Open questions — retinoblastoma¶
Last curated: 2026-09-01 (independent audit, same date). Tier assignments are curator judgment. Every asserted evidence gap was searched live on 2026-09-01 — at build and again at audit — and remains provisional, not proof of permanent absence.
Tier 1 — would change practice; designable now¶
OQ-1. Which delay component causes the income-stratified mortality gradient?¶
The global cohort quantifies survival and stage gradients but not the causal share of recognition, referral, diagnosis, treatment start and completion (Global Retinoblastoma Study Group 2022, PMID 35839812). A 1,120-child Indian cohort measured those intervals—44%, 26% and 31% of total lag—but did not connect intervention on each component to survival (Das 2025, PMID 40719713). A prospective multi-country mediation study is feasible. → global-access-and-outcome-disparity.md
OQ-2. Which support package prevents treatment abandonment?¶
Stage III/IV abandonment reached 38.5%/46.6% in one resource-limited extraocular cohort (Pant 2017, PMID 29337595), and the natural history of stopping is documented: median 13.7 months from intraocular diagnosis to orbital disease, 2.6 months to metastasis and 2.0 months to death, with 100% mortality by 48 months (Zhao 2021, PMID 34359552). Programme-level funding across 55 projects raised estimated five-year paediatric-cancer survival by a median of 5.1% but was not designed to isolate which component worked (Howard 2018, PMID 29726390). Navigation, transport, cash support and decentralized follow-up can be factorially or cluster-randomized with treatment completion and survival endpoints. → global-access-and-outcome-disparity.md
OQ-3. Does IAC improve useful vision, not only globe salvage?¶
Randomized IAC improved two-year progression-free globe salvage from 27% to 53% but reported no standardized binocular-function outcome (Wen 2023, PMID 37536351). The instruments exist: contrast sensitivity and saccade testing detect deficits acuity misses (Reynolds 2023, PMID 37442536), strabismus persists in 69% of salvaged patients at mean 93.7 months and is predicted by foveal involvement (Fabian 2018, PMID 30009948), and vision-related quality of life is measurable and worse in bilateral survivors (Belson 2023, PMID 36520266). Acuity, contrast, fields, amblyopia treatment and participation should be prespecified in route-comparison follow-up. → intra-arterial-chemotherapy.md, vision-outcomes-and-patient-experience.md
OQ-4. What intravitreal dose controls seeds with the least retinal toxicity?¶
Series report high seed control but ocular toxicity in 59% of eyes at standard 20–30 µg dosing, with focal seeds responding better than diffuse (Shields 2016, PMID 26630319; Yousef 2021, PMID 34322023). The two available drug comparisons disagree by endpoint — melphalan gave better seed resolution (92% vs 72%, P = 0.069) but more posterior-segment complications and lower globe salvage (60% vs 77%) than topotecan in 64 eyes with unequal follow-up (Agarwal 2025, PMID 39566884) — and carboplatin's phase I trial closed early after a retinal toxicity event (King 2023, PMID 36372348). Prospective dose/exposure–retinal-function modeling is designable. → intravitreal-and-intracameral-chemotherapy.md
OQ-5. Which isolated high-risk histopathology features require adjuvant chemotherapy?¶
Observational evidence suggests a large benefit for isolated massive choroidal invasion, a global cohort shows a very large effect for postlaminar optic-nerve invasion (metastasis subhazard ratio 5.39, 95% CI 1.75–16.60 without adjuvant chemotherapy), and a randomized trial supports three rather than six cycles across a defined high-risk set (Feng 2023, PMID 37603354; Lin 2026, PMID 41475682; Ye 2024, PMID 39432296). The obstacle is definitional before it is statistical: only three features command unanimous agreement among specialists (Kaliki 2022, PMID 34762098) and central review disagrees with local reporting in 17–23% of cases (Chévez-Barrios 2019, PMID 31539297). Central pathology review and feature-stratified randomization could reduce overtreatment. → enucleation-and-high-risk-pathology.md
OQ-6. Should serial brain MRI be restricted by baseline pineal morphology?¶
Meta-analysis predicts missed asynchronous pineoblastomas and models 311 scans per asymptomatic tumour detected and 776 per life saved under six-monthly imaging to 36 months, whereas prospective follow-up required 494 scans per pineal tumour when all cystic glands were followed and 22 when restricted to suspicious glands (de Jong 2020, PMID 32061409; de Jong 2022, PMID 33939299; de Bloeme 2024, PMID 38992673). A harmonized prospective protocol can compare detection yield, stage, anaesthesia and false positives. → extraocular-metastatic-and-trilateral-disease.md
OQ-7. What surveillance improves mortality in adult heritable survivors?¶
