Multimodal Imaging Features of Bilateral Diffuse Uveal Melanocytic Proliferation: A Systematic Review of 82 Patients

Abstract

Introduction

Bilateral diffuse uveal melanocytic proliferation (BDUMP) is a rare paraneoplastic syndrome characterized by proliferation of benign melanocytes throughout the uveal tract. Our purpose was to quantify the prevalence of multimodal imaging findings in BDUMP and identify features that most reliably characterize this rare paraneoplastic syndrome.

Methods

We conducted a systematic review following PRISMA 2020 guidelines, searching PubMed/MEDLINE, Web of Science, and Embase through September 2025. We included case reports and series providing patient-level data on BDUMP with associated imaging. Prevalence rates were calculated using a documented denominator approach with 95% Wilson score confidence intervals.

Results

Sixty-seven studies encompassing 82 patients were included. Median age was 67 years (range 35–88); 54.9% were female. Gynecologic (25.9%), gastrointestinal (22.2%), and lung (17.3%) cancers were most common. Optical coherence tomography revealed retinal pigment epithelium irregularities in 91.8% and subretinal fluid in 90.2% of patients, exceeding the prevalence of classical orange-red patches on fundus photography (82.6%). Fundus autofluorescence showed alternating hypo/hyperautofluorescence in 76.6%. Choroidal thickening was present in 88.9% on ultrasonography. Bacillary layer detachment was identified in 16.4% of cases.

Conclusion

OCT-based structural abnormalities are more consistently present than Gass’s classical funduscopic signs. These quantitative data may assist earlier recognition of BDUMP and suggest that updated diagnostic criteria should incorporate multimodal imaging findings.

Introduction

Bilateral diffuse uveal melanocytic proliferation (BDUMP) is a rare paraneoplastic syndrome characterized by proliferation of benign melanocytes throughout the uveal tract, occurring in patients with systemic malignancies [1]. Gass established the diagnostic framework in 1990, describing five cardinal features: multiple orange-red subretinal patches, exudative retinal detachment, rapidly progressive cataracts, multifocal hyperfluorescent lesions, and elevated pigmented uveal lesions [2]. However, diagnosis remains challenging due to phenotypic heterogeneity and overlap with other chorioretinal conditions, including metastatic disease, chronic serous chorioretinopathy, posterior scleritis and medication-related ocular toxicities, particularly from immune checkpoint inhibitors and targeted biologic agents used in oncology [3].

The spectrum of associated malignancies has expanded since initial descriptions, with gynecologic cancers representing 25–30% of cases, followed by gastrointestinal (20–25%) and lung malignancies (15–20%) [4–7]. Rare associations with hematologic malignancies have also been documented [8, 9]. In a subset of cases, the ocular manifestations precede cancer diagnosis, making prompt recognition critical for both visual and systemic outcomes [10]. Early initiation of plasmapheresis and systemic cancer treatment has been reported to stabilize or improve visual function in some cases, although definitive evidence for prognostic benefit remains lacking [11].

Recent advances in multimodal imaging have revealed features beyond clinical examination. Optical coherence tomography (OCT) demonstrates bacillary layer detachment (BALAD), previously unreported before 2020 [12]. Fluorescein angiography may show pinpoint macular hyperfluorescence or mild leakage in some cases. The possibility of an underlying inflammatory contribution has been noted in prior descriptions and more recent reports, though its role in BDUMP pathophysiology remains incompletely understood [2, 13]. However, the relative diagnostic value of different imaging modalities remains undefined, and no systematic synthesis exists of imaging findings across the spectrum of BDUMP cases. We performed a systematic review to quantify the prevalence of multimodal imaging findings in BDUMP and establish which features most reliably identify this condition.

Methods

Search Strategy and Study Selection

We conducted and reported this systematic review in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines. The PRISMA checklist is provided as online supplementary material (for all online suppl. material, see https://doi.org/10.1159/000550536). We performed a comprehensive literature search of PubMed/MEDLINE, Web of Science, and Embase for all articles published from inception through September 2025 using the search terms “bilateral diffuse uveal melanocytic proliferation” and “BDUMP.” We manually searched reference lists of included studies and relevant reviews to identify additional cases. The last search in all databases was done on September 30, 2025.

