Allogeneic hematopoietic cell transplantation (HCT) is frequently complicated by acute and chronic graft-versus-host disease (GVHD) [1, 2]. Although considerable progress has been made in the development of methods to prevent or treat acute GVHD, similar progress in chronic GVHD has languished by comparison after the clinical and pathologic features of this syndrome were first described in 1980 [3]. Interest in this debilitating complication of HCT was rejuvenated when recommendations of the National Institutes of Health Consensus Conference on Criteria for Clinical Trials in Chronic Graft-versus-host disease were published in 2005-2006 [4-9].

The Consensus Conference recognized two major categories of GVHD, each with 2 subcategories [4]. Acute GVHD with onset before day 100 was defined as "classic GVHD." A separate category was recognized for persistent, recurrent or late-onset acute GVHD beyond day 100 after HCT. Chronic GVHD was separated from acute GVHD not by time from HCT but by the presence of diagnostic criteria or by distinctive findings supported by biopsy or other procedures. Classic chronic GVHD was defined by unambiguous chronic GVHD manifestations in the absence of abnormalities such as cutaneous erythema, liver function abnormalities, or gastrointestinal manifestations typical of acute GVHD. "Overlap syndrome" is a subcategory of chronic GVHD characterized by chronic GVHD in the presence of one or more manifestations typical of acute GVHD.

Chronic GVHD is a pleomorphic syndrome with "autoimmune" features that sometimes resemble clinical findings in scleroderma and Sjögren syndrome. The onset usually occurs between 3 and 15 months after HCT [10-13]. Risk factors associated with an increased risk of chronic GVHD include the use of a mobilized blood cell graft or an HLA-mismatched or unrelated donor, older patient age and a history of acute GVHD [12]. The risk of chronic GVHD can be decreased by exhaustive depletion of T cells from the graft or by treatment of the recipient with rabbit antibodies specific for human T cells as part of the conditioning regimen before HCT [12, 14, 15]. Without these measures, approximately 30% to 50% of HCT recipients develop chronic GVHD [2, 16].

Chronic GVHD can affect multiple organs and sites, including the skin and subcutaneous connective tissues, lacrimal and salivary glands, oral mucosa, lungs, esophagus, joints, gastrointestinal tract and liver. The disease is characterized by immune dysfunction with an increased risk of infections and a 30% to 50% risk of mortality during the first 5 years after diagnosis [10, 11]. Chronic GVHD has been associated with a reduced risk of recurrent malignancy after HCT [17-22], but despite this benefit, survival is not improved [22]. A prognostic scoring system has recently been proposed based on factors present at the time of chronic GVHD diagnosis [13].

To date, only 6 randomized phase III studies have ever been reported for initial treatment of chronic GVHD [23-28]. The study by Koc et al. [26] was the only one that indicated benefit. Results of this study suggested that treatment with cyclosporine reduced the amount of glucocorticoid treatment needed to control the disease, as indicated by a decreased frequency of avascular necrosis. The generally recommended approach for treatment of chronic GVHD involves continued administration of the calcineurin inhibitor used for GVHD prophylaxis together with prednisone initially at 1 mg/kg/day [2, 29, 30]. Strategies for tapering the dose of prednisone vary considerably, but as a general principle, efforts should be made to use the minimum dose that is sufficient to control GVHD manifestations. At our center, the dose of prednisone is tapered to an alternate-day schedule of administration after initial clinical improvement, which generally occurs within 6 weeks after starting treatment. The dose of prednisone is then tapered to 0.5 mg/kg every other day and generally continued until reversible manifestations of the disease resolve. The dose of prednisone is then gradually tapered with careful monitoring for recurrent manifestations of chronic GVHD. Doses of the calcineurin inhibitor are gradually decreased after treatment with prednisone has been withdrawn.

The median duration of treatment for chronic GVHD is approximately 2 years in patients who had HCT with marrow cells and approximately 3.5 years in those who had HCT with growth factor-mobilized blood cells [31]. The current therapeutic approach functions primarily to prevent immune-mediated damage, while awaiting the development of tolerance. Evidence to suggest that current treatments accelerate the development of immunological tolerance is lacking. The mechanisms that facilitate development of tolerance have not been defined.

