol/L increase, as apart from considering vitamin D as a continuous variable, they also analyzed it by categories (<20, 20–60, 60–85, 85–100, and 100+ nmol/L). Similarly to in the first study, they adjusted for Breslow thickness but not ulceration.

Timerman et al.12 reported worse melanoma-specific mortality in patients with a low vitamin D serum level, particularly in those with metastatic melanoma. They used the standard cutoff of 20ng/mL, which is the established level for adequate bone metabolism. Blood sampling was performed at different times during follow-up. No adjustments were made for Breslow thickness or tumor ulceration.12

In the study by Weinsten et al.,16 season-specific vitamin D levels categorized by quantiles were associated with worse prognosis in melanoma following adjustment for age.

Bade et al.14 observed shorter overall survival in patients with vitamin D serum levels in the lowest quartile compared with those in the top quartile. Only univariate Kaplan–Meier curves, however, were used.

Unlike the above authors, Saiag et al.17 found no differences in overall survival between patients with a low vitamin D level at diagnosis, although they did find that variations over time (both increases and decreases from baseline) had prognostic value. Similarly to the other authors, they used quartile values as cutoff points. They adjusted for both Breslow thickness and ulceration.17

Fang et al.11 used the most similar methodology to ours. They used recursive partitioning to identify the split associated with the lowest P value in the log-rank test. On applying the cutoff identified—16ng/mL—they found that patients with lower levels had significantly worse melanoma-specific, disease-free, and overall survival after adjustment for logarithmic C-reactive protein, age, sex, clinical stage, and blood draw season in the multivariate model.

We identified a cutoff of 9.25ng/mL in our study using an iterative loop and log-rank testing. This methodology, while different to that of Fang et al.,11 enabled us to identify the cutoff showing the most significant survival difference between groups. Another methodological difference was that we adjusted for Breslow thickness and tumor ulceration rather than clinical stage (III/IV vs. I/II) in the multivariate model. We also adjusted for mitotic rate. While this factor is no longer used to determine clinical stage in the latest melanoma staging system of the American Joint Committee on Cancer,’ it is still an important determinant of prognosis.1

Methodologies such as that used in the present study are interesting because most studies to date have used reference values related to bone metabolism to analyze vitamin D levels. These values, however, may not necessarily correlate with prognosis in cutaneous melanoma. The prognostic value of vitamin D in melanoma needs to be studied by analyzing levels as a continuous variable across a wide range of measurements and in larger studies.

While our findings show a direct association between vitamin D serum levels and prognosis in melanoma, they also point to an indirect association. They showed, for example, a link between low vitamin D levels (<9.25ng/mL) and ulceration. Previous studies have reported associations between low levels and greater Breslow thickness,10,11,14,17–19 a higher mitotic rate,18,20 a higher clinical stage,11,17 and ulceration.11,15,17,18,20 Bade et al.14 found that older melanoma patients had lower vitamin D levels, although other studies have reported an inverse relationship.11,12

Neither our findings nor those of previous studies indicate a cause-effect relationship between vitamin D serum level and survival in cutaneous melanoma. Rather, they show a direct and indirect correlation between low levels and prognosis. More prospective studies and clinical trials are needed to investigate the impact of vitamin D supplementation in this setting.

The strengths of this study include the prospective recording of data, adjustment for key prognostic factors in melanoma, and the use of iterative looping to determine the vitamin D cutoff level most closely associated with survival.

Limitations include the small number of patients in certain categories and the considerable difference in the size of the 2 groups (95.5% of patients had a vitamin D level≥9.25ng/mL).

In conclusion, vitamin D serum levels <9.25ng/mL are associated with worse prognosis in cutaneous melanoma. More studies analyzing larger samples and vitamin D levels as a continuous variable are needed to identify a more accurate cutoff associated with survival in this setting and to examine the usefulness of vitamin D measurements in routine clinical practice.

The authors declare that they have no conflicts of interest.