Volume=20 36, Supplement=20 2, Pages S69-S76 (August = 2009)

10 of 10=20 =

ABSTRACT

FULL=20 TEXT

FULL-TEXT PDF (216=20 KB)

Statistical and Trial Design = Considerations=20 in Central Nervous System Prophylaxis Studies

Jennifer=20 A. Smitha 3D"Corresponding

, Michael Glantzb

The prognosis is very poor = for patients=20 with cancer who develop central nervous system (CNS) = involvement.=20 Prophylactic therapy may be a viable strategy for patients = at high=20 risk of CNS metastases. This article explores the rationale, = feasibility, and ethics of prophylactic therapy. We discuss = various=20 study design considerations for CNS prophylaxis trials, with = particular focus on statistical issues, and provide guidance = to=20 clinicians trying to decide how to investigate prophylactic = therapy=20 The pool of patients eligible for inclusion in clinical = trials is=20 limited. To answer pressing clinical questions, innovative = trial=20 designs are needed, along with operational strategies that = include=20 risk factor enrichment, target-based end point selection and = validation, flexible study accrual and monitoring, and = techniques=20 that permit early termination when initial treatment = outcomes are=20 poor. Two proposed trials for patients with glioblastoma = multiforme=20 and non-Hodgkin lymphoma are presented to highlight design=20 considerations.

Article Outline

=E2=80=A2 Abstract

=E2=80=A2 End=20 Points for CNS Disease

=E2=80=A2 Time=20 to Neurologic Progression (clinical)

=E2=80=A2 QoL=20 and Neurocognitive End Points

=E2=80=A2 QoL

=E2=80=A2 Neurocognitive=20 Function

=E2=80=A2 Tumor=20 Response

=E2=80=A2 Survival

=E2=80=A2 Trial=20 Design Considerations

=E2=80=A2 Simon=20 Two-Stage Designs

=E2=80=A2 Proposed=20 GBM Phase III Study

=E2=80=A2 Proposed=20 NHL Prophylaxis Study

=E2=80=A2 Conclusion

=E2=80=A2 References

=E2=80=A2 Copyright

Despite progress in controlling = systemic=20 cancer, the prognosis for individuals affected by meningeal = and=20 other central nervous system (CNS) metastases is poor; = affected=20 patients rarely survive for more than a few months. In = addition, the=20 spread of cancer cells to the CNS, and cancer cell growth in = the=20 CNS, can cause diverse and debilitating neurologic = complications,=20 including paralysis, loss of vision, disturbances of gait = and=20 speech, altered mental status, and impaired cognitive = ability, which=20 severely compromise quality of life (QoL).

A number of factors suggest an = increased risk=20 of CNS relapse in patients with non-Hodgkin lymphoma (NHL)1, 2, 3, 4, 5, 6, 7, 8, 9, 10, = 11, = 12 = and a variety=20 of solid tumors.13, = 14, = 15, = 16, = 17, = 18 = Prophylactic=20 therapy may be a viable treatment alternative for patients = at high=20 risk for CNS metastases.19 = However, the=20 role of intra-cerebrospinal fluid (CSF) prophylaxis remains=20 controversial, and questions persist regarding the = effectiveness of=20 available prophylactic interventions. This is largely due to = the=20 lack of well-powered and properly controlled trials. The = small=20 number of patients available for intra-CSF prophylaxis = studies=20 presents a challenge and has important implications for = trial=20 design. Moreover, selecting patients who will most likely = benefit=20 from prophylactic therapy with the least risk of toxicity = may not be=20 straightforward.

This review highlights issues of = trial design=20 in patients at risk for developing CNS disease, beginning = with end=20 point selection and measures of clinical benefit. Further = topics of=20 discussion include patient selection criteria, issues = surrounding=20 randomization, blinding, stratification, and sample size. = The review=20 concludes with a discussion of several design strategies = that may be=20 used in future intra-CSF prophylaxis trials.

End Points for CNS=20 Disease

The primary end points selected to = assess=20 clinical outcome should balance the competing goals of = patient=20 benefit and scientific rigor sufficient to satisfy the = demands of=20 institutional review boards and regulatory agencies.20

Time to Neurologic = Progression=20 (clinical)

Time to neurologic progression, = which censors=20 patients at the time of death, and progression-free survival = (PFS),=20 which includes both progression and death as events, = represent end=20 points that attempt to combine time and quality of survival = in one=20 measure (Table 1). Traditionally, these outcome = measures=20 do not account for treatment-related neurotoxicities, which = would be=20 necessary to evaluate the net benefit of treatment. = Meticulous=20 prespecification of the clinical definition of neurologic=20 progression is essential. Once clearly defined, the = assessment of=20 neurologic progression must be blinded, since significant = bias is=20 likely when the individual making outcome determinations is = not=20 masked to the treatment assignment. Assessments performed by = a=20 designated outside physician (someone other than the = treating=20 physician) may help overcome this potential problem. Data = review and=20 end point determination by an independent expert committee = may be=20 another alternative.21, = 22

Table 1.

Traditional End Points = for=20 Neurooncology Clinical=20 Trials


	End Point	Confounders and Concerns
	Tumor response	Lack of correlation with=20 QoL Test=20 imprecision
	Survival	Variation in subsequent=20 care Death from non=E2=80=93central = nervous system=20 cancer Ignores quality of=20 survival
	Cause of death	Variation in subsequent=20 care Lack of reliable=20 criteria Poor correlation with other end=20 points
	Surrogate = markers	Restricted=20 availability Lack of correlation with=20 QoL
	Time to tumor = progression=20 (radiographic or cytologic)	Lack of correlation with=20 QoL Test=20 imprecision
	Time to = neurologic progression	Usually = unblinded Test = imprecision Cumbersome

Abbreviation: QoL, = quality of=20 life.