Administration of medications to prevent infection with Pneumocystis jirovecii and encapsulated bacteria is necessary during treatment for chronic GVHD [32]. Some patients may need topical or systemic treatment to prevent mucocutaneous candida infection. Patients at risk of Varicella zoster activation should be given antiviral prophylaxis, and CMV monitoring and preemptive treatment is necessary in patients at risk of CMV infection [33]. Activation of CMV during the first 3 months after HCT suggests an increased risk of subsequent reactivation in patients with chronic GVHD. Systemic immunosuppressive treatment should be administered at the lowest effective dose in order to minimize the risk of infections and other complications. Many steroid-related complications can be avoided or at least minimized by an alternate-day schedule of administration [34], and topical treatment can be used to minimize the need for systemic treatment [35]. Bone mineral density should be monitored yearly, and losses should be minimized through weight bearing exercise, administration of calcium and vitamin D supplements and hormone replacement.

Indications for secondary treatment include worsening manifestations in a previously affected organ, development of manifestations in a previously unaffected organ, absence of improvement after 1 month of treatment, or inability to decrease the dose of prednisone below 1.0 mg/kg/day within 2 months [30, 36]. Numerous clinical trials have been carried out to evaluate approaches for secondary treatment of chronic GVHD. To date, no consensus has been reached regarding the optimal choice of agents for secondary treatment, and clinical management is generally approached through empirical trial and error [36]. Treatment choices are based on physician experience, ease of use, need for monitoring, risk of toxicity and potential exacerbation of pre-existing co-morbidity.

Progress in the clinical management of chronic GVHD has been slowed by limited insight into the pathogenesis of the disease and the mechanisms that lead to development of immunological tolerance. In the absence of pathophysiologic understanding, physicians must rely on personal or published empirical experience in making decisions regarding treatment. In principle, results of treatment in patients with "steroid-refractory" or "steroid-resistant" chronic GVHD could be used to identify promising agents for initial treatment. Effective agents would be expected to decrease reliance on glucocorticoids and could conceivably decrease the duration of time needed for resolution of the disease.

A variety of retrospective and phase II studies suggested that MMF could be used successfully for secondary treatment of chronic GVHD. In results of a survey published in 2002, nearly 80% of clinicians reported that they had used mycophenolate mofetil (MMF) with great success or at least some success for treatment of chronic GVHD [37]. In another survey proposing a hypothetical scenario describing a case of high-risk chronic GVHD, 54% of the respondents indicated that they would add MMF for initial treatment of chronic GVHD [38]. These results supported a formal test of the hypothesis that the addition of MMF to standard initial treatment could improve outcomes for patients with chronic GVHD.

We therefore conducted a prospective, double-blind, randomized phase III clinical trial to test this hypothesis [27]. The primary endpoint was resolution of reversible manifestations of chronic GVHD within 2 years after enrollment, before death or the onset of recurrent malignancy and without the need for secondary systemic treatment. It was expected that the use of MMF would shorten the time to response, decrease systemic steroid exposure, and decrease the risk of transplant-related mortality without increasing the risk of recurrent malignancy, thereby potentially improving overall survival. Results of the trial, however, did not show any benefits of treatment with MMF. Potential reasons for the negative results were thoroughly explored. The absence of success in this randomized trial could not be attributed to an imbalance of risk factors between the arms, sub-optimal dosing of MMF or non-adherence with administration of the study drug. Hence, this clinical trial definitively demonstrated that addition of MMF to standard initial treatment did not improve outcomes for patients with chronic GVHD.

These unexpected results conflicted with previously prevailing clinical impressions and motivated a careful review of prior reports evaluating the use of MMF for treatment of chronic GVHD. Overall results of 9 such studies suggested that secondary treatment with MMF produced a 20% complete response rate and a 65% complete or partial response rate (Table 1) [39-47]. One of these studies also evaluated results in 10 patients who received MMF as part of the initial treatment regimen for chronic GVHD [45]. Seven of the 10 patients had a complete response, and 2 had a partial response, yielding an overall 90% rate of complete or partial response. In addition, a further study from our center had shown that the proportion of patients who discontinued systemic immunosuppressive treatment after resolution of reversible abnormalities increased progressively from 9% to 17% and 26% at 1, 2 and 3 years, respectively, after starting treatment with MMF [48].

