Long-term relative survival in uveal melanoma: a systematic review and meta-analysis.

Background: A large proportion of patients with uveal melanoma develop metastases and succumb to their disease. Reports on the size of this proportion vary considerably.
Methods: PubMed, Web of Science and Embase were searched for articles published after 1980. Studies with ≥100 patients reporting ≥five-year relative survival rates were included. Studies solely reporting Kaplan-Meier estimates and cumulative incidences were not considered, due to risk for competing risk bias and classification errors. A meta-analysis was performed using random-effects and weighted averages models, as well as a combined estimate based on curve fitting.
Results: Nine studies and a total of 18 495 patients are included. Overall, the risk of selective reporting bias is low. Relative survival rates vary across the population of studies (I2 48 to 97% and Q p < 0.00001 to 0.15), likely due to differences in baseline characteristics and the large number of patients included (τ2 < 0.02). The 30-year relative survival rates follow a cubic curve that is well fitted to data from the random-effects inverse-variance and weighted average models (R 2 = 0.95, p = 7.19E-7). The estimated five, ten, 15, 20, 25 and 30-year relative survival rates are 79, 66, 60, 60, 62 and 67%, respectively. Conclusions: The findings suggest that about two in five of all patients with uveal melanoma ultimately succumb to their disease. This indicates a slightly better prognosis than what is often assumed, and that patients surviving 20 years or longer may have a survival advantage to individuals of the same sex and age from the general population.

Introduction

Melanomas of the uvea are the most common primary intraocular malignant tumors in adults, affecting more than 7000 individuals each year worldwide[1]. No substantial survival differences have been observed between commonly used treatment modalities, patient sex and age or calendar period during the last several decades[2-6]. Eventually, a large proportion of patients develop distant metastases after which median survival is about one year[7]. Currently available treatment options for primary tumors have limited effect on patient survival[3]. Similarly, there are no clinically available treatments with meaningful impact on survival in metastatic disease[8].

Estimations of the proportion of patients that develop metastases vary considerably. It is often stated that one half of patients will die from their disease[9-14]. Other publications report significantly lower mortality rates in the range of 20–25% for patients with similar patient baseline characteristics over similar periods of time[2,15,16]. The reasons for this variance may at least partially be found in dissimilarities in the methods used for calculation of the mortality rates. No method of estimation of the long-term mortality in disease is without flaws, but some methods of estimation of mortality rates may be less suitable than others. Actuarial methods including life tables and Kaplan–Meier estimates are excellent for evaluation of all-cause mortality, but are likely to overestimate disease-specific mortality in the presence of competing risks, i.e., death from other causes[17-19]. Cumulative incidences of melanoma-related mortality rely on accurate classifications of the cause of death. This has relatively small impact when studying diseases with low mortality, but may be more biased in studies of a disease with a mortality that approaches 50%[20]. A previous study found that uveal melanoma-related death was misclassified in more than half of cases included in a national cause of death registry[21]. Relative survival, in which the observed overall survival of a cancer population is divided by the overall survival in a reference population without the cancer is less prone to bias provided that the sample size is sufficiently large, that the disease is rare in the general population and that it does not have risk factors that are strongly associated with other causes of death (e.g., smoking)[20]. Analysis of relative survival may therefore be well suited for uveal melanomas.

What do we answer patients that ask us the basic question about how high the mortality is in their disease? Patients with choroidal or ciliary body melanomas rarely undergo biopsy or other tumor sampling prior to treatment and before we have results from radiological and detailed ophthalmological examinations, an individualized prognosis is not available. This study is intended to help us inform patients and relatives with a systematic review and meta-analysis of long-term relative survival rates. Based on nine included studies and a total of 18,495 patients, we estimate 5, 10, 15, 20, 25, and 30-year relative survival rates of 79, 66, 60, 60, 62, and 67%, respectively. This indicates a slightly better prognosis than what is often assumed, and that patients surviving 20 years or longer may have a survival advantage to individuals of the same sex and age from the general population.

