A comparison of the efficacy and safety of oral and intravenous fludarabine in chronic lymphocytic leukemia in the LRF CLL4 trial




An oral formulation of fludarabine was introduced for use in chronic lymphocytic leukemia in 2001 following studies demonstrating the bioequivalence of a 40 mg/m2 oral dose with a 25 mg/m2 intravenous dose. We assessed retrospectively the efficacy of these two routes of administration in the LRF CLL4 trial.


A total of 777 patients were randomized from 1999-2004 to receive fludarabine, alone or with cyclophosphamide, or chlorambucil. In 2001, a protocol amendment allowed the oral formulation. There were 117 assessable patients who received fludarabine intravenously and 252 who received it orally. A total of 387 patients given chlorambucil acted as a control group.


Patients given oral fludarabine were less likely to receive the full dose (P = .0004) and experienced more, predominantly gastrointestinal, toxicity. Progression-free survival (PFS) and overall survival were not affected by the route of administration (PFS hazard ratio, 1.10; 95% confidence interval, 0.87-1.40), but the overall rate of response to treatment appeared to be lower with the oral formulation (P = .003). However, patients recruited since 2001 were older (P = .03) and were more likely to have TP53 deletion, and response rates after 2001 were also lower in the chlorambucil group. After excluding patients with TP53 deletion, no significant difference in outcome was attributable to the route of administration.


Although the LRF CLL4 data suggest no important difference in the effectiveness of oral compared with intravenous fludarabine, randomized trials are needed to reliably evaluate this comparison, particularly in combination with rituximab. Meanwhile, it is important to monitor compliance and gastrointestinal side effects with the oral route and to switch to intravenous therapy if a reduced dose is being received. Cancer 2011. © 2010 American Cancer Society.

In 2001, an oral formulation of fludarabine was licensed for the treatment of chronic lymphocytic leukemia (CLL). The oral formulation is now available in over 40 countries, and it was approved by the US Food and Drug Administration in December 2008. A phase 2 study suggested there was no difference in clinical efficacy between the oral and intravenous (iv) drug,1 and subsequent studies have confirmed this impression, whether used alone2, 3 or in combination with oral cyclophosphamide.4-6 Furthermore, the cost-effectiveness and convenience of oral over iv administration has provided a driver for the introduction of oral fludarabine in many countries. Not only is it 5% cheaper than the iv formulation, but there are no associated clinic and nursing costs, reducing the overall cost per cycle of treatment to an estimated half of that of the iv route. Elderly patients, in particular, can benefit from being treated at home and thereby participate more readily in clinical trials. However, there has never been a comparison between oral versus iv fludarabine in a phase 3 trial, and questions remain as to whether the efficacy and toxicity are the same.

A multicenter randomized clinical trial was undertaken in the United Kingdom (LRF CLL4) to compare the safety and efficacy of chlorambucil to fludarabine, used either alone or with cyclophosphamide (FC), in previously untreated patients with CLL requiring therapy. During the course of the trial, oral fludarabine became available, and a protocol amendment allowed its use at a bioequivalent dose.7 Although the comparison of oral versus iv fludarabine was not planned as a prospective analysis, the availability of the oral form provided an opportunity to compare the efficacy and safety of the 2 routes of administration. The protocol for patients randomized to receive chlorambucil remained unchanged, and these patients therefore formed a control group against which to test whether any changes in clinical outcomes after the introduction of oral fludarabine were attributable to the oral formulation or were the result of other changes that took place over time. We present here the results of this retrospective analysis.


Over a 66-month period from February 1999 to October 2004, 387 patients in the LRF CLL4 trial were randomized to receive chlorambucil, 194 received fludarabine alone, and 196 received fludarabine with cyclophosphamide. The diagnosis of CLL was confirmed via central review of morphology and immunophenotype with 5 markers: CD5, CD23, CD79b, FMC7, and surface immunoglobulins with antibodies against light chains.

Other variables recorded at trial entry (either in all or in the majority of patients) and included in analyses were: stage of disease, age, gender, β2 microglobulin level, lactate dehydrogenase level, and absolute lymphocyte count. A range of other prognostic markers was analyzed centrally for the majority of trial patients: immunoglobulin heavy chain variable gene (IGHV) mutation status, CD38, ZAP-70 expression, and cytogenetics via fluorescence in situ hybridization (FISH). Five FISH probes were used: for trisomy 12 and deletions at 6q21, 11q23, 13q14, and 17p13 (the TP53 locus).