Similar discrepancies have been observed in studies to evaluate the efficacy of thalidomide for treatment of chronic GVHD. Results of 6 retrospective studies and prospective phase II clinical trials suggested favorable outcomes with the use of thalidomide for secondary treatment of chronic GVHD [49-54]. The two randomized prospective studies testing the use of thalidomide for primary treatment of chronic GVHD, however, showed no benefit [24, 25].

Results of the randomized trials defied expectations coming from at least 16 studies evaluating the use of MMF or thalidomide for treatment of chronic GVHD. At least 2 explanations could be proposed to explain this discrepancy. 1) Results of secondary treatment might not predict efficacy as an added agent for primary treatment, perhaps because most patients do not need additional agents in order to gain maximal benefit from initial treatment. Experience at our center, however, has indicated that systemic treatment is changed in approximately 60% of patients during the first 3 years because of inadequate response to primary therapy for chronic GVHD [55]. 2) Alternatively, previous studies might have had unrecognized limitations leading to overstated expectations.

Previous publications have identified quality indicators for evaluation of phase III clinical trials [56, 57], but to our knowledge, similar quality criteria have not been previously proposed for phase II trials and retrospective studies. Therefore, before embarking on a detailed review of the 10 previous studies evaluating the use of MMF, we developed a list of 10 quality indicators that could be used to characterize an ideal prospective phase II clinical trial or retrospective study of treatment for GVHD. The proposed quality indicators are summarized below.

Two individuals (YI and PM) independently reviewed the 10 prior reports of studies testing the use of MMF for secondary treatment of chronic GVHD [39-48]. Reports were evaluated according to whether each quality criterion was met or not, based on careful reading of the text. Differences in scores were reconciled by joint review to arrive at a consensus. Since the purpose of publication is to persuade others, application of the criteria was very strict, and no credit was given if the text did not address the criterion or if the text was not clear. Therefore, in many cases, deficiencies in the report might not have been representative of a study as it was actually conducted.

Results for the 10 studies of MMF are summarized in Table 2. Scores at the bottom of the table represent the total number of criteria met by each report. One report failed to meet any of the 10 criteria. Two reports met 4 criteria, and none had higher scores. The mean score for the 10 reported studies was 2.0. Scores at the right margin of the table represent the number of reports that met each criterion. None of the reports attempted to demonstrate that bias had been minimized in the selection of patients, used an historical or contemporaneous benchmark or tested a statistical hypothesis. Only one report had a specified time of assessment, and only two had objective response criteria and well-defined overall response criteria. Three reports employed a consistent treatment regimen, while 7 accounted for possible effects of concomitant treatment.

Results of the review of reports for studies testing MMF prompted a more comprehensive review of studies testing systemic agents for secondary treatment of chronic GVHD published between 1990 and 2011. We searched the Medline (PubMed) database using a broad search strategy to identify studies evaluating secondary treatment of chronic GVHD. The search was conducted using the terms "Chronic graft versus host disease" and "Treatment" excluding "Review." Relevant references in the publications identified were also reviewed. Both retrospective and prospective studies were included, but studies with cohorts containing fewer than 10 patients (N=26), phase III studies and case reports were excluded. A total of 60 studies were selected for review [39-54, 58-101]. Initial agreement between the two reviewers was high, ranging between 72% and 98% (Table 3).

Across the 60 studies, 17 different agents were evaluated (Fig. 1). Extracorporeal photopheresis was the most frequently studied agent (N=17) followed by mycophenolate mofetil (N=10), thalidomide (N=6), sirolimus or everolimus (N=4) and rituximab (N=4). The distribution of scores representing the total number of criteria met by each report ranged from a low of 0 (N=6) to 8 (N=1) [61] (Fig. 2). The mean score for all 60 reports was 2.5. The mean score for prospective studies (N=31) was 3.1, compared to 1.8 for retrospective studies (N=29). The mean score for multicenter studies (N=7) was 3.6, compared to 2.3 for single-center studies (N=53).