The frequency of assessment of = neurologic=20 progression, and the interval between repeated tests, are = critical=20 as well. Assessment frequency is in large part determined by = the=20 size of the expected effect; if treatment is expected to = delay=20 progression by only a few weeks, the assessment must be more = frequent then if a longer delay is expected. This desire for = frequent testing is balanced against the increased cost and=20 feasibility of testing. Both increased frequency of testing = and=20 increased sample size increase the power of the study. = Generally,=20 the more frequently neurologic progression is assessed, the = fewer=20 the number of patients required to complete the trial. To = avoid the=20 introduction of bias, assessment frequency must be equal = between=20 comparison groups.

QoL and = Neurocognitive End=20 Points

Direct measures of clinical benefit = include=20 QoL and neurocognitive end points. These measurements allow = the=20 clinician to evaluate the trade-off between clinical benefit = and=20 treatment-related toxicity (both neurologic and = non-neurologic) and=20 assess the relative impact of treatment on the patient's = QoL.

QoL

QoL end points include strategies = such as=20 Quality-adjusted Time Without Symptoms of disease or = Toxicity of=20 treatment (Q-TWiST)23 = and the=20 Independent Living Score (ILS).24 = Q-TWiST=20 provides a single measure that integrates quality and = quantity of=20 life. This is a formal mathematical technique that allows = the=20 clinician to assess the amount of time a patient experiences = treatment-related toxicity, the proportion of the total = survival=20 time during which a patient is free of symptoms related to = either=20 treatment or disease progression, and the amount of time = during=20 which symptoms are present because of disease progression.25 = The rationale=20 for the ILS is the understanding that increasing impairment=20 associated with cancer progression decreases independent=20 functioning. Therefore, a simple way to measure QoL may be = to=20 quantify the amount of time a patient remains independent in = his or=20 her activities. While reductionist, and a simplification of = the=20 concept of QoL, this measurement approach makes the = reasonable=20 assumption that one critical global variable (independence) = measured=20 over time will reflect other aspects of a patient's QoL.24

Neurocognitive=20 Function

Brain tumors may profoundly affect = cognitive=20 function (eg, memory, decision-making), as may the = treatment.=20 Neurocognitive performance may thus reflect the net benefit = of=20 treatment over toxicity. Many instruments measure = neurocognitive=20 function (Table 2).26, = 27, = 28, = 29, = 30, = 31, = 32, = 33 = However, given=20 the multifactorial causes of cognitive deficits in patients = with CNS=20 metastases, accurate assessment may present a considerable=20 challenge.26, = 34 No = single,=20 simple screening tool reliably identifies the degree and = course of=20 impairment and also defines the contributions of multiple=20 etiologies. Some tests measure discrete processes such as = expressive=20 language function, while others measure more distributed,=20 nonspecific, functions throughout the brain.34 = Because=20 neurocognitive deficits may limit a patient's ability to = complete=20 the tests, the reason for non-completion also must be = captured,=20 thereby increasing the power of this end point. The utility = of most=20 tests depends on the psychometric properties of the measure, = the=20 nature of the populations and treatment under study, and the = effect=20 on neurocognitive function of those treatments and of the = disease=20 itself.

Table 2.

Instruments Available = to Assess=20 Neurocognitive = Function


Test	Characteristic Measured	Time to Administer (min)
Hopkins Verbal = Learning=20 Test27	Verbal memory	5
Trail Making Test = Part A28	Visual-motor = speed	2
Trail Making Test = Part B28	Executive = function	5
Controlled Oral Word=20 Association31	Verbal fluency	5
Digit Span Test32	Attention and = concentration	10
Folstein MMSE33	Global cognitive = function	10
Grooved Pegboard29=20 (non-dominant and dominant hands)	Motor dexterity and = speed	2-3
Categoric Word=20 Fluency30	Executive=20 function	5

Abbreviation: MMSE, = Mini-Mental=20 State = Examination.

Few studies have prospectively = followed=20 neurocognitive function among patients treated in randomized = clinical trials who are suffering from CNS metastases.35, = 36 = Corn et al=20 used the Folstein Mini-Mental State Examination (MMSE), a = simple and=20 widely used tool to assess dementia.35 = Mehta et al=20 used a sophisticated array of neurocognitive tests, = including the=20 Hopkins Verbal Learning tests, Controlled Oral Word = Association,=20 Grooved Pegboard Test, and TrailMaking tests A and B.36 = Performance on=20 neurocognitive testing also depends, in part, on the = patient's=20 compliance with treatment and testing. Compliance observed = in the=20 Mehta study was high (94%-96%), and almost all patients in = the Corn=20 study completed at least one MMSE. These studies support the = feasibility of formally evaluating neurocognitive outcome = among=20 patients treated for brain metastases. Although many of = these tests=20 are relatively simple to administer, a high degree of = professional=20 expertise is required for their accurate interpretation.26, = 34