Case report forms were kept simple to maximize recruitment and answer the primary study question. Questions concerning the treatment given were: date started, number of courses, administration route, whether the full dose was given and date completed. Information was not collected concerning the details of any dose reductions, nor whether these were protocol-mandated (ie, due to persistent neutropenia or thrombocytopenia or reduced kidney function) or the result of individual physician or patient decisions. The definition of full dose was as per protocol. No guidance was given on how much of a reduction from this should be classified as not full dose, but was according to the individual physician's judgement.

Chlorambucil was administered orally, 10 mg/m2/d for 7 days, every 28 days, for up to 12 cycles. Fludarabine was administered initially iv, 25 mg/m2/d for 5 days, or for 3 days together with cyclophosphamide 250 mg/m2/d for 3 days. When the protocol amendment allowed oral fludarabine, after February 2001, patients receiving oral fludarabine monotherapy were given fludarabine at a dosage of 40 mg/m2/d for 5 days, and those receiving oral FC were prescribed fludarabine at a dosage of 24 mg/m2/d together with cyclophosphamide at a dosage of 150 mg/m2/d for 5 days. The change from 3 to 5 days for the oral combination was undertaken to improve gastrointestinal tolerance and patient compliance. Whether administered iv or orally, fludarabine and FC were given every 28 days for up to 6 cycles.

All patients provided written informed consent. The trial was approved by a United Kingdom multicenter research ethics committee and followed UK Medical Research Council guidelines for good clinical practice. The study was registered as an International Standard Randomized Controlled Trial (ISRCTN58585610). Primary results were published in 2007.8

Statistical Methods

The response recorded was the best achieved at any time due to first-line treatment. Overall survival was calculated from randomization to death from any cause. Progression-free survival (PFS) was time from randomization to relapse needing further treatment, or to other progression date if reported, or death from any cause. (Relapse needing further treatment was defined as any of the following: lymphocyte doubling time <12 months; downward trend in hemoglobin level and/or platelet count; ≥50% increase in size of liver, spleen, or lymph nodes; appearance of lymphadenopathy, hepatomegaly, or spenomegaly if not already present; or B-symptoms not attributable to other causes.) For nonresponders and those with progressive disease, the date of progression was when nonresponse or progressive disease was recorded.

All main analyses were intention to treat; all patients were analyzed according to the initial treatment given and the initial route of administration. Comparisons between categorical variables were made using the chi-squared test; comparisons between quantitative variables were made using the Wilcoxon rank sum test. Cochran-Mantel-Haenszel statistics were used to examine associations between toxicities and route of administration or treatment, allowing for the other variable. Logistic regression models were used to determine whether response was associated with iv versus oral fludarabine independently of other factors. Multivariate Cox regression analysis was used to determine which factors were independently associated with PFS and survival. All P values were 2-sided.



The route of administration of fludarabine was known in 369 cases. Of these, 32% received it intravenously (half of whom were treated before the oral option became available in 2001, the other half of whom were treated subsequently) and 68% received it orally (Table 1). In 2001, 70% of patients given fludarabine received the oral formulation, increasing to 84% in 2002, 88% in 2003, and 89% in 2004 (P for trend = .001).

Table 1. Number of Patients Receiving Intravenous and Oral Fludarabine
 Fludarabine (n=194)Fludarabine + Cyclophosphamide (n=196)Total (N=390)
  • *

    Among these patients, 3 received chlorambucil, 1 received fludarabine + cyclophosphamide, 1 died before treatment, and 1 failed to return to the clinic.

  • Among these patients, 2 received chlorambucil and 1 died before treatment.

Unknown whether intravenous/oral7512
Allocated but not given6*39

In the trial as a whole, the 641 patients randomized in 2001-2004 were older (median age, 65 years; range, 35-86 years) than the 136 patients randomized in 1999-2000 (median age, 63 years; range, 42-84 years) and, in those tested, more likely to be in the poor risk category (>10% TP53-deleted cells),9 but there were no differences in the men:women ratio or disease stage at entry (Table 2). Patients who received fludarabine orally were no different in gender or disease stage from those who received it intravenously, but they were older, and all 18 patients with TP53 deletion who were allocated to fludarabine or FC received it orally (Table 2). The proportion of patients who received the full dose was similar over time in the chlorambucil group, but for those on fludarabine or FC, it was greater overall among those randomized in 1999-2000 (Table 3). However, it was similar over time in the iv group, and this difference was due to it being less likely that the full dose was received if the drug was given orally. The older the patient, the more likely they were to receive a reduced dose, whether given iv or orally (Table 4). Reasons for the reduced dose were not recorded. There was no significant difference in the number of cycles of treatment given by route of administration (data not shown).