Approximately 35% to 45% of all reports provided adequate information regarding eligibility criteria, organ response criteria, overall response criteria, concomitant treatment and overall survival (Table 4). Only 22% of the reports had a specified time for assessment of response, and less than 10% of the reports documented an absence of bias in the selection of patients, used a consistent treatment regimen, or tested a formal statistical hypothesis on the basis of a benchmark from a contemporaneous or historical cohort. The percentage of reports fulfilling quality indicators was generally higher for prospective studies than for retrospective studies (Table 4).

Despite their many shortcomings, all 10 reports evaluating MMF offered favorable overall assessments, 8 in the abstract, and 2 in the discussion. All 10 reports called for additional studies, 3 in the abstract, and 7 in the discussion. The contrast with results of the prospective phase III trial testing MMF for initial treatment of chronic GVHD raises a general concern that other previously tested agents also do not provide as much benefit as suggested in the reports. The approach used in most reports relies on the assumption that any improvement after new treatment must have resulted from the new treatment, but most of the studies did not attempt to assess the durability of response. Taken as a whole, the collection of reports does not facilitate comparisons of efficacy from one agent to the next, and readers are left to conclude that everything works, more or less.

Investigators prefer new treatment to be effective, and under the "publish or perish" pressures of academic life, authors may lose objectivity and attempt to portray results as positively as possible. None of the 60 reviewed results indicated negative overall results, strongly suggesting a powerful bias by authors and journals to publish only the results of "positive" studies. Conclusions from retrospective studies and phase II clinical trials should be stated more cautiously. For example, we suggest that an appropriate conclusion from the studies of MMF would be the following: "Our results demonstrate the feasibility of using MMF to treat chronic GVHD. The true merits of using MMF for this indication can be evaluated only in a prospective controlled trial". Small retrospective studies have very limited value for assessing results of a new treatment, and the distinction between retrospective studies and prospective studies is important. Nonetheless, many prospective phase II studies still fall far short of the ideal.

Progress would be enhanced if studies could be conducted in a way that allows results to be compared from one study to the next in a more informative way. Aggregation of results for secondary therapy with those for third, fourth and subsequent lines of treatment makes such comparisons impossible, due to large variation in prior treatment and concomitant therapy. Comparisons are also impeded by an inability to estimate the baseline prognosis of patients enrolled in any given study as compared to those enrolled in other studies.

The current state of affairs has many harmful effects. Most reviews that summarize previous literature regarding treatment of chronic GVHD focus on overall complete and partial responses, leading readers to uncritical acceptance of conclusions that agents are effective, when in fact, they are not. Agents that are accepted as effective could actually cause unrecognized harm, as suggested by results of the phase III MMF study [27]. Clinicians who believe that they already know what is best have little incentive to participate in clinical trials. As a consequence, progress has stalled, and no one is able to identify new treatments that are truly effective. Progress would be enhanced if investigators could identify truly promising results from phase II trials and move forward more quickly to testing in definitive phase III trials.

Progress would be greatly enhanced by standardization in 4 areas: eligibility criteria, organ response criteria, overall response criteria, and time of assessment. Eligibility criteria should focus on true secondary systemic therapy, rather than allowing enrollment at any point beyond primary therapy. This strategy will provide a more homogeneous population and more interpretable results than one that allows enrollment at any point in the development of the disease. Criteria defining "steroid-resistant" and "steroid-refractory" chronic GVHD should be standardized. Lack of adequate improvement after at least one month of treatment with prednisone at 1 mg/kg per day represents one possible definition, although some chronic GVHD manifestations would not be expected to improve within a month after starting treatment.

