Evaluation of MRI of the Pineal Gland in Retinoblastoma

1. INTRODUCTION AND RATIONALE Currently baseline brain MRI with extended follow-up of pineal cysts is systematically performed in all new retinoblastoma patients since children with hereditary retinoblastoma have an increased risk of primitive neuroectodermal tumors (PNET) that are histopathologically identical to the retinal tumors. This combination is also known as trilateral retinoblastoma (TRb). The prevalence of developing a PNET in combination with unilateral or bilateral hereditary retinoblastoma is 5-15%. The risk of developing TRb in Rb patients, is less than 0.5% for sporadic unilateral disease. The PNET usually arises in the pineal gland (PG) (77%) but can also be a parasellar or suprasellar tumor. A meta-analysis from 1999 including TRb patients from literature covering 1966 to 1998; they found that the interval between the ocular tumor and the intracranial tumor was 1 versus 22 months respectively in TRbs diagnosed in patients with screening versus patients who developed symptoms. Of TRb cases found at screening half were present when Rb was diagnosed and that 75% might be detected within one year after Rb diagnosis. Treatment is difficult and the prognosis is poor as only few survivors are reported. In a previous retrospective multicenter study from the ERIC-group small asymptomatic PNETs (<15 mm), that were present synchronously with the eye cancer (on baseline MRI) showed to have a better prognosis. Furthermore it was discovered that the majority of early stage pineoblastoma showed a cystic aspect, sometimes hardly distinguishable from benign pineal cysts (9,11,12). The prevalence of pineal cysts in hereditary Rb patients is 5.3%. In the general population at an age of 1 to 5 years 1.6% (1.1% in males and 2.2% in females) of all patients have pineal cysts. Current practice is based on evidence that early depiction of TRbs in retinoblastoma patients (during baseline MRI and follow-up of selected patients with pineal cysts) detects these aggressive tumors in a curable stage.
2. OBJECTIVES The main objective of this prospective multicenter study is to evaluate the current clinical strategy of baseline MRI screening of the brain in newly diagnosed retinoblastoma patients, with extended follow-up of selected patients with simple and complicated pineal cysts.

To investigate this objective the investigators will try to answer the following question: "What is the diagnostic accuracy (sensitivity and specificity) of the baseline MRI screening and extended follow-up of patients with pineal cysts for the diagnosis of TRb?" and: "How can the investigators optimize this screening strategy in order to minimise patient burden by keeping number of MRIs to a minimum and to maximise diagnostic accuracy?" The investigators will evaluate the classification (see section 3.7) of pineal cysts (probably benign versus suspicious) and pineoblastoma or supra- / parasellar PNET.

2.1 Secondary objectives

1. Evaluate epidemiological parameters:
2. incidence of TRb patients at baseline MRI (synchronous), follow-up (metachronous) MRI, and missed by MRI, compared to historical data;
3. survival of patients with positive TRb findings on MRI at baseline or follow-up compared to TRb patients missed by MRI and compared to historical data.
4. Evaluate prognostic parameters (age, tumor size, aspect, time of diagnosis relative to Rb diagnosis etc).
5. STUDY DESIGN The investigators propose a prospective multicenter cohort study (to investigate the diagnostic accuracy of baseline MRI screening of the pineal gland in new patients with retinoblastoma, with extended follow-up of selected patients with pineal cysts for early detection of pineoblastoma.

3.1 Study population Within ERIC about 150 new retinoblastoma patients are diagnosed every year. About 10 percent of all new retinoblastoma patients will be diagnosed at the VUmc.

3.2 Inclusion criteria

• All new hereditary and non-hereditary retinoblastoma patients undergoing a baseline MRI diagnosed at one of the ERIC centers.
• Availability of MR sequences (see section 3.7.1) required by the study

3.3 Exclusion criteria

A patient will be excluded from the study if:

• baseline MRI has not been performed;
• MR protocol or quality not adjusted to the required protocol
• Follow-up not possible or available

3.4 Study quality In 2003 The STARD (standards for the reporting of diagnostic accuracy studies) statement has been published by Bussuyt et al. as a tool to help researchers to improve the accuracy and completeness of reporting of studies of diagnostic accuracy. This should allow readers to assess internal validity (potential bias) and external validity (generalisability) of a study. The investigators will use STARD to ensure that - where possible - the design of this study meets those standards and that the final publication is clear on possible risk of bias and generalisability.

3.5 Sample size calculation In the VUmc and the other ERIC centers 150 new retinoblastoma patients are diagnosed each year. All new Rb patients receive a baseline MRI for (clinical) diagnostic purposes. When a pineal cyst is detected on baseline MRI a follow-up MRI will be done (see section 3.7 for the follow-up protocol). The investigators consider the baseline MRI and the follow-up MRI as one diagnostic test (the index test) and the investigators will calculate sample sizes based on that. As mentioned in section 1. 75% of TRb cases can be diagnosed during follow-up within one year, of which 50% can be diagnosed synchronously with Rb. The other 25% of cases develop at a later stage; the investigators will not consider those as false negatives of our test. There are no previous studies on diagnostic accuracy on comparable tests, but on the basis of clinical experience the investigators estimate a sensitivity of at least 95% and a specificity of about 80%. Prevalence in the hereditary Rb group (40% of patients) is about 10, giving us a prevalence of four per cent.