Tumor=20 Response

Direct measures of CNS tumor burden = include=20 magnetic resonance imaging (MRI) and CSF cytology. = Conventional=20 cytology is not quantitative37 = and like=20 conventional histology is subject to large inter- and = intra-observer=20 variability. As discussed elsewhere in this supplement, the=20 reliability of CSF cytology is also very dependent on the = volume of=20 fluid submitted for analysis, the speed of sample = processing, the=20 number of samples obtained, and the site (lumbar v=20 ventricular) from which the CSF is sampled. CSF cytology and = MRI=20 analysis may be combined; however, end point complexity = becomes a=20 serious concern with multimodal combinatorial assessments.=20 Quantitative and semi-quantitative techniques are available = for MRI=20 data, but a strong subjective component remains.38 In = addition,=20 the increasingly widespread use of vascular targeted agents = such as=20 bevacizumab have rendered traditional measurement = strategies, which=20 rely on tumor dimensions on gadolinium-enhanced, T1-weighted images, obsolete.39, = 40, = 41

Survival

Survival can be used as a measure of = clinical=20 benefit. However, the utility of this end point depends on = the=20 particular study and the clinical context. For example, = survival may=20 not be the most sensitive end point when there are competing = risks=20 of death from the systemic and CNS cancers, and the = investigational=20 treatment is directed only at the CNS disease component (eg, = treatment of neoplastic meningitis or brain metastases in = the=20 setting of active systemic cancer). In this case, survival = as a=20 measure of treatment effect may be =E2=80=9Cdiluted=E2=80=9D = by non=E2=80=93CNS-related=20 deaths, and CNS-related mortality may be more useful. = Survival or=20 PFS, where both death and progression are considered events, = might=20 be more sensitive end points in trials where the mechanism = for=20 overall disease progression depends on the status of the CNS = component (eg, the glioblastoma multiforme [GBM] study = discussed=20 below) or where the CNS acts as a reservoir to = =E2=80=9Cresupply=E2=80=9D the brain=20 (eg, GBM) or systemic disease (eg, acute lymphocytic = leukemia), or=20 where the treatment itself may hasten death (eg, bone marrow = transplant). In these latter situations, all-cause = mortality, time=20 to (any) disease progression, or PFS may be valid end=20 points.

Trial Design=20 Considerations

An effective trial design calls for = adequate=20 numbers of prespecified primary end point events. An = adequate number=20 of events can be achieved by enrolling a large enough number = of=20 patients, enrolling a smaller number of selected patients = more=20 likely to achieve the specified event, liberalizing the = definition=20 of the primary outcome measure, or by extending the duration = of=20 follow-up. Each of these strategies has its own benefits and = drawbacks. Accruing large numbers of patients and prolonging = the=20 length of the trial increase expenses and make = administration of the=20 trial more complex. Selecting an at-risk population more = likely to=20 reach a study end point is more economical but may slow down = enrollment and limit the ability to generalize the results. = One=20 important advantage of using an at-risk population in the = context of=20 a prophylaxis trial is a reduction in the number of patients = who=20 receive treatment unnecessarily, ie, patients who would not = have=20 suffered the outcome of interest even without the study=20 intervention. Another way of looking at this trade-off are = the=20 concepts of =E2=80=9Cnumber needed to treat=E2=80=9D (NNT) = and =E2=80=9Cnumber needed to=20 harm=E2=80=9D (NNH), concise expressions of the number of = patients who would=20 need to undergo the proposed intervention in order for one = patient=20 to benefit from or be harmed by the intervention. An example = of=20 these concepts as they apply to a trial of CSF prophylaxis = in=20 patients with NHL may be found in the article by Herrlinger = et al in=20 this supplement.

In a large-scale retrospective = study,=20 Hollender et al sought to determine the incidence and risk = factors=20 for CNS relapse in patients with NHL.4 A = primary=20 objective of this study was to establish risk assessment = guidelines=20 to help clinicians identify patient subpopulations that = might be=20 ideal targets for CNS prophylaxis. The probability of CNS = recurrence=20 within 5 years was calculated as a function of the number of = risk=20 factors present. CNS prophylaxis was recommended for = patients with=20 high-grade NHL and at least four risk factors (see Table 5 in the article by Herrlinger et = al in=20 this supplement).

Risk models such as this may indeed = help=20 clinicians identify high-risk patients and decide whether to = implement CNS prophylactic therapy. From a clinical trials=20 perspective, the advantage of using such a model to guide = patient=20 enrollment is the greater power to detect a treatment effect = in a=20 smaller sample size. As noted, however, the ability to = generalize=20 study results and a reduced pool of study-eligible patients = (and=20 hence slower patient accrual) are potential trade-offs.

Randomizing small patient numbers = may lead to=20 other methodologic challenges. For example, the desire to = stratify=20 patients (by study site or other variables) using the usual = blocked=20 lists for randomization may become too predictable. One = possible way=20 to avoid this predictability is to use a centralized = adaptive=20 randomization that achieves balance across stratification = variables=20 and across sites without using the usual prespecified = blocked lists.=20 With this approach, each subsequent patient randomization = depends on=20 the characteristics of the preceding randomization decision, = and in=20 this way, allocation concealment is preserved.

Because the power of a study (for = time to=20 event end points) is driven by the number of events, not the = entire=20 sample size, the usual strategy for powering a trial assumes = a=20 targeted event rate and calculates the number of patients to = achieve=20 the desired number of events. There are four basic methods = for=20 preserving power in a given trial: (1) increase the number = of=20 observed events either by extending the follow-up period, = increasing=20 the frequency of testing, or redefining the end point; (2) = increase=20 the hypothesized effect size (a strategy that is only = acceptable if=20 the investigators believe the revised effect size is valid); = (3)=20 increase alpha; or (4) use a one- versus two-tailed test. = (In some=20 contexts, a two-tailed test is not necessary; for example, = the=20 traditional Simon Two-Stage Design is a one-tailed = test.)