Methods

Search strategy and selection criteria

We did a meta-analysis to evaluate the long-term relative survival in uveal melanoma. Data was acquired with a comprehensive literature search in the PubMed, Web of Science and Embase databases for peer reviewed published articles that described relevant results. The following search terms were used and matched to appropriate medical subject headings: (“uveal melanoma” OR “choroidal melanoma” OR “ciliary body melanoma”) AND “relative survival”. The search strategy was restricted to titles and/or abstracts of human clinical studies published after January 1st 1980 in English or any language for which an English translation was readily available[22]. The latest search was performed on August 19, 2021. All available studies were included and could be accessed in full via the University Library, Karolinska Institutet. Study authors were contacted if discrepancies existed, for clarifications or if we thought that additional unpublished data could be useful for this analysis. Trial registries, unpublished studies, gray literature, animal studies, laboratory studies, letters to the editor, correspondence, notes, editorials, and conference abstracts were not considered. Reference lists of included articles were searched for additional studies. As both clinical trials and observational studies could be considered, the search method was based on the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) and on the checklist for Meta-analyses Of Observational Studies in Epidemiology (MOOSE)[21,23]. The PRISMA and MOOSE checklists are available as related manuscript files. The protocol was registered and published in advance on PROSPERO (CRD42021265504).

The selection of articles for this analysis was performed in four steps: identification, abstract and full-text screening, eligibility assessment, and inclusion. Abstract screening of articles identified in the literature search was done independently by the two authors, with any disagreements resolved by discussion. Publications were included for full-text screening if they reported (1) ≥5-year relative survival rates (or data that could be readily converted to relative survival) for patients with uveal melanoma, (2) consecutively or prospectively included patients, (3) at least 100 patients. Studies were excluded if they (1) only reported survival rates for prognostically relevant subgroups (e.g., tumors with specific mutations, gene expression profiles, histological appearance, or size categories), (2) were earlier versions of a series of articles from the same database or center, (3) reported patients that were already included in another publication, or (4) did not provide confidence intervals or standard errors for their relative survival estimates. Studies including patients with primary conjunctival or orbital melanomas or metastatic lesions were not considered. The same inclusion and exclusion criteria applied to full-text screening (if not evident in title or abstract). Additionally, articles could be excluded if they were deemed to have sub-par methodological quality, as described below.

Quality assessment of studies

All articles that reached the eligibility assessment step was evaluated with a modified version of the Newcastle-Ottawa Scale-Education (NOS-E)[24]. At the eligibility assessment step, no article was excluded because it was deemed to have sub-par methodological quality according to NOS-E.

Data collection, qualification of searchers and risk for bias assessment

Relative survival rates were extracted from downloaded full texts of each included study. The data was not coded. Dr. Herrspiegel is an ophthalmologist and ocular oncology and pathology researcher. Dr. Stålhammar is a board-certified ophthalmologist and pathologist, and his qualifications include a research group leadership of Ocular Oncology and Pathology at Karolinska Institutet, Stockholm, Sweden. Risk of bias was evaluated according to the recommendations of the Cochrane collaboration[22].

Statistical analysis

The meta-analysis was based on three pre-specified methods for calculating relative survival rates. The a priori determined outcome measure was the long-term relative survival rate, reported in 5-year intervals. The variance of survival rates across the population of studies was evaluated with τ2, which reflects the amount of true heterogeneity regardless of number of included studies or sample size[25]. The alternative measurements of heterogeneity Q and I2 were included for comparison[25,26]. Differences with a p < 0.05 were considered significant, all p values being two-sided. When derived from cumulative overall survival, the 95% CI of the relative survival rates was calculated by dividing the standard error of the observed cumulative survival rate by the expected survival rate, which is a common method in cancer epidemiology (Eq. 1):[27-29]

Equation (1): 95% confidence interval (CI) of a relative survival rate.

Firstly, the random-effects inverse-variance statistical method was used, with 95% CI. The weight given to each study was the inverse of the variance of the relative survival rate. To obtain the standard error from stated 95% CIs, the latter was divided by 3.92 (Eq. 2).

Equation (2): Standard error calculated from 95% CI.

Thus, larger studies which have smaller standard errors were given more weight than smaller studies, which have larger standard errors. This weighting method minimizes the imprecision of the pooled effect estimate[22]. This statistical analysis was performed using the Review Manager by the Cochrane Collaboration (RevMan Version 5.4. Copenhagen: The Nordic Cochrane Center; The Cochrane Collaboration, 2014).