After removing duplicates, two independent reviewers (A.A. and A.S.S.) screened titles and abstracts for eligibility. We retrieved full texts of potentially relevant articles and assessed them against the inclusion criteria. We included case reports and case series providing original, patient-level data on paraneoplastic BDUMP with associated imaging. We excluded review articles without original data and studies of non-paraneoplastic etiologies. Discrepancies were resolved through consensus discussion.

Data Extraction

We developed a standardized data extraction form to capture: study characteristics (author, year, country), patient demographics (age, sex), primary malignancy, imaging modalities utilized (fundus photography, OCT, fluorescein angiography, fundus autofluorescence [FAF], indocyanine green angiography [ICGA], ultrasound, ultrasound biomicroscopy [UBM], optical coherence tomography angiography [OCTA], anterior segment OCT), and specific imaging findings. Two reviewers independently extracted data, with discrepancies resolved by consensus. For each imaging finding, we recorded whether it was explicitly assessed and reported as present or absent or not mentioned in the original report. This distinction is critical for calculating accurate prevalence estimates in the context of heterogeneous case report literature.

Unit of Analysis

The primary unit of analysis was the patient. We considered a patient positive for an imaging finding if it was present in at least one eye. We chose this approach to account for the bilateral nature of the disease and avoid statistical errors from treating paired eyes as independent observations. We compiled eye-level data for descriptive purposes.

Statistical Methods and Handling of Missing Data

We calculated prevalence rates with 95% Wilson score confidence intervals, which are appropriate for small sample sizes and extreme proportions. We used a “documented denominator” approach: only patients with explicit documentation of a finding as present or absent were included in prevalence calculations. Findings described as “not mentioned” were excluded from the denominator to avoid underestimating prevalence due to incomplete reporting. This conservative approach means denominators vary across findings based on reporting completeness in the source literature. We performed all statistical analyses and created data visualizations, including the forest plot of prevalence rates, using R version 4.3.1 (R Foundation for Statistical Computing, Vienna, Austria).

Results

Study Selection and Cohort Characteristics

The systematic search identified 395 records. Following deduplication and screening, 67 studies met the inclusion criteria, providing data on 82 unique patients [1–63]. The PRISMA flow diagram is shown in Figure 1.

Fig. 1.

PRISMA flow diagram.

The methodological quality of the included case reports and case series was evaluated using the Joanna Briggs Institute Critical Appraisal Checklist for Case Reports. Two reviewers independently assessed completeness of clinical, imaging, and diagnostic reporting. Disagreements were resolved by consensus. As most studies adequately described clinical findings and imaging characteristics, the overall risk of bias was considered low to moderate.

A meta-analysis was not conducted because the available data were descriptive and heterogeneous. Instead, findings were synthesized narratively to highlight the prevalence and spectrum of multimodal imaging features reported in BDUMP. Due to the rarity of the disease and reliance on single-case reports, a formal assessment of publication bias was not feasible. Still, it is acknowledged that atypical or negative cases are less likely to be published, which may influence the observed distribution of findings.

The median patient age was 67 years (range: 35–88), with 54.9% being female. The most prevalent associated malignancies were gynecologic (25.9%), gastrointestinal (22.2%), and lung (17.3%). In 7 (8.6%) patients, the ocular diagnosis of BDUMP preceded the discovery of the primary cancer (Table 1).

Table 1.

Patient demographics and distribution of primary malignancies in 82 patients with bilateral diffuse uveal melanocytic proliferation

Characteristic	Value
Total patients, n	82
Age at presentation, years
Median (range)	67 (35–88)
Sex, n (%)
Female	45 (54.9)
Male	37 (45.1)
Primary malignancy (n = 81)a, n (%)
Gynecologic	21 (25.9)
Gastrointestinal	18 (22.2)
Lung	14 (17.3)
Other systemic cancersb	11 (13.6)
Unknown at presentation	7 (8.6)
Bladder/urothelial	6 (7.4)
Hematologic	4 (4.9)

^aPrimary malignancy data were available for 81 of 82 patients.