Sensitivity depends on the number of cases; therefore the investigators calculate the needed number of cases according to the estimated sensitivity of 95%. The sample size the investigators calculate has a probability of at least 85% (1 - β) that the 95% (1 - α) lower confidence limit is ≥ 60%. These input variables result in a required number of cases of ten. Entering the number of cases in the previously mentioned formula gives us 240 controls. The investigators can then calculate the 95% lower confidence limit of the expected specificity of 80%. With a probability of at least 85% this number of controls gives us a 95% lower confidence limit of ≥ 72.

The number of required Rb patients is 240 controls + 10 cases = 250, and since 75% of TRb cases can be diagnosed in the first year the investigators need 334 Rb patients for this study. About 10% of Rb patients included in this study will be from the VUmc. The investigators expect almost all Rb patients to participate in the study. With a study duration of about two to three years the investigators expect to be able to include the required number of Rb patients. This sample size calculation is based on rough estimates and could therefore vary. Because of the low number of cases it is especially important to obtain a sufficient number of TRb cases for this study and depending on the number of found cases the duration of this study may vary.

The prevalence of pineal cysts in hereditary Rb patients is about 5% and the investigators estimate the prevalence of pineal cysts in the non-hereditary group is similar to the general population, which is about 1.6%. These numbers amount to (5.3% • 40% + 1.6% • 60%) • 334 = 10 patients with a pineal cyst who will receive follow-up MRIs.

3.6 Study procedures

The ERIC guideline for imaging retinoblastoma states that brain screening for pineoblastoma should be performed in every new patient with retinoblastoma (uni- and bilateral). Pineal lesions depicted on baseline MRI are classified as:

1. normal pineal gland; no further follow-up;
2. probably benign pineal cyst; this group contains patients with a cystic pineal gland with a discrete rim enhancement and a thin smooth wall. Pineal glands with thin intracystic septa will also be classified as probably benign pineal cysts. Follow-up MR imaging will be done once after 3 months and further follow-up is not necessary if the cysts remains stable*;
3. suspicious pineal cyst; irregularly thickened (> 2 mm) cyst wall or fine nodular aspect of the wall. MRI follow-up also after 3 months. If stable, no further follow-up. If any doubt persists, another follow-up after 3 months*;
4. pineoblastoma or supra- / parasellar PNET.

• In case a patient receives chemotherapy a scan 3 months after chemotherapy is done, to exclude a possible chemotherapy effect on the PNET growth.

3.6.1 MRI protocol This screening is primarily achieved by a post-contrast 3D T1-weighted sequence with 1 mm slice thickness or post Gadolinium thin slice (≤3mm) T1-weighted. The preferred sequence for initial evaluation and follow-up are thin slice (≤ 3mm) T2-weighted (T2-TSE or CISS-images). Follow-up should be performed with the same T2-weighted and post-contrast 3D-T1 weighted sequence.

3.7 Follow-up of subjects withdrawn from treatment Survival data and disease progression will be followed up in all included Rb patients.

1. STATISTICAL ANALYSIS Sensitivity and specificity will be the main measures of outcome of this study. The low number of expected TRb cases will probably not allow for subset analyses. The diagnostic index test the investigators are evaluating (the baseline MRI and the follow-up MRI) lacks a uniform reference standard, however, it is possible to construct a composite reference standard. For positive index test results it might be difficult to determine false positives in some cases, because in case of a positive index test treatment will have been started, which could make it hard to distinguish effect of treatment from a false positive index test result, when a reference test is carried out at a later stage than the MRI. A good reference standard is especially critical for patients where MRI results are not clear cut. See table 1 for the expected test results based on the data the investigators used in our sample size calculation (see section 3.5).

Reference standard for patients with a positive index test (any combination of):

• tumor cells in cerebrospinal fluid;
• histopathological confirmation (gold standard);
• clinical disease progression during follow-up;
• follow-up MRI diagnostics;
• absence of alpha-fetoprotein (AFP), beta human chorionic gonadotropin (β-hCG) in cerebrospinal fluid or blood.

Reference standard for patients with a negative index test:

• clinical diagnosis of TRb within one year of the last MRI.

The investigators have set a limit of one year for the reference standard of a negative index test to try to avoid patients who develop TRb after the last MRI. The investigators realize that this could falsely classify patients with fast growing tumors that develop after the last MRI, who present with clinical symptoms within a year as true positives. The investigators will evaluate the effect of 'this one year limit' on sensitivity and specificity by varying this limit.

4.1 Secondary objectives After a follow-up period of 5 years survival data will be analyzed. Survival curves will be presented as Kaplan-Meier plots. The log-rank test will be used to compare our date with data from historical patient series. Results will be corrected for potential confounders like differences in treatment, and lead time bias. Stratifications will be made according to tumor size, pineal TRb versus ectropic intracranial TRb, and time between Rb diagnosis and TRb diagnosis.