Secondly, weighted averages were calculated based on a model previously used for meta-analysis of shorter-term mortality in uveal melanoma[30]. The number of deaths at each point in each study, n†, was multiplied with the same study’s sample size, n. The resulting product was divided by the total sum of weights Σn†n to arrive at a weight for each individual study. Σn†n was then multiplied by n†/n, resulting in a weighted average for each study, p, which was then summed for a pooled weighted average (Eq. 3). 95% CI was calculated from the standard deviation of .

Equation (3): Pooled weighted average (). n†, number of deaths at each point in each included study. n, study sample size. Σn†n, total sum of weights in all included studies.

Thirdly and last, curve fitting was performed based on the results of the random-effects and weighted averages models to arrive at a combined estimate. 95% CI for the combined estimate was calculated by dividing the standard deviation of the random-effects and weighted averages models by the square root of the number of samples (n = 2) to obtain a standard error from which the 95% CI was derived (±1.96 × standard error). This statistical analysis was performed using SPSS statistics version 26 (IBM, Armonk, NY, USA).

Role of the funding source

The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. CH and GS had full access to all the data in the study and GS had final responsibility for the decision to submit for publication.

Table 1

Summary of primary studies included in the meta-analysis. Y, year(s).

Relative survival
#	Authors (ref.)	n	Country	Period of diagnosis	Tumor size	Group	5 years (%)	95% CI	10 years (%)	95% CI	15 years (%)	95% CI	20 years (%)	95% CI	25 years (%)	95% CI	30 years (%)	95% CI
1	Kujala et al.[17]	289	Finland	1962–1981	Mean T 7 mm, mean ⌀ 13 mm		63	60–66	57	54–61	54	50–58	50	45–55	62	54–69	82	71–92
2	Isager et al.[34]	2178	Denmark	1943–1997	N/a	Men	66	62–69	55	51–60	49	44–54
						Women	69	66–73	57	52–61	54	48–59
3	Burr et al.[35]	2876	England and Wales	1986–1999	N/a		72	70–75
4	Virgili et al.[36]	1081	Western and Eastern Europe*	1983–1994	N/a	Western Europe	63	58–69
						Eastern Europe	65	61–70
5	Caminal et al.[39]	155	Spain	1995–2004	Small 13%, Medium 58%, Large 34%**		90	84–96
6	Chew et al.[42]	308	Australia	1981–2005	⌀ ≤10 mm 22%, >10 mm 26%, unknown 52%. T, ≤5 mm 19, >5 mm 26, unknown 55%		81	76–87	71	63–78
7	Baily et al.[16]	253	Ireland	2010–2015	Mean T 6.3 mm, mean ⌀ 12.6 mm		77	68–84
8	Stålhammar et al.[23]	677	Sweden	1980–1999	Mean T 5.2 mm, mean ⌀ 10.1 mm		79	76–82	71	66–75	66	60–72	64	55–73	74	60–87	86	58–115
9	Radivoyevitch et al.[49]	10,678	United States	1975–2016	N/a		81	80–82	70	69–72	64	63–66	60	57–63	60	56–64	61	56–67
	Sum	18,495

T thickness, ⌀ diameter, N/a not available, y years

*Only patients from non-overlapping geographical regions included.

**Tumors classified as small if their apical thickness were between 1 and 3 mm and their longest basal diameter at least 5 mm, as medium if their apical thickness were between 3.1 and 10 mm and their longest basal diameter no more than 16 mm, and as large if their apical thickness were greater than 10.1 mm and/or their basal diameter greater than 16 mm.

Table 2

Pooled estimates of relative survival in five-year intervals after uveal melanoma diagnosis in random effects and weighted averages models.

	5 years		10 years		15 years		20 years		25 years		30 years
	RS (%)	95% CI	RS (%)	95% CI	RS (%)	95% CI	RS (%)	95% CI	RS (%)	95% CI	RS (%)	95% CI
Random effects estimate	75	70–80	65	58–72	59	52–66	58	50–65	63	57–69	74	56–92
Weighted average	79	76–82	69	63–75	63	55–72	60	49–71	60	49–72	61	50–72

RS relative survival.