Subsequent-neoplasm risk is large, but consensus found no evidence that routine imaging of asymptomatic survivors improves outcomes (Schonfeld 2021, PMID 33473166; Tonorezos 2020, PMID 32422154). International pragmatic evaluation of skin, symptom-triggered and targeted protocols is feasible. This question does not apply wholesale to non-heritable survivors. → second-cancers-and-survivorship.md
OQ-8. Has subsequent-cancer burden fallen in the modern local-therapy era?¶
Radiation dose–sarcoma effects are established in historical heritable cohorts, while contemporary IAC/intravitreal cohorts are not old enough for lifetime outcomes (Wong 1997, PMID 9333268; Kleinerman 2019, PMID 31622129). Linked registries with exposure dosimetry can begin now even though mature answers require time. → second-cancers-and-survivorship.md
OQ-9. What testing workflow most reliably excludes heritable disease after unilateral retinoblastoma?¶
Constitutional and mosaic variants occur in isolated unilateral disease, and modern NGS can detect variant classes missed by limited methods (Lohmann 1997, PMID 9311732; Li 2016, PMID 27155049). Paired plasma and buffy-coat sequencing found RB1 mosaicism in 14.7% of 136 consecutive patients, four of whom had tested germline-negative at outside laboratories (Gao 2025, PMID 40338593), while germline detection rates run 86.2% in bilateral versus 19% in unilateral disease (Gupta 2021, PMID 34294096). Paired tumour–blood benchmarks across laboratories are designable. → genetic-testing-and-counselling.md
OQ-10. Can awareness programs reduce cT4 presentation?¶
Red-reflex/leukocoria education is biologically plausible and widely promoted, and newborn red-reflex examination for retinoblastoma is described as widely accepted despite little evidence of validity and effectiveness (Jullien 2021, PMID 34496780). Lag is reliably associated with stage (Das 2025, PMID 40719713; Sherief 2022, PMID 36938376), but a live PubMed search on 2026-09-01 combining retinoblastoma with awareness, education, screening or early detection and restricted to randomized or controlled-trial publication types returned five records, all of them treatment or diagnostic-imaging trials or a general cancer-awareness review — none an awareness intervention with a stage endpoint in retinoblastoma. Expert opinion in low- and middle-income settings currently favours strengthening referral pathways and abandonment reduction over universal screening (Chantada 2025, PMID 40619694). Cluster trials should use cTNMH stage and referral time, not awareness scores alone. → clinical-presentation-and-diagnosis.md
OQ-11. Which outcomes should define successful conservative treatment?¶
Current studies separate survival, globe salvage and occasional acuity; small survivor studies identify contrast and saccadic deficits beyond acuity (Reynolds 2023, PMID 37442536). A core outcome set is designable with survivor and parent participation. → vision-outcomes-and-patient-experience.md
OQ-12. Can resource-adapted IAC programs reproduce vascular safety?¶
The randomized trial reported 18% stenosis and 3% occlusion; high-volume retrospective series show learning effects but infrastructure differs (Wen 2023, PMID 37536351; Shields 2014, PMID 24656794). A prospective implementation registry should report technical failure, vascular harm, survival and opportunity cost. → intra-arterial-chemotherapy.md
Tier 2 — important, blocked by rarity, tools or Tier-1 answers¶
OQ-13. Is MYCN-amplified, RB1-wild-type retinoblastoma a distinct therapeutic disease?¶
Two metastatic cases showed rapid chemoresistant disease, but the denominator is too small to define treatment (Zugbi 2020, PMID 32971811). International molecular ascertainment must precede a subtype trial. → molecular-pathogenesis-and-cell-of-origin.md
OQ-14. Which single-cell state predicts metastasis prospectively?¶
Single-cell and multi-omic studies identify cycling cone precursors and a dedifferentiated high-risk subtype, but clinical validation is absent (Yang 2021, PMID 34815392; Liu 2021, PMID 34552068). → molecular-pathogenesis-and-cell-of-origin.md
OQ-15. Can aqueous-humour cfDNA guide treatment safely?¶
Aqueous cell-free DNA reproduces the tumour's RB1 genotype (3/3 concordant in the first matched series, and generally high concordance across 14 matched-sample studies in a 2026 systematic review), and 6p gain carries roughly 10-fold odds of enucleation in two successive cohorts (Raval 2022, PMID 35577019; Darajati 2026, PMID 41521223; Berry 2018, PMID 30061186; Xu 2020, PMID 32434859). Prospective companion-diagnostic use has so far been reported in 26 eyes of 21 patients (Berry 2024, PMID 38040321), and mosaic RB1 variants persist in cell-free DNA after treatment in the absence of disease, creating a false-positive route to overtreatment (Gao 2025, PMID 40338593). What is missing is a prospective study in which cfDNA-informed management is compared with clinical management on eye and child outcomes; routine tissue biopsy remains unsafe. → molecular-pathogenesis-and-cell-of-origin.md
OQ-16. How should cT2b be subdivided?¶