Standardized response criteria have been proposed but have not been used because of their complexity and lack of validation [7]. Death, recurrent malignancy or a further change in systemic treatment other than the dose of prednisone before the assessment point should not qualify as a response. Measures of response will have to be simplified and validated in order to gain wider acceptance. For example, the NIH scales for mild (score 1), moderate (score 2) and severe (score 3) chronic GVHD could be used to standardize organ response criteria. The same scale could be used to standardize overall response criteria, although the appropriate classification for cases with improvement in one organ with progression in another would still pose difficulty. Alternatively, changes measured according to the NIH scale for global severity of chronic GVHD could be used to standardize overall response, since several studies have shown that the NIH global severity at initial diagnosis correlates with survival [102, 103], although changes in global severity have not yet been correlated with survival. Further evidence from retrospective or prospective studies will be needed to reach consensus on standards for assessment of treatment response in patients with chronic GVHD.

Many studies have used short-term response to assess new therapies for chronic GVHD, but at least one prior study showed that response at 3 or 6 months does not predict resolution of the disease [104]. On the other hand, results of this study showed that several definitions of response could be used together with the additional criterion of a prednisone dose ≤0.25 mg/kg/day to predict the risk of subsequent failure, defined as death, onset of bronchiolitis obliterans, or introduction of a new systemic treatment because of new or progressive manifestations of chronic GVHD. Hence, patients with response by the proposed composite definitions had lower risks of subsequent failure, as compared to those who did not have responses by the same definitions.

Progress has been hampered by the absence of any established benchmark of success that could be used as a comparison point for studies of new treatment. We have previously reviewed outcomes after secondary systemic treatment for chronic GVHD at our center [29]. Approximately 50% of patients died or had a qualitative change in systemic therapy during the first year, and an additional 10% had recurrent malignancy (Fig. 3). The proportion of patients who were alive without a subsequent change in systemic therapy and without recurrent malignancy was approximately 60% at 6 months and 40% at 1 year. The proportion of patients with complete or partial response at 1 year was not determined, but it cannot exceed 40%. These results might not be representative of current outcomes, since historical criteria were used to define chronic GVHD.

Editors of the Journal of Clinical Oncology have recognized the importance of improving the conduct and reporting of phase II trials testing treatments for cancer [105]. As a guideline for authors, the editors have summarized criteria for the types of phase II studies that would be most appropriate for consideration by the journal. The editors also expressed the hope that their views might assist other journals that may be struggling to prioritize the most important types of phase II trials for publication.

The editorial position at the Journal of Clinical Oncology is that phase II studies will be considered only if they include 1) a clear definition of the primary end point, 2) a hypothesized value of the primary end point that justified the planned sample size, and 3) a discussion of possible weaknesses, such as any comparison to historical controls [106]. Only one of the 31 prospective studies of secondary treatment for chronic GVHD had a formal estimation of sample size with a well-justified historical benchmark. The Journal of Clinical Oncology also requires on-line publication of a redacted version of the study protocol, thereby enabling reviewers and readers to recognize any differences between the reported results and the study as originally planned. Improved treatment for chronic GVHD is more likely to emerge if reviewers and journal editors hold authors to higher standards in evaluating manuscripts for publication.

Standardization of methods for clinical trials would enable comparison of results in different trials and thereby accelerate progress in evaluation of new treatments for chronic GVHD. An urgent priority is the development of a benchmark of success based on results of unbiased retrospective reviews or prospective studies that include all patients who received secondary systemic therapy for chronic GVHD. Robust phase II studies could then be carried out to evaluate whether new therapies offer any genuine improvement compared to the benchmark. For example, a clinical trial could be designed to test whether a new treatment improves outcomes compared to the 40% historical response rate at 1 year, as described above. If the true response rate with the new treatment were 60%, enrollment of 42 patients would offer 80% statistical power with a 0.05 one-side type I error, and successful outcomes in at least 22 patients would encourage further studies. Alternatively, randomized trials with a "pick-the-winner" design could be used to identify approaches that truly warrant further evaluation.

Promising candidate treatments identified in robust phase II studies could be taken forward in phase III studies of secondary treatment, and successful results in such a study would establish a new benchmark for future phase II studies of secondary therapy. Promising candidate treatments could also be tested in phase II studies of primary treatment. Most importantly, successful results in phase III studies of either secondary or primary treatment would improve patient outcomes and establish new standards of care.