Table 2. Demographic Characteristics of LRF CLL4 Patients by Trial Entry Date, Drug, and Intravenous Versus Oral Fludarabine
 All Trial PatientsFludarabine (Alone or with Cyclophosphamide)*Chlorambucil
1999-2000 (n=136)2001-2004 (n=641)PIntravenous (n=117)Oral (n=252)P1999-2000 (n=66)2001-2004 (n=321)P
  • All data are presented as n (%).

  • NS indicates not significant.

  • *

    Excludes 21 patients whose route of administration was unknown or who did not receive the treatment allocated.

  • Wilcoxon rank sum test.

Age, y  .04  .01  NS
 <6056 (41)199 (31) 48 (41)74 (29) 27 (41)100 (31) 
 60-6945 (33)243 (38) 40 (34)96 (38) 22 (33)122 (38) 
 ≥7035 (26)199 (31) 29 (25)82 (33) 17 (26)99 (31) 
Gender  NS  NS  NS
 Men105 (77)468 (73) 87 (74)185 (73) 15 (23)86 (27) 
 Women31 (23)173 (27) 30 (26)67 (27) 51 (77)235 (73) 
Binet stage  NS  NS  NS
 A progressive30 (22)161 (25) 28 (24)61 (24) 12 (18)84 (26) 
 B64 (47)288 (45) 59 (50)113 (45) 32 (48)140 (44) 
 C42 (31)192 (30) 30 (26)78 (31) 22 (33)97 (30) 
Tested for TP53 deletion108463 77198 52228 
TP53-deleted2 (2)31 (7).050 (0)18 (9).0062 (4)13 (6)NS
Table 3. Number of Patients Receiving Full Dose Chemotherapy by Treatment, Route of Administration, and Trial Entry Date
TreatmentRoute of AdministrationTrial Entry DateNo. of Patients AssessableNo. of Patients Receiving Full Dose (%)P
  1. Patients who were not given their allocated treatment (chlorambucil, n = 3; fludarabine, n = 6; fludarabine with cyclophosphamide, n = 3) or for whom dose information was missing (chlorambucil, n = 5; fludarabine, n = 3; fludarabine with cyclophosphamide, n = 2) were excluded.

  2. NS indicates not significant; F, fludarabine; FC, fludarabine with cyclophosphamide; IV, intravenous.

Chlorambucil 1999-20006454 (84)NS
  2001-2004315261 (83)
FAll1999-20003230 (94).06
  2001-2004153122 (80)
 IV1999-20003230 (94)NS
  2001-20042525 (100)
 IVAll years5755 (96).0004
 Oral2001-200412291 (75)
FCAll1999-20003329 (88)NS
  2001-2004158127 (80)
 IV1999-20003329 (88)NS
  2001-20042623 (88)
 IVAll years5952 (88)NS
 Oral2001-200412799 (78)
F and FCAll1999-20006559 (91).03
  2001-2004311249 (80)
 IVAll years116107 (92).0003
 Oral2001-2004249190 (76)
Table 4. Number of Patients Receiving Full Dose by Drug, Route of Administration, and Age
Drug and Route of AdministrationAge, yNo. of Patients AssessableNo. of Patients Receiving Full Dose (%)P for Trend
  1. Patients who were not given their allocated treatment or for whom dose information was missing were excluded. The interaction between age and route of administration was not significant.

Intravenous fludarabine<604847 (98).05
 60-694036 (90)
 ≥702824 (86)
Oral fludarabine<607362 (85).009
 60-699473 (78)
 ≥708255 (67)
Chlorambucil<60123103 (84).3
 60-69142123 (87)
 ≥7011489 (78)


Toxicity by treatment group has been reported previously.8 Briefly, neutropenia, nausea and vomiting, alopecia, and any World Health Organization grade 3 or 4 toxicity were significantly more frequent with FC than with either chlorambucil or fludarabine alone (all P < .0001), as were diarrhea (P = .01) and “other toxicity” (P = .004). Hemolytic anemia was less frequent with FC (P = .01).10 These differences remained significant after adjustment for the oral or iv route for fludarabine, apart from grade 3 or 4 diarrhea, which was no longer significantly different between fludarabine and FC. Hemolytic anemia was more common with the oral than the iv fludarabine formulation, as were diarrhea and “other toxicity,” but no other differences in toxicity between the routes of administration were seen (Table 5). It is not known how many patients experiencing toxicity with the oral form were switched to the iv route, but the number was estimated to be low and was unlikely to have influenced these results.