Commonly, enrollment is based on a = fixed=20 sample size that was generated by assuming a targeted event = rate.=20 Because the power of a trial is critically dependent on this = potentially uncertain rate, a safer approach may be to base=20 enrollment on the estimated number of events, not on sample = size=20 (ie, continue enrollment until a predefined number of events = is=20 achieved). This approach maintains statistical power but = comes with=20 the trade-off of an uncertain sample size that cannot be = specified=20 at study outset. Furthermore, this approach requires a = relatively=20 short time between the occurrence of the event and = collection of=20 pertinent data by the study personnel.

Simon Two-Stage=20 Designs

This methodology describes a = one-armed phase=20 II design that requires a relatively small sample size.42 = Two-stage=20 designs are widely used in oncology trials to reduce the = number of=20 patients placed on therapies that might be ineffective. This = design=20 provides the option to stop the trial early when clear-cut=20 termination rules are met. In the conventional two-stage = design, the=20 number of patients treated in the two stages is fixed. Once = the=20 target number of patients is enrolled in stage one, response = is=20 assessed using an interim analysis, and the study proceeds = to the=20 second stage, enrolling the remainder of patients if a = specified=20 number of responses are achieved in the first stage. In the = event=20 that the first stage fails to achieve the prespecified = number of=20 responses, the trial is terminated.

A large number of two-stage designs = (with=20 varying sizes of stages one and two) can be generated that=20 correspond to specified values of the null response rate = (treatment=20 has no effect), specific alternative response rate of = interest=20 (corresponding to a treatment effect), type I error rate, = and power.=20 Then a particular design can be selected from the large pool = based=20 on additional specified criteria. For example, Simon = originally=20 proposed selection criteria that generated what he termed = the=20 optimal and the minimax designs. The optimal design is = associated=20 with the smallest expected sample size assuming the = treatment has no=20 effect. The minimax design is associated with the smallest = overall=20 sample size. The total N for this design can be = significantly=20 smaller then the total N for the optimal design; however, = there is=20 generally a larger stage one sample size. This larger stage = one=20 sample size might be preferable for heterogeneous = populations, as a=20 very small stage one sample size may not be representative = of the=20 entire eligible population.

There are several potential problems = with any=20 two-stage design. Investigators may fail to identify a = subgroup that=20 could potentially benefit from the intervention if a trial = with=20 small numbers is prematurely terminated. In addition, the = purpose of=20 the trial might be to gather relevant safety (or other end = point)=20 information, and this might not be sufficiently accomplished = if a=20 trial is stopped early on the basis of response rate data.=20 Similarly, investigators may wish to continue the trial even = though=20 the treatment does not meet the prespecified response = criteria,=20 because they believe even the more modest observed effect is = clinically important.

Proposed GBM Phase = III=20 Study

The phase III GBM trial proposed = elsewhere in=20 this supplement considers two possible end points: survival = (4-year=20 accrual, 18-month follow-up) and PFS (4-year accrual, = 12-month=20 follow-up) (see the article by Glantz et al in this = supplement).=20 This trial assumes a median survival of 14.6 versus 19.5 = months and=20 a median PFS of 6.9 versus 9.2 months in the standard versus = investigational treatment arms, both with a hazard ratio = (HR) of=20 0.75. Survival and PFS estimates for the standard treatment = arm are=20 based on published results from a well-designed phase III = trial.43 = Survival and=20 PFS targets for the investigational arm are based on the=20 investigators' judgment about what would constitute a = clinically=20 meaningful response advantage. Table 3 provides the number of patients = required=20 to detect these differences for various values of alpha and = power=20 levels. Of note, although the medians differ for the two end = points=20 (survival v PFS), the proposed treatment effect (HR) = is the=20 same. The follow-up time and the number of patients needed = to=20 achieve this effect size are smaller when PFS is chosen as = the end=20 point, because the PFS end point results in an earlier and a = greater=20 number of events.

Table 3.

Glioblastoma Phase III = Study=20 Design (survival end point, 4-year accrual, 18-month=20 follow-up; progression-free survival end point, 4-year = accrual, 12-month follow-up)


14.6 Versus 19.5 Months Median = Survival=20 (HR, 0.75)
	Sample Size Per Arm
	85%=20 Power	90% Power
Alpha =3D 0.05	275	321
Alpha = =3D 0.10	220	262

6.9 Versus 9.2 Months Median PFS (HR, = 0.75)
	Sample Size Per Arm
	85%=20 Power	90% Power
Alpha =3D 0.05	236	276
Alpha =3D 0.10	189	225
Alpha = =3D 0.20	142	173

Abbreviations: PFS,=20 progression-free survival; HR, hazard=20 ratio.

Assumes log-rank tests = for all=20 comparisons.

Alternatively, dramatically smaller = sample=20 sizes and time required for patient accrual could be = achieved with a=20 single-arm trial using a PFS end point. However, the lack of = blinding in these types of single-arm trials has the = potential to=20 introduce substantial observer bias, and the lack of = randomization=20 requires faith that the historical control population is = very=20 comparable to the study population and that the historical = outcome=20 data are reliable.