^b“Other” includes melanoma, thyroid, and breast cancer.

Multimodal Imaging Profile

A comprehensive summary of findings is presented below, organized by imaging modality. All prevalence data are reported at the patient level (n = 82).

There was noticeable variability among the included studies in terms of imaging techniques, diagnostic approaches, and the level of clinical detail provided. Some reports focused on fundus or OCT findings, while others included additional modalities, such as FA, ICG, or UBM. These methodological and reporting differences likely explain part of the variation observed in imaging feature frequencies across studies. Given the descriptive nature of the data, no formal statistical assessment of heterogeneity was undertaken.

No formal sensitivity analysis was performed, as this review synthesized descriptive data from individual case reports and small series. However, the overall trends observed across studies remained consistent despite variations in reporting detail and imaging modalities.

Eye-Level Descriptive Analysis

Data were available for 162 individual eyes. The prevalence of key findings calculated at the eye level is presented here for descriptive purposes. Retinal pigment epithelium (RPE) irregularities were documented in 92.6% of eyes (113/122) and subretinal fluid in 91.0% (111/122). On fundus photography, orange-red patches were present in 75.4% of eyes (104/138) and nevi-like lesions in 61.6% (85/138). A comparison with patient-level data reveals differences for certain findings. The prevalence of orange-red patches at the eye level (75.4%) was lower than at the patient level (82.6%), and the prevalence of nevi-like lesions was lower (61.6% vs. 66.7%).

Imaging Modality Utilization

Fundus photography was performed in 69 (84.1%) patients, OCT in 61 (74.4%) patients, and fluorescein angiography in 59 (72.0%) patients. FAF was utilized in 47 (57.3%) patients. B-scan ultrasonography was performed in 36 (43.9%) patients, ICGA in 22 (26.8%) patients, UBM in 16 (19.5%) patients, OCTA in 7 (8.5%) patients, and anterior segment OCT in 2 (2.4%) patients (Table 2).

Table 2.

Utilization of imaging modalities in the systematic review cohort (n = 82)

Imaging modality	Patients imaged, n	Utilization rate, %
Fundus photography	69	84.1
Optical coherence tomography	61	74.4
Fluorescein angiography	59	72.0
Fundus autofluorescence	47	57.3
B-scan ultrasonography	36	43.9
Indocyanine green angiography	22	26.8
Ultrasound biomicroscopy	16	19.5
Optical coherence tomography angiography	7	8.5
Anterior segment optical coherence tomography	4	4.9

Multimodal Imaging Findings

A summary of key findings from the most frequently utilized imaging modalities is presented in Table 3 and Figure 2.

Table 3.

Comprehensive prevalence of multimodal imaging findings in bilateral diffuse uveal melanocytic proliferation (patient-level analysis)

Modality (patients, n)	Finding	Prevalence, % (n) [95% CI]
Fundus photography (69)	Orange-red patches	82.6 (57) [71.7, 90.0]
	Nevi-like lesions	66.7 (46) [54.3, 77.1]
	Giraffe-skin pattern	34.8 (24) [24.1, 47.1]
	Leopard spots	24.6 (17) [15.8, 36.4]
	Reticular orange pigment	18.8 (13) [11.1, 30.1]
	Subretinal fibrosis	5.8 (4) [2.3, 13.9]
	Vascular attenuation	2.9 (2) [0.8, 10.0]
Fundus autofluorescence (47)	Leopard/giraffe pattern	70.2 (33) [55.4, 81.9]
	Alternating hypo/hyperautofluorescence	76.6 (36) [62.0, 86.8]
Fluorescein angiography (59)	Early hyperfluorescence	78.0 (46) [65.2, 87.0]
	Window defects	20.3 (12) [11.7, 32.7]
	Blocked fluorescence	15.3 (9) [8.0, 26.9]
	Optic disc hyperfluorescence	16.9 (10) [9.1, 29.0]
Optical coherence tomography (61)	RPE irregularities	91.8 (56) [81.9, 96.5]
	Subretinal fluid	90.2 (55) [79.8, 95.6]
	Ellipsoid zone disruption	31.1 (19) [20.8, 43.8]
	Bacillary layer detachment	16.4 (10) [8.9, 28.1]
	Hyperreflective foci (all layers)	14.8 (9) [7.8, 26.0]
	Subretinal hyperreflective material	14.8 (9) [7.8, 26.0]
B-scan ultrasound (36)	Diffuse choroidal thickening	88.9 (32) [73.8, 95.9]
	Retinal detachment/elevation	66.7 (24) [49.8, 80.1]
	Vitreous opacities	11.1 (4) [4.4, 25.3]
Ultrasound biomicroscopy (16)	Ciliary body thickening	81.3 (13) [54.2, 94.0]
	Iridociliary cysts	68.8 (11) [42.3, 87.2]
	Anterior ciliary body rotation	18.8 (3) [6.4, 42.2]
	Angle closure	18.8 (3) [6.4, 42.2]
Anterior segment optical coherence tomography (4)	Shallow anterior chamber	​
	Angle closure	​
Indocyanine green angiography (22)	Blocked fluorescence from pigmented lesions	13.6
	Areas of choroidal hyperpermeability	9.1
	“String of sausages” pattern	4.5 (1)
Optical coherence tomography angiography (7)	Choriocapillaris flow voids	28.6