Table 5. Percentage of Fludarabine-Treated Patients Experiencing Toxicity by Intravenous Versus Oral Treatment Route
Toxicity*No. of Patients Assessable% with ToxicityP (Adjusted)
  • NS indicates not significant.

  • *

    World Health Organization grades 3 and 4 only, unless specified otherwise.

  • Includes only patients randomized to fludarabine or fludarabine plus cyclophosphamide who received the allocated treatment by known route (n=369) and where individual toxicity question was completed.

  • Allowing for age and treatment.

  • §

    One or more episodes.

  • ||

    Where the grade was not specified, grade 1 or 2 toxicity was assumed. Cases of “other toxicity” (n=106) included skin rash (25 cases), infection (22 cases), cardiovascular event (10 cases), fatigue (6 cases), and constipation (5 cases).

Hemolytic anemia364310.03 (.06)
Febrile episodes§3562835NS
Nausea and vomiting358119NS
Nausea and vomiting (all grades)3584243NS
Diarrhea (all grades)3471528.01 (.01)
Other toxicity||298310.05 (.07)
Other toxicity (all grades)2982341.002 (.005)

Allowing for age, treatment, and treatment route, doses were significantly more likely to be reduced in patients experiencing febrile episodes (P = .01), thrombocytopenia (P = .007) and “other toxicity” (P = .02), and may have been reduced in some cases of neutropenia or anemia. Doses were not apparently reduced as a result of any of the other categories of recorded toxicities. In an analysis of dose by route, with age, treatment, and anemia or “other toxicity” included in the model, the route still affected the dose significantly (P = .006).

Response to Treatment

As previously reported, the overall response rate and “good” response rate (complete response plus nodular partial response) were significantly better with FC than with either chlorambucil or fludarabine alone (P < .0001).8 Responses appeared less good with the oral formulation of fludarabine than with iv, including in those who received the full dose; however, the differences were less marked when the TP53-deleted cases were excluded (Table 6). Response rates were lower in patients receiving a reduced dose compared with patients receiving the full dose, irrespective of the route of administration (Table 6). In addition, a (nonsignificant) decline in response rates was seen during the course of the trial in all age groups, including in the chlorambucil group (Table 7).

Table 6. Response Rates by Oral Versus Intravenous Fludarabine
 IVOralOral Excluding TP53 DeletionP
IV vs OralIV vs Oral Excluding TP53 Deletion
  • Only patients who received their allocated treatment and for whom response was assessable were included.

  • IV indicates intravenous; CR, complete response; NodPR, nodular partial response; NS, not significant.

  • *

    P values for all patients receiving versus not receiving the full dose were: overall response, P = .07 (not significant); CR and NodPR, P = .001.

Fludarabine alone     
 Total patients56116108  
 Overall response51 (91%)87 (75%)84 (78%).01.03
 CR and NodPR27 (48%)48 (42%)47 (44%)NSNS
Fludarabine plus cyclophosphamide   
 Total patients57118110  
 Overall response56 (98%)108 (92%)106 (96%)NSNS
 CR and NodPR39 (68%)67 (57%)66 (60%)NSNS
Full dose received*     
 Total patients104180167  
 Overall response99 (95%)153 (85%)148 (89%).009NS
 CR and NodPR64 (62%)96 (53%)94 (56%)NSNS
Full dose not received*     
 Total patients85249  
 Overall response7 (87.5%)41 (79%)41 (84%)NSNS
 CR and NodPR2 (25%)18 (35%)18 (37%)NSNS
 Total patients113234218  
 Overall response107 (95%)195 (83%)190 (87%).003.03
 CR and NodPR66 (58%)115 (49%)113 (52%)NSNS
Table 7. Response Rates by Treatment and Trial Entry Datea
 Trial Entry DateP
  • a

    Only patients who received their allocated treatment and for whom response was assessable were included.