Proposed NHL = Prophylaxis=20 Study

The phase III NHL prophylaxis study = proposed=20 by Herrlinger et al elsewhere in this supplement also would = consider=20 two end points: time to progression (2 years accrual, 2 or 3 = years=20 follow-up) and PFS (2 years accrual, 2 or 3 years follow-up) = (Table 4). Required risk factors for = study=20 eligibility would be elevated lactate dehydrogenase and = either=20 involvement of more than one extranodal site or a specific = site of=20 disease associated with a high risk of CNS relapse. The = sample size=20 calculations would assume a two-sided alpha of 0.05, both = risk=20 factors present, and a time to progression event rate of 10% = versus=20 4% and PFS of 60% versus 68.2% at 2 years (Table 4). The authors also would = consider the=20 benefit-to-risk trade-off by comparing the NNT to prevent = one case=20 of CNS relapse with the NNH. Assuming a 10% CNS relapse = rate, the=20 NNT would be 13.6 and the NNH 3.5, with an odds ratio in = favor of=20 prophylaxis of 3.9 (95% confidence interval, 1.4-11.0). = Using the=20 risk factors and assumptions outlined in their proposal, = these=20 ranges fall within what is usually considered acceptable in = other=20 disease models.

Table 4.

Non-Hodgkin Lymphoma = Phase III=20 Prophylaxis Study Design (time to progression end = point,=20 2-year accrual, alpha =3D 0.05; progression-free = survival end=20 point, 2-year accrual, alpha =3D 0.05)


TTP: Two or More Risk Factors: = 2-Year=20 Event Rate 10% v 4%
	Sample Size Per Arm
	85%=20 Power	90% Power
2-Year follow-up	236	276
3-Year=20 follow-up	180	210

PFS: 60% v 68.2% Event-Free at 2 = Years (HR,=20 0.75)
	Sample Size Per Arm
	85%=20 Power	90% Power
2-Year follow-up	460	538
3-Year=20 follow-up	379	443

Abbreviations: TTP, = time to=20 progression; PFS, profression-free survival; HR, = hazard=20 ratio.

Assumes log rank tests = for all=20 comparisons.

Once again, the length of time = patients are=20 followed (2 v 3 years) has a significant impact on = the number=20 of patients required to complete the study, since a longer = follow-up=20 period translates into a higher number of events. The = desired power=20 (85% v 90%) also influences study size and = reliability. A=20 lower power conveys a higher probability of a type II error. = For=20 example, an 80% power incurs a 20% chance of a type II = error. In=20 earlier drug development settings (phase II), adequate = control of=20 the type II (false-negative) error may be extremely = important,=20 because a false-negative result at this stage could = prematurely=20 terminate the development of a promising agent. In late = stages of=20 development, the burden shifts to avoiding a false-positive = result=20 (a type I error) to avoid the widespread use of a drug that = is=20 ineffective.44

Herrlinger et al also offer a = randomized phase=20 II =E2=80=9Cpick the winner=E2=80=9D trial design .45 In = this=20 design, the two treatment arms would each be compared = independently=20 to a prespecified historical control but not to each other. = This=20 design may be practically superior, since study size would = be=20 unreasonably large if a phase III trial design was used, and = successful accrual to a phase III trial might be very = difficult to=20 achieve in this patient population. Table 5 provides sample size estimates = for=20 various response assumptions and power choices for this = randomized=20 phase II =E2=80=9Cpick the winner=E2=80=9D design, a = randomized phase III design=20 with the end point of relapse rate, and a single-arm phase = II design=20 with an historical control comparison. These calculations = illustrate=20 the substantial variation in sample size dictated by choice = of study=20 design and by other factors such as estimates of effect size = and=20 choice of power.

Table 5.

Sample Size = Calculations for a=20 Proposed non-Hodgkin Lymphoma Prophylaxis Study Using = Various=20 Designs (primary outcome variable is defined as = relapse in the=20 central nervous = system)


Randomized Controlled Phase = III Design=20 (end point: relapse rate)
Relapse Rates	Sample Size (per arm)
Treatment 1 (L-AraC)	Treatment 2 (MTX)	80%=20 Power	90% Power
0.01	0.07	180	225
0.02	0.06	400	520
0.03	0.05	1,600	2,100

Single-Arm Phase II Model=E2=80=A0
Relapse Rates	Sample Size
Treatment (L-AraC or MTX)	Historical High-Risk Population	80%=20 Power	90% Power
0.01	0.15	35	35
0.02	0.15	35	46
0.03	0.15	46	56
0.05	0.15	75	102

Randomized Phase II =E2=80=9CPick the = Winner=E2=80=9D Design=E2=80=A1
Relapse Rates	Sample Size (per arm)
Treatment (L-AraC or MTX)	Historical High-Risk Population	80%=20 Power	90% Power
0.01	0.15	35	35
0.02	0.15	35	46
0.03	0.15	46	56
0.05	0.15	75	102
0.06	0.15	102	127
0.07	0.15	127	160

NOTE. All calculations = assume a=20 two-sided alpha of 0.05.

Abbreviations: L-AraC, = liposomal=20 cytarabine; MTX, methotrexate; NHL, non-Hodgkin=20 lymphoma.

Experimental treatment = reduces the=20 relapse rate to 1% to 3%, while the relapse rate = varied=20 between 5% and 7% in patients treated with MTX. Test: = Fisher's=20 exact test.

=E2=80=A0

Test: one-sample exact = test of a=20 binomial proportion.

=E2=80=A1

Test: Two separate = one-sample exact=20 tests of a binomial proportion. Power is for each test = = independently.