cT2b is heterogeneous, and diffuse versus focal seeds differ in salvage (42% vs 62%, P < 0.0001) (Yousef 2021, PMID 33769386). The same problem recurs one group higher: reclassifying group D to E moved 17% of eyes and produced 12-month ocular survival of 79%, 59%, 49% and 1% across D, E1, E2 and E3 (Singh 2024, PMID 38830602), while all three intraocular schemes predict high-risk pathology about equally (Kurian 2025, PMID 39922380). Revision needs international reproducibility and treatment-independent validation. → classification-and-staging.md
OQ-17. What therapy materially improves CNS metastatic disease?¶
ARET0321 achieved one-year EFS 28.3% in IVb/trilateral disease despite intensive treatment and caused two therapy-related deaths overall (Dunkel 2022, PMID 35820112). Molecularly targeted trials are blocked by rarity and need international pooling. → extraocular-metastatic-and-trilateral-disease.md
OQ-18. What is the long-term functional effect of repeated anaesthesia?¶
Retinoblastoma care can require repeated examinations under anaesthesia and arterial procedures, and surveillance protocols add to that count — following all cystic pineal glands cost 494 scans per tumour detected (de Bloeme 2024, PMID 38992673). NCT03546387 (RECRUITING, ClinicalTrials.gov v2 API, verified 2026-09-01) is studying cognitive function after multiple anaesthesia exposures, but separating anaesthesia from disease, treatment and visual impairment as causes remains difficult. → vision-outcomes-and-patient-experience.md
OQ-19. Which prosthetic-eye pathway optimizes orbital growth and psychosocial outcome?¶
Perioperative reviews describe age-specific implant/prosthesis needs, but comparative long-term functional and psychosocial evidence is sparse (Leclerc 2020, PMID 31886534). → enucleation-and-high-risk-pathology.md
OQ-20. How should reproductive counselling be delivered where testing is unavailable?¶
Heritable status alters offspring risk and surveillance, yet access is quantified only at country level: genetic testing was available in 74 of 145 countries (51.0%) in 2024, against intravenous chemotherapy in 95.9% (Global Retinoblastoma Study Group 2026, PMID 42463304), and no study links testing availability to the survival gradient (Li 2016, PMID 27155049; Global Study Group 2022, PMID 35839812). → genetic-testing-and-counselling.md
Dots not yet connected¶
| # | Dot A | Dot B | The missing junction | Powers |
|---|---|---|---|---|
| D1 | Essential treatment available almost everywhere (PMID 35839812) | survival 99.5% to 57.3% by income (PMID 35839812) | separate availability from timely delivered/completed treatment | OQ-1, OQ-2 |
| D2 | Parent/system lag fractions measured (PMID 40719713) | global mortality hazard measured (PMID 35839812) | causal mediation of each delay interval | OQ-1 |
| D3 | IAC improves globe salvage (PMID 37536351) | contrast/saccade deficits persist (PMID 37442536) | randomized route comparison with functional vision | OQ-3, OQ-11 |
| D4 | Intravitreal seed control high (PMID 26630319) | ocular toxicity common in another cohort (PMID 34322023) | cumulative dose–retinal-function curve | OQ-4 |
| D5 | Shorter adjuvant therapy is noninferior (PMID 39432296) | pathology definitions vary (PMID 34762098) | treatment de-escalation by centrally reviewed feature | OQ-5 |
| D6 | Baseline MRI misses asynchronous tumours (PMID 33939299) | broad cyst follow-up has low yield (PMID 38992673) | risk-adapted serial imaging trial | OQ-6 |
| D7 | Heritable SMN incidence is high (PMID 33473166) | routine surveillance benefit is unproven (PMID 32422154) | mortality-effectiveness study | OQ-7 |
| D8 | Radiation dose predicts sarcoma (PMID 9333268) | modern routes displace radiation (PMID 37536351) | exposure-linked modern survivor cohort | OQ-8 |
| D9 | Mosaic variants exist in unilateral disease (PMID 9311732) | NGS can detect low-level variants (PMID 27155049) | inter-laboratory sensitivity benchmark | OQ-9 |
| D10 | Red-reflex screening is widely accepted despite little validity evidence (PMID 34496780) | lag correlates with stage (PMID 40719713) | controlled stage endpoint for awareness | OQ-10 |
| D11 | MYCN/RB1-wild-type cases appear aggressive (PMID 32971811) | subtype-directed agents remain preclinical: abemaciclib in xenografts (PMID 42212883), oncolytic adenovirus at phase 1 (PMID 30674657) | international molecular basket trial | OQ-13, OQ-17 |
| D12 | Parent and survivor QOL reports differ (PMID 36385462) | 143 PROMs are in use but only one is retinoblastoma-specific (PMID 32770435) | patient-engaged core outcome set | OQ-11 |
| D13 | Heritable disease raises melanoma risk (PMIDs: 33473166, 33090227) | adult consensus found no evidence that routine imaging of asymptomatic survivors helps (PMID 32422154) | coordinated cross-condition surveillance without evidence duplication | OQ-7 |
| D14 | Non-heritable overall SIR is 0.8 (PMID 33473166) and 1.52 in a second national cohort (PMID 33090227) | many survivorship resources address all survivors undifferentiated | risk-stratified communication that avoids over-surveillance | OQ-7 |