  • NS indicates not significant; CR, complete response; NodPR, nodular partial response.

Fludarabine alone   
 Total patients32145 
 Overall response29 (91%)113 (78%)NS
 CR and NodPR15 (47%)61 (42%)NS
Fludarabine plus cyclophosphamide 
 Total patients33147 
 Overall response32 (97%)137 (93%)NS
 CR and NodPR23 (70%)86 (58.5%)NS
 Total patients63303 
 Overall response52 (83%)214 (71%)NS
 CR and NodPR21 (33%)77 (25%)NS

Because the comparison of oral and iv administration was not randomized, an adjustment for possible confounding factors is necessary. A multivariate analysis of response was therefore conducted for several variables, including age, randomized treatment, disease stage, gender, and route of administration. Because no patients in the iv-treated group had TP53 deletion, and this group is known to respond extremely poorly ,9, 11 they were excluded from the analysis. In the remaining 331 patients receiving fludarabine or FC as allocated and whose administration route was known, treatment was the only significant variable (P < .0001), with better response rates for FC (odds ratio [OR], 2.43; 95% confidence interval [CI] , 1.58-3.73). The OR for iv treatment was 1.48 (95% CI, 0.94-2.33; P = .09). Restriction to the group that received the full dose did not materially alter this finding (OR, 1.42; 95% CI, 0.86-2.35). Other risk factors (11q deletion, elevated β2 microglobulin level, and unmutated IGHV genes and/or IGHV3-21 usage) were available for only 152 of the patients in this group, and in this subset the OR for iv treatment was 1.20 (95% CI, 0.61-2.36).

PFS and Overall Survival

At a median follow-up of 8 years for the iv group and 6 years for the oral group, there was no significant difference between groups in either PFS or overall survival, either overall (P = 1.00, 0.25, respectively), or within the fludarabine and FC treatments (Figure 1). In multivariate analyses in the trial as a whole, prognostic factors found to be significant were unmutated IGHV genes and/or IGHV3-21 usage, 11q deletion, elevated β2 microglobulin level and treatment allocation, for predicting shorter PFS, and older age, elevated β2 microglobulin level, unmutated IGHV genes and/or IGHV3-21 usage and treatment allocation, for predicting shorter overall survival.9 In Cox regression analyses, excluding TP53-deleted cases, the hazard ratios (HR) for oral versus iv administration were 1.02 (95% CI, 0.80-1.31) for PFS and 1.07 (95% CI, 0.78-1.48) for overall survival, in a model including only treatment and route. In subsets with complete data, the HRs were modified from 0.85 (95% CI, 0.59-1.21) to 0.91 (95% CI, 0.63-1.31) for PFS and from 0.81 (95% CI, 0.52-1.27) to 0.82 (95% CI, 0.52-1.29) for overall survival via inclusion of all relevant prognostic factors into the model.

Figure 1.

A comparison of intravenous and oral fludarabine is shown for (A) progression-free survival and (B) overall survival. There were no significant differences between the oral and intravenous routes. Cases with TP53 deletion occurred only in patients given the oral formulation and are not included.


On first examination of the results, it seems clear that the overall response rate is higher in the group of patients receiving iv versus oral fludarabine, alone or in combination with cyclophosphamide. However, iv versus oral fludarabine (with or without oral cyclophosphamide) was not a randomized comparison in the LRF CLL4 trial. It is therefore necessary to make adjustments for confounding factors when interpreting the results. Of note, a decline in response rates over the course of the trial affected the chlorambucil arm, for which no treatment change was introduced, as well as the fludarabine and FC arms. Therefore, other variables acting over time must have affected outcome for patients regardless of the treatment they received. Two main factors are implicated: 1) the recruitment of older age patients, possibly as a result of the more readily administered oral treatment becoming available; 2) the increased numbers of patients with TP53 deletion. The latter information was only available retrospectively, and genetic risk was not controlled for in the trial randomization. Unexpectedly, all the patients with TP53 deletion, which was associated with low response rates across all treatment arms,9 received oral rather than iv fludarabine.