Conclusion

Definitive evidence for the = effectiveness of=20 CNS prophylaxis is lacking except in small cell lung = cancer,46, = 47 = acute=20 leukemia, and very aggressive lymphomas (ie, Burkitt = lymphoma and=20 lymphoblastic lymphoma). In other patient cohorts at high = risk for=20 CNS relapse, prophylactic therapy may represent an important = treatment option; however, the patient pool for clinical = trial=20 participation is relatively small. A small number of = patients=20 magnifies the challenges of effective trial design. = Researchers=20 involved in trial design should be particularly mindful of = the=20 following issues that could be detrimental to a successful = study. Do=20 not manipulate the effect size simply to achieve a = manageable=20 estimate of sample size. This will likely result in too few = patients=20 to arrive at a meaningful conclusion at the end of the = study.=20 Consider centralized randomization to avoid predictable = treatment=20 allocation that could bias trial results. Consider both the = risks=20 (slower accrual, possibly decreased generalizability) and = the=20 benefits (smaller study size, shorter duration, fewer = patients=20 exposed to the intervention in prophylactic trials) of = limiting the=20 study to an at-risk population. Avoiding type II error is = perhaps=20 the most important goal of early-phase trials; therefore, = consider a=20 one-tailed test or a two-stage test that does not have=20 higher-than-desired probability of early termination. Do not = modify=20 an end point simply to increase the number of events (or = power) at=20 the expense of using the most clinically relevant or=20 treatment-responsive end point. Blinded outcome assessment,=20 randomization, and stratification for prognostically = important=20 variables are also critically important considerations. More = research directed at optimal trial design and outcome = measures is=20 needed for CNS prophylaxis trials.

References

1.=20 1Boehme V, Zeynalova S,=20 Kloess M, Loeffler M, Kaiser U, = Pfeurndschuh M,=20 et al. Incidence and risk=20 factors of central nervous system recurrence in aggressive=20 lymphoma=E2=80=94a survey of 1693 patients treated in = protocols of the=20 German High-Grade Non-Hodgkin's Lymphoma Study Group=20 (DSHNHL). Ann Oncol. 2007;18:149=E2=80=93157. MED= LINE=20 | CrossRef=20

2.=20 2S=C3=B8rlie T, Perou CM,=20 Tibshirani R, Aas T, Geisler S, = Johnsen H,=20 et al. Gene = expression=20 patterns of breast carcinomas distinguish tumor subclasses = with=20 clinical implications. Proc Natl Acad Sci U S A. 2001;98:10869=E2=80=9310874. = M= EDLINE=20 | CrossRef=20

3.=20 3Haioun C, Besson C, = Lepage E,=20 Thieblemont C, Simon D, Rose C,=20 et al. Incidence and risk=20 factors of central nervous system relapse in histologically=20 aggressive non-Hodgkin's lymphoma uniformly treated and = receiving=20 intrathecal central nervous system prophylaxis: a GELA study = on 974=20 patients (Groupe d'Etudes des Lymphomes de l'Adulte). = Ann=20 Oncol. 2000;11:685=E2=80=93690. MED= LINE=20 | CrossRef=20

4.=20 4Hollender A, Kvaloy S, = Nome O,=20 Skovlund E, Lote K, Holte H. Central nervous system involvement = following=20 diagnosis of non-Hodgkin's lymphoma: a risk model.=20 Ann=20 Oncol. 2002;13:1099=E2=80=931107. ME= DLINE=20 | CrossRef=20

5.=20 5van Besien K, Ha CS, = Murphy S,=20 McLaughlin P, Rodriguez A, Amin K,=20 et al. Risk = factors,=20 treatment, and outcome of central nervous system recurrence = in=20 adults with intermediate-grade and immunoblastic = lymphoma.=20 Blood. 1998;91:1178=E2=80=931184. ME= DLINE

6.=20 6Zinzani PL, Magagnoli M,=20 Frezza G, Prologo G, Gherlinzoni F, = Bendandi M,=20 et al. Isolated = central=20 nervous system relapse in aggressive non-Hodgkin's lymphoma: = the=20 Bologna experience. Leuk Lymphoma. 1999;32:571=E2=80=93576. MED= LINE

7.=20 7Tomita N, Kodama F, = Sakai R,=20 Koharasawa H, Hattori M, Taguchi J,=20 et al. Predictive factors=20 for central nervous system involvement in non-Hodgkin's = lymphoma:=20 significance of very high serum LDH concentrations.=20 Leuk=20 Lymphoma. 2000;38:335=E2=80=93343. MED= LINE

8.=20 8Herman TS, Hammond N, = Jones SE,=20 Butler JJ, Byrne GE, McKelvey EM. = Involvement of the central nervous = system by=20 non-Hodgkin's lymphoma (The Southwest Oncology Group=20 experience). Cancer. 1979;43:390=E2=80=93397.

9.=20 9Liang R, Chiu E,=20 Loke SL. Secondary = central=20 nervous system involvement by non-Hodgkin's lymphoma: the = risk=20 factors. Hematol Oncol. 1990;8:141=E2=80=93145. MEDL= INE=20 | CrossRef=20

10.=20 10Keldsen N, Michalski W,=20 Bentzen SM, Hansen KB, Thorling K. = Risk factors for central nervous = system=20 involvement in non-Hodgkin's lymphoma=E2=80=94a multivariate = analysis. Acta Oncol. 1996;35:703=E2=80=93708. CrossRef=20

11.=20 11Litam JP, Cabanillas F,=20 Smith TL, Bodey GP, Freireich EJ. = Central nervous system relapse in = malignant=20 lymphomas: risk factors and implications for = prophylaxis.=20 Blood. 1979;54:1249=E2=80=931257. ME= DLINE

12.=20 12Ersboll J, Schultz HB,=20 Thomsen BL, Keiding N, Nissen NI. = Meningeal involvement in non-Hodgkin's = lymphoma:=20 symptoms, incidence, risk factors and treatment. = Scand J=20 Haematol. 1985;35:487=E2=80=93496.

