Prostate Cancer Clinical Trial End Points: “RECIST”ing a Step Backwards

Materials and Methods

To examine the manifestations of progressing metastatic prostate cancers that are detectable on an imaging study, the sites and size of lesions from patients enrolled and treated at MSKCC on sequential Institutional Review Board–approved trials for patients in the states of clinical metastases noncastrate and clinical metastases castrate disease were analyzed for eligibility under RECIST (10). Enrollment on these trials required progression of disease defined as any one of the following: a minimum of three increasing PSA levels obtained ≥2 weeks apart where the range of values is at least 25%, new or progressive (25% bidimensional increase) soft tissue masses on computerized tomography or magnetic resonance imaging, or new lesions on radionuclide bone scan. All imaging studies were reviewed in a blinded fashion by a single radiologist (L.H. Schwartz). The metastatic sites for each patient on the protocol were categorized as a “target” or “nontarget” lesion using the RECIST criteria. Sites considered were recurrent/persistent disease in the prostate or prostate bed, lymph nodes, viscera (e.g., liver and lung), and bone.

All of the manifestations were then related to the end points developed to assess objective tumor regression as defined by RECIST. Separately, the definitions for complete response, partial response, and stable and progressive disease were applied to the specific sites of prostate cancer spread. Overall response assessments defined by RECIST were also analyzed and related to the therapies currently approved for the systemic treatment of prostate cancer. A similar analysis was conducted for the manifestations of 31 patients with noncastrate metastatic disease treated on a study of rapid hormonal cycling and chemotherapy (11), and separately for patients in the states of localized disease and a rising PSA (noncastrate and castrate). An alternative set of phase II end points are proposed that are applicable to patients who occupy a given clinical state.

Results

Eligibility criteria

Target lesions. RECIST categorizes sites of tumor spread as either measurable (target lesions) or nonmeasurable (nontarget lesions), with the recommendation that up to 10 target lesions per patient be recorded and monitored. The RECIST criteria further recommend that only patients with target lesions be registered to clinical trials. We examined the distribution of disease in contemporary group of patients representing the different prostate cancer clinical states to determine what proportion had target lesions that would satisfy the eligibility criteria for enrollment on clinical trials under RECIST. Per RECIST, metastases to lymph nodes, liver, or lung may qualify as target lesions if they have a minimum size that is more than twice the slice thickness of the scanning instrument used (10).

Table 1 shows the proportion of patients with castrate metastatic disease with measurable visceral and/or nodal spread (target lesions), and osseous disease (nontarget lesions) from 124 men enrolled on three sequential Institutional Review Board–approved protocols (12–14). As noted, 54 (44%) had disease limited to bone, 12 (10%) disease limited to soft tissue, whereas 58 (47%) had disease in bone and soft tissue. Overall, 63 patients (51%) had lesions that could be measured in soft tissue and 7 (6%) had lesions that were considered too small to measure. In the 63 patients with measurable tumor, 54 patients (44%) had a total of 104 lesions that were ≥2 cm in size. The median number of lesions ≥2 cm per patient was 2 (range, 1-6). The sites of spread were lymph nodes in 76% (48% retroperitoneal, 36% pelvic, 7% mediastinal, 5% abdominal, 2% axillary, and 2% unspecified), liver in 13%, lung in 3%, recurrent in the prostate bed in 2%, and adrenal spread in 1% of cases. Nine (7%) additional patients had lesions in the range of 1.0 to 1.9 cm, which would qualify for enrollment using spiral computed tomography. The median size of all lesions was 2.5 cm (range, 1-8 cm); the median size is 3 cm (range, 2-8 cm) if only lesions ≥2 cm in size were considered. The distribution of lesions seen in the MSKCC series compared with those reported in recently completed multicenter trials is shown in Table 2 (15–17).

Table 3 shows a comparable analysis of 31 patients with clinical metastasis noncastrate disease. As shown, 13% had disease limited to bone, 16% to soft tissue, whereas 71% had both osseous and soft tissue disease. In this cohort, 12 (39%) patients had 15 target lesions of which 7 (47%) were ≥2 cm in size. Lymph node disease was again predominant. Five had soft tumor disease and no osseous spread.

Bone. Bone lesions are “truly nonmeasurable” and are recorded as nontarget lesions. Tumors that have eroded through bone into the epidural space or base of skull may in some cases be measurable. However, for most phase II trials, neurologic compromise is an automatic exclusion because these patients require immediate relief of symptoms using radiation therapy or surgery. As shown in Table 1, 54 (44%) of patients in the castrate data set and four patients (13%) in the noncastrate data set had disease limited to bone. These patients, by RECIST criteria, would be ineligible for trial entry.

Localized or locally recurrent disease. Disease in this location occurs in the newly diagnosed patient with clinically localized or metastatic disease, and those who have recurred after surgery or radiation therapy with or without metastases. In the MSKCC data set of patients with castrate metastatic disease, only one patient had a target lesion recorded in the prostate bed.

Tumor markers. The median PSA values for patients with castrate metastatic disease was 91 ng/mL (range, 0.5-2,282), for noncastrate metastatic disease was 19.8 ng/mL (range, 0.6-75), and in a series of patients with an increasing PSA was 3.2 ng/mL (range, 0.08-275; ref. 11). By RECIST criteria, marker values are recorded as “normal” or “abnormal” and increasing values are not required for entry (10). However, what constitutes an abnormal value varies by state, as reflected in the data above. For example, in a patient with localized disease who has been treated with external beam radiation therapy, a PSA of 1.5 ng/mL may signify the nonmalignant prostate tissue that also produces PSA after the malignant component has been eradicated, and not persistent or recurrent tumor. In contrast, any PSA above 0.2 or 0.4 ng/mL in a patient who has undergone radical prostatectomy indicates recurrent disease.

Most advanced-disease therapeutic trials require increasing values on entry (18, 19). Where trials may differ is in the absolute PSA value and/or the defined change over time (e.g., doubling time) required for entry. These aspects of PSA are not specified in RECIST, and increasing values are not required for trial entry.

Symptoms. The presence or absence of symptoms is not considered an eligibility criterion under RECIST criteria.

Eligibility criteria by Response Evaluation Criteria in Solid Tumors—a state-by-state view.Table 4 summarizes the disease manifestations across the disease states in relation with eligibility under RECIST. As shown, all patients with localized disease and an increasing PSA are excluded because there are no target lesions in these states. Even in patients with detectable metastatic lesions on imaging, whether castrate or noncastrate, the majority of patients do not have target lesions that would qualify for enrollment under RECIST.

Outcome measures (response)

Table 5 lists sites of disease and clinical state and relates them to the definitions of complete response, partial response, and stable and progressive disease. These are considered below.

Target lesions. The majority of soft tissue target lesions in patients with clinical metastases enrolled on trials are lymph nodes. A partial response classification by RECIST requires a ≥30% decrease, and progression of disease a ≥20% increase in the sum of the largest diameters of the individual target lesions. Using the target lesions shown in Table 1, this change represents a decrease from 3 to 2.1 cm, or increase from 3 to 3.6 cm in the greatest diameter. Lesions of this size are subject to greater measurement error and subsequent misclassification than larger tumors even when they are read by a single radiologist (20). In addition, lymph nodes do not disappear even if the tumor is eradicated and a response may represent a shrinkage of minimally enlarged node to normal size, e.g., 2.2 to 1.8 cm (21).

A separate consideration is whether changes