The trial results showed that, although older as well as younger patients responded better to FC than to fludarabine alone or chlorambucil, older age was associated with slightly poorer responses overall.8 The changes in age demographics over the course of the trial therefore will have had an impact on the results. It is more difficult to judge any specific impact of older age on the efficacy of oral versus iv administration. Older patients were less likely than younger ones to receive the full dose, particularly if they received the oral formulation. Age-related physiological changes may affect the pharmacology (absorption, distribution, metabolism, renal clearance), but there are no reliable data to confirm this. Furthermore, the fact that older patients may be receiving more concomitant medications, and also may be less reliable in administering the oral drugs, could have an impact.

Patients of all ages were more likely to receive a reduced dose if they were given oral rather than iv fludarabine (24% vs 8%). Responses were poorer with the oral formulation even among patients receiving the full dose, but were also poorer in patients who received a reduced dose compared with patients receiving the full dose, irrespective of the route of administration. Clearly, some patients received a reduced dose due to toxicity, but it is not known which patients changed from oral to iv, and it is therefore not possible to assess whether this change-over explains why some patients with gastrointestinal toxicity were still able to receive the full dose. The dosing schedule for oral administration is complex, involving a different number of tablets on different days to make up the full dose and usually involving other medications, such as anti-emetics, co-trimoxazole, and fluconazole, each with an individual dosing schedule. For patients likely to find compliance difficult, it may be helpful if they are either offered treatment by the iv route or given aids to increase compliance, such as a chart showing the schedule for all medications on a daily basis.

In multivariate analysis, route of administration was not significantly associated with response, either when the model included randomized treatment, disease stage, age, and gender or, in the subset with genetic risk factors measured, when it included randomized treatment and other risk factors. Nor was the result any different when only patients who received the full dose were included. This suggests that the route of administration did not influence the response rate after adjustment for known confounding factors.

Importantly, there was no significant difference in either PFS or overall survival between the oral and iv fludarabine treatment groups. There was some increased toxicity with the oral formulation, including more reported cases of diarrhea, which could have resulted in some patients failing to receive optimal doses of treatment, and “other toxicity,” including skin rashes, infections and cardiovascular events, which may also have resulted in dose reductions. This observation raises awareness of the need to switch patients from oral to iv therapy if they experience gastrointestinal toxicity and, where possible, in cases of other toxicity.

The changes in results over time in the LRF CLL4 trial raise a number of issues in relation to the conduct of trials in general, especially those recruiting patients over a protracted time period. There is a need to be aware that the generalizability of results is related both to the patient population recruited and to management issues. Even in the chlorambucil arm, there would have been different response rates seen at an interim recruitment point compared with the end of the trial. This highlights the difficulty of extrapolating data from small nonrandomized phase 2 trials and comparing results with historical studies. In addition, it has raised our awareness of the inherent difficulties in interpretation of the data after a trial amendment is made part-way through a study, even if no effect on outcome is anticipated.

Although the treatments given in the LRF CLL4 trial have now been superseded by the introduction of chemo-immunotherapy combinations, the question of the efficacy of oral fludarabine compared with iv remains highly relevant. Fludarabine is still a major component of most regimens for the treatment of CLL (notably FC plus rituximab, which has been administered iv thus far12, 13) and the ability to deliver the drug as an oral therapy offers practical and cost advantages. Overall, there was no evidence of any clinically relevant difference in trial endpoints for patients treated on LRF CLL4 which could be attributed to the route of administration of fludarabine. However, this was a retrospective comparison, and large randomized trials are needed to reliably evaluate the efficacy of oral versus iv fludarabine, particularly in combination with rituximab and other chemo-immunotherapy agents. Meanwhile, it is important to recognize issues both of compliance and tolerability with the oral form, particularly gastrointestinal toxicity, which can lead to a reduced dose being received, and to switch to iv therapy in such cases to ensure that the optimal dose is given.


We thank the National Cancer Research Institute Chronic Lymphocytic Leukaemia Working Group, and all doctors and patients involved, for their participation in this study.


The LRF CLL4 trial was funded by a core grant from Leukaemia Research UK. Laboratory studies were funded by an educational grant from Schering Health Care (United Kingdom) and Schering AG (Germany). The Clinical Trial Service Unit received research support from the Medical Research Council and Cancer Research UK. M.E. was supported by the Arbib Foundation. C.E.D. has acted as a consultant for Roche, BSPharma, and Genzyme Inc. D.C. has acted as a consultant for and has received research funding from Roche, GlaxoSmithKline , Parexel, and Schering AG. P.H. has acted as a consultant and received research support from Genzyme Inc, Roche, and GlaxoSmithKline .