13.=20 13Kesari S, = Batchelor TT. Leptomeningeal metastases. = Neurol = Clin.=20 2003;21:25=E2=80=9366. Full=20 Text | Full-Text=20 PDF (651 KB) | MEDLINE=20 | CrossRef=20

14.=20 14Bromberg JE, Breems DA,=20 Kraan J, Bikker G, van der Holt B, = Smitt PS,=20 et al. CSF flow = cytometry=20 greatly improves diagnostic accuracy in CNS hematologic=20 malignancies. Neurology. 2007;68:1674=E2=80=931679. CrossRef=20

15.=20 15Benesch M, Urban C. = Liposomal cytarabine for leukemic and=20 lymphomatous meningitis: recent developments. = Expert = Opin=20 Pharmacother. 2008;9:301=E2=80=93309. CrossRef=20

16.=20 16Ceresoli GL, Reni M, = Chiesa G,=20 Carretta A, Schipani S, Passoni P,=20 et al. Brain = metastases=20 in locally advanced nonsmall cell lung carcinoma after = multimodality=20 treatment: risk factors analysis. Cancer. 2002;95:605=E2=80=93612.

17.=20 17Evans AJ, James JJ,=20 Cornford EJ, Chan SY, Burrell HC, = Pinder SE,=20 et al. Brain = metastases=20 from breast cancer: identification of a high-risk = group.=20 Clin=20 Oncol (R Coll Radiol). 2004;16:345=E2=80=93349. MED= LINE=20 | CrossRef=20

18.=20 18Melisko ME, Glantz M,=20 Rugo HS. New = challenges and=20 opportunities in the management of brain metastases in = patients with=20 ErbB2-positive metastatic breast cancer. Nat Clin = Pract=20 Oncol. 2009;6:25=E2=80=9333. CrossRef=20

19.=20 19Mazhar D, Stebbing J,=20 Bower M. Non-Hodgkin's=20 lymphoma and the CNS: prophylaxis and therapy in = immunocompetent and=20 HIV-positive individuals. Expert Rev = Anticancer=20 Ther. 2006;6:335=E2=80=93341. MEDL= INE

20.=20 20Thall PF. A review of phase 2-3 clinical trial=20 designs. Lifetime Data Anal. 2008;14:37=E2=80=9353. CrossRef=20

21.=20 21Mehta MP, Shapiro WR,=20 Glantz MJ, Patchell RA, Weitzner MA, = Meyers CA,=20 et al. Lead-in = phase to=20 randomized trial of motexafin gadolinium and whole-brain = radiation=20 for patients with brain metastases: centralized assessment = of=20 magnetic resonance imaging, neurocognitive, and neurologic = end=20 points. J Clin Oncol. 2002;20:3445=E2=80=933453. CrossRef=20

22.=20 22Meyers CA, Smith JA, = Bezjak A,=20 Mehta MP, Liebmann J, Illidge T,=20 et al. Neurocognitive=20 function and progression in patients with brain metastases = treated=20 with whole-brain radiation and motexafin gadolinium: results = of a=20 randomized phase III trial. J Clin = Oncol.=20 2004;2:157=E2=80=93165.

23.=20 23Gelber RD, Cole BF, = Gelber S,=20 Goldhirsch A. Comparing=20 treatments using quality-adjusted survival: the Q-TwiST=20 method. Am Stat. 1995;49:161=E2=80=93169.

24.=20 24Recht L, Glantz M,=20 Chamberlain M, Hseih CC. Qualitative measurement of quality = outcome in=20 malignant glioma patients using an independent living score=20 (ILS). J Neurooncol. 2003;61:127=E2=80=93136. CrossRef=20

25.=20 25Cole BF, Glantz MJ,=20 Jaeckle K. Quality-of-life-adjusted survival = comparison of=20 sustained release cytosine arabinoside versus intrathecal=20 methotrexate for treatment of solid tumor neoplastic=20 meningitis. Cancer. 2003;97:3053=E2=80=933060.

26.=20 26Glantz M, Conlee J. = High grade gliomas. = In: Meyers C, = Perry A=20 editor. Cognition and=20 cancer. Cambridge, UK: Cambridge University Press; = 2008;p. 156=E2=80=93169. =

27.=20 27Benedict RHB, Schretlen D, = Groninger L, Brandt J. Hopkins Verbal Learning Test-Revised: = normative=20 data and analysis of inter-form and test-retest = reliability.=20 Clin=20 Neuropsychol. 1998;12:43=E2=80=9355.

28.=20 28Lezak MD, Howleson DB,=20 Loring DW. Neuropsychological=20 assessment. New = York: Oxford=20 University Press; 1994;.=20

29.=20 29Lezak MD. Neuropsychological assessment (ed = 3).=20 New York: Academic Press; = 1995;.

30.=20 30Benton AD. Differential behavioral effects in = frontal lobe=20 disease. Neuropsychologia. 1968;6:53=E2=80=9360. CrossRef=20

31.=20 31Benton AL, Hamsher K = de S.=20 Multilingual Aphasia = Examination.=20 Iowa City, IA: AJA = Associates; 1989;.