Fig. 2.

Forest plot of key imaging finding prevalence in bilateral diffuse uveal melanocytic proliferation.

Discussion

This systematic review analyzed multimodal imaging findings from 82 patients with BDUMP across 67 studies. Our analysis provides quantitative prevalence rates for imaging features and may contribute to understanding this rare paraneoplastic syndrome.

In our review, BDUMP preceded the diagnosis of the underlying malignancy in 8.6% of patients, highlighting its potential role as an early clue of undiagnosed malignancy. Although most individuals had a known primary tumor at presentation, the fact that ocular manifestations preceded cancer detection in several cases aligns with prior literature reporting that a substantial proportion of BDUMP cases initially lack an identifiable malignancy. This highlights the need for detailed systemic evaluation when BDUMP is suspected. Also, the multimodal imaging findings of BDUMP can mimic entities, such as Vogt-Koyanagi-Harada disease or primary vitreoretinal lymphoma; therefore, clinicians should maintain a broad differential and carefully exclude more common inflammatory or masquerade syndromes before establishing the diagnosis of BDUMP.

The high prevalence of RPE irregularities (95.1%) and subretinal fluid (90.2%) on OCT in our cohort suggests these features may be more consistent than the classical orange-red patches (82.6%) originally described by Gass et al. [2]. This apparent superiority of OCT findings likely reflects technological evolution rather than true prevalence differences – OCT can detect subtle structural changes invisible on clinical examination. Notably, OCT was utilized in only 61/82 (74.4%) patients, primarily from studies after 2005, while fundus photography was available across all decades (69/82, 84.1%). Our findings expand upon Klemp et al.’s [3] review, which preceded widespread OCT adoption and thus relied primarily on clinical and angiographic findings.

One of the most important findings from recent literature is the identification of BALAD [12, 13, 43, 64]. This finding was identified in 16.4% of patients in our review. The relatively low prevalence in our cohort likely reflects under recognition rather than true rarity, as earlier OCT studies would not have specifically evaluated for this recently defined finding. Retrospective re-examination of OCT images from pre-2020 cases might reveal additional instances of BALAD that were previously unrecognized or misinterpreted as simple neurosensory detachment. While our cross-sectional analysis cannot establish prognostic significance, the association of BALAD with photoreceptor disruption in other retinopathies suggests it may indicate more severe disease [62]. Prospective studies using current high-resolution OCT technology and specifically evaluating for BALAD could better define its prevalence and clinical significance in BDUMP.

Among the 61 patients who underwent OCT, spectral-domain OCT was the most commonly used technique reported in 40 patients, and swept-source OCT was used in 6 patients. EDI-OCT was specifically reported in 9 patients. Others did not report the type of OCT used. Because deeper-penetration modalities (EDI-OCT and swept-source OCT) are better suited to assess choroidal structures, variability in OCT type may have influenced detection rates of choroidal thickening and BALAD.

Choroidal thickening was observed in 88.9% of patients who underwent B-scan ultrasonography. We acknowledge that measurements were not standardized across studies and may have been influenced by different equipment and techniques. These findings are consistent with enhanced depth imaging OCT observations by Jabbarpoor Bonyadi et al. [38], who demonstrated bilateral subfoveal choroidal thickening in BDUMP, and Sarigul Sezenoz et al. [13], who reported diffuse choroidal thickening in all examined eyes.

Anterior segment involvement was common when specifically evaluated, though only 16 (19.5%) patients underwent UBM. In this subset, ciliary body thickening (81.3%) and iridociliary cysts (68.8%) were frequently observed, along with previously unreported findings of ciliary body detachment and angle closure in 18.8% each. The limited use of UBM across studies prevents accurate estimation of anterior segment involvement prevalence in the broader BDUMP population. However, these findings, combined with the well-established rapid cataract formation in BDUMP, suggest the disease process extends beyond the posterior segment [3]. Whether anterior segment changes contribute directly to visual decline or represent an epiphenomenon of the underlying paraneoplastic process remains unclear. Prospective evaluation with standardized anterior segment imaging could clarify the frequency and clinical significance of these findings.

FAF revealed an alternating hypo/hyperautofluorescent “leopard” or “giraffe” pattern in 76.6% of evaluated cases, with hypoautofluorescent areas representing RPE loss and hyperautofluorescent zones indicating lipofuscin accumulation. The distinct alternating pattern differs from the confluent hypoautofluorescence of geographic atrophy or Stargardt disease, suggesting potential diagnostic utility [65]. The alternating hypo/hyperautofluorescent pattern differs from the more confluent patterns seen in other chorioretinal disorders and may help distinguish BDUMP from conditions with overlapping uveal involvement, such as idiopathic uveal effusion syndrome or Vogt-Koyanagi-Harada disease. Also, this imaging modality shows a characteristic inverse relationship with fluorescein angiography, hypoautofluorescent areas on FAF corresponded to hyperfluorescent window defects on fluorescein angiography, providing complementary information about RPE integrity [66]. However, FAF was performed in only 57.3% of patients (47/82), predominantly in studies published after 2010 when the technology became widely available. This temporal bias and limited utilization prevent definitive conclusions about sensitivity and specificity. Given FAF’s apparent ability to detect early RPE changes invisible on fundoscopy, routine incorporation into BDUMP evaluation protocols warrants consideration.

Our comparison of eye-level versus patient-level data provides a crucial clinical insight into disease asymmetry. While the underlying pathology is systemically driven and fundamentally bilateral, its clinical expression can be markedly asymmetric, particularly in the early stages. The patient-level prevalence of orange-red patches (82.6%) was significantly higher than the eye-level prevalence (75.4%), confirming that a patient may present with significant disease in one eye and very subtle changes in the other. This highlights the risk of misdiagnosing BDUMP as a unilateral condition and underscores the absolute necessity of performing a thorough multimodal evaluation of both eyes in any suspected case.

Clinical Implications

Our findings suggest several practical recommendations for clinicians evaluating suspected BDUMP. First, OCT should be considered a first-line imaging modality given the high prevalence of RPE irregularities and subretinal fluid (both >90%). These structural findings may be detectable even when funduscopic signs are subtle. Second, comprehensive bilateral imaging is essential, as disease presentation may be asymmetric. Third, when available, FAF may provide additional diagnostic value through its characteristic alternating pattern. Fourth, anterior segment evaluation with UBM should be considered in cases with unexplained visual decline or angle abnormalities, as ciliary body involvement is common when assessed. Finally, clinicians should specifically evaluate for BALAD on OCT. Although it was present in only 16.4% of patients and its prognostic value in BDUMP is not yet established, BALAD is known to be associated with significant photoreceptor disruption in other conditions. It may therefore represent a potential indicator of disease severity, warranting careful consideration.

Considering the substantial clinical overlap between BDUMP and other inflammatory or neoplastic entities, a structured diagnostic approach integrating imaging findings, clinical features, and systematic investigation is essential to avoid misdiagnosis. Importantly, because BDUMP may precede the detection of an underlying malignancy, repeated systemic assessment should be considered when initial investigations are inconclusive.

This systematic review has several important limitations inherent to studying ultra-rare diseases. This review was not prospectively registered, as it represents a retrospective synthesis of published case reports and case series focusing on multimodal imaging findings in BDUMP. Because the study did not involve interventional data or individual-level clinical outcomes such as treatment response, disease progression, or survival, prospective registration was not deemed mandatory according to PRISMA 2020 guidance for descriptive reviews. While the absence of registration may limit reproducibility, all methodological steps – including the literature search strategy, study selection, and data extraction – were transparently and comprehensively reported to ensure clarity and adherence to PRISMA standards. However, the evidence base consists entirely of retrospective case reports and case series, which are susceptible to publication bias. Atypical presentations or successfully treated cases may be overrepresented, while cases with poor outcomes may be underreported. Additionally, heterogeneity in data reporting across 67 studies spanning 3 decades resulted in variable completeness of clinical descriptions and imaging documentation.

To address incomplete reporting, we adopted a conservative “documented denominator” approach, calculating prevalence only among patients for whom findings were explicitly reported as present or absent. While this improves accuracy, it results in variable denominators across findings and may still be influenced by reporting bias. Utilization of modern imaging modalities was particularly inconsistent – while foundational techniques like fundus photography and fluorescein angiography were standard, newer modalities like UBM (19.5%), ICGA (26.8%), and OCTA (8.5%) were used in small subsets. Prevalence estimates for findings identified exclusively by these technologies have wide confidence intervals and should be considered preliminary.

Finally, while we prioritized patient-level analysis to address the unit-of-analysis issue, our descriptive eye-level data suggest clinical asymmetry but cannot quantify actual rates of asymmetric onset or progression from retrospective data. The prevalence of certain signs may be underestimated because they were not part of standard examination or reporting at the time of publication. For example, the relatively low documented prevalence of BALAD (16.4%) may reflect recent recognition of this biomarker rather than its actual frequency. This review represents published knowledge but is necessarily constrained by the original authors’ focus and reporting practices. In addition, emerging studies have proposed that pathways such as the HGF/c-Met axis may contribute to the underlying biology of BDUMP, reflecting ongoing efforts to better understand its pathogenesis and identify potential future therapeutic targets. Together, these advances may help refine diagnostic criteria and ultimately improve early recognition and management of this rare paraneoplastic syndrome [47].

This systematic review of 82 patients provides the first quantitative synthesis of BDUMP imaging findings. The triad of RPE irregularities, subretinal fluid, and choroidal thickening emerged as more consistent than classical funduscopic signs, though prospective validation is needed. While our analysis is limited by the retrospective nature and small samples inherent to rare disease research, these prevalence data may assist earlier recognition of a condition that often heralds occult malignancy. Future studies should validate these findings and explore the clinical significance of newly identified features.

Statement of Ethics

Ethical approval was not required for this study, as it is a systematic review based exclusively on previously published case reports and case series. No new patient data were collected, and all information analyzed was already available in the public domain. Patient consent to participate was not required for this study, as all data were obtained from previously published reports in which informed consent had already been secured by the original authors.

Conflict of Interest Statement

The authors have no conflicts of interest to declare.

Funding Sources

Hakan Demirci is supported by Richard N. and Marilyn K. Witham Professorship. The funder had no role in the design, data collection, data analysis, manuscript preparation, or the decision to submit the article for publication.

Author Contributions

A.S.S. and A.A.: conceptualization, data curation, formal analysis, investigation, methodology, validation, writing – original draft, and writing – review and editing; S.J., B.G.B., B.A., and H.D.: conceptualization, investigation, methodology, and writing – review and editing.

Funding Statement

Hakan Demirci is supported by Richard N. and Marilyn K. Witham Professorship. The funder had no role in the design, data collection, data analysis, manuscript preparation, or the decision to submit the article for publication.

Data Availability Statement

The data supporting the findings of this review were derived from published studies identified through systematic database searches. Data collection forms; data extracted from included studies; data used for all analyses; analytic code; any other materials used in the review are not publicly available due to space limitations but can be obtained from the corresponding author upon reasonable request.