32.=20 32Wechsler D. Wechsler Adult Intelligence=20 Scale=E2=80=93Revisited. San Antonio,=20 TX: The Psychological Corp; 1981;.

33.=20 33Folstein MF, Folstein SE,=20 McHugh PR. Mini-mental state=E2=80=9D:=20 a practical method for grading the cognitive state of = patients for=20 the clinician. J Psychiatr Res. 1975;12:189=E2=80=93198. MED= LINE=20 | CrossRef=20

34.=20 34Smith J, Wefel J. = Neurocognitive testing in clinical=20 trials. In: = Meyers C, = Perry A=20 editor. Cognition and=20 cancer. Cambridge, UK: Cambridge University Press; = 2008;p. 320=E2=80=93328. =

35.=20 35Corn BW, Moughan J,=20 Knisely JPS, Fox SW, Chakravarti A, = Yung WK,=20 et al. Prospective=20 evaluation of quality of life and neurocognitive effects in = patients=20 with multiple brain metastases receiving whole-brain = radiotherapy=20 with or without thalidomide on Radiation Therapy Oncology = Group=20 (RTOG) trial 0118. Int J Radiat Oncol Biol Phys. = 2008;71:71=E2=80=9378. Abstract=20 | Full-Text=20 PDF (759 KB) | CrossRef=20

36.=20 36Mehta MP, Rodrigus P,=20 Terhaard CHJ, Rao A, Suh J, Roa W,=20 et al. Survival = and=20 neurologic outcomes in a randomized trial of motexafin = gadolinium=20 and whole-brain radiation therapy in brain = metastases.=20 J Clin=20 Oncol. 2003;21:2529=E2=80=932536. CrossRef=20

37.=20 37Melisko ME, Kunwar S, = Prados M,=20 Berger MS, Park JW. Brain=20 metastases of breast cancer. Expert Rev = Anticancer=20 Ther. 2005;5:253=E2=80=93268. MEDL= INE

38.=20 38Shah BD, Kesari S, = Xu R,=20 Batchelor TT, O'Neill AM, Hochberg FH,=20 et al. Comparison of=20 linear and volumetric criteria in assessing tumor response = in adult=20 high-grade gliomas. Neurooncology. 2006;8:38=E2=80=9346. MEDLI= NE=20 | CrossRef=20

39.=20 39Ananthnarayan S, Bahng J,=20 Roring J, Nghiemphu P, Lai A, = Cloughesy T,=20 et al. Time = course of=20 imaging changes of GBM during extended bevacizumab = treatment.=20 J=20 Neurooncology. 2008;88:339=E2=80=93347.

40.=20 40Norden AD, Young GS,=20 Setayesh K, Muzikansky A, Klufas R, = Ross GL,=20 et al. Bevacizumab for=20 recurrent malignant gliomas: efficacy, toxicity, and = patterns of=20 recurrence. Neurology. 2008;70:779=E2=80=93787. CrossRef=20

41.=20 41de Groot JF, = Yung WK. Bevacizumab and irinotecan in the = treatment of=20 recurrent malignant gliomas. Cancer = J. 2008;14:279=E2=80=93285.

42.=20 42Simon R. Optimal two-stage designs for phase II = clinical=20 trials. Control Clin Trials. 1989;10:1=E2=80=9310. MEDLI= NE=20 | CrossRef=20

43.=20 43Stupp R, Mason WP, van den = Bent MJ, Weller M, Fisher B, = Taphoorn MJ,=20 et al. Radiotherapy plus=20 concomitant and adjuvant temozolomide for = glioblastoma.=20 N Engl J=20 Med. 2005;352:987=E2=80=93996. CrossRef=20

44.=20 44Lee JJ, Feng L. = Randomized phase II designs in cancer = clinical=20 trials: current status and future directions. = J Clin = Oncol.=20 2005;23:4450=E2=80=934457. CrossRef=20

45.=20 45Simon R, Wittes RE,=20 Ellenberg SS. Randomized phase=20 II clinical trials. Cancer Treat Rep. 1985;69:1375=E2=80=931381. ME= DLINE

46.=20 46Aup=C3=A9rin A, = Arriagada R,=20 Pignon JP, Le P=C3=A9choux C, Gregor A, = Stephens RJ,=20 et al. Prophylactic=20 cranial irradiation for patients with small-cell lung cancer = in=20 complete remission (Prophylactic Cranial Irradiation = Overview=20 Collaborative Group). N Engl J Med. 1999;341:476=E2=80=93484. ME= DLINE=20 | CrossRef=20

47.=20 47Lee JJ, Bekele BN, = Zhou X,=20 Cantor SB, Komaki R, Lee JS. Decision analysis for prophylactic = cranial=20 irradiation for patients with small-cell lung cancer. = J Clin=20 Oncol. 2006;24:3597=E2=80=933603. CrossRef=20

a Geron Corp, Menlo Park, CA

b Department of Neurosurgery, Penn = State Hershey=20 Medical Center, Hershey, PA

Address = correspondence=20 to Jennifer A. Smith, PhD, Geron Corp, 230 Constitution Dr, = Menlo=20 Park, CA 94025

STATEMENT OF CONFLICT OF = INTEREST:=20 M.G. has received honoraria and research support from Enzon=20 Pharmaceuticals. J.A.S. has nothing to disclose.

PII: S0093-7754(09)00098-0

doi:10.1053/j.seminoncol.2009.05.008

=C2=A9 2009 = Published by=20 Elsevier Inc.

3D""=20
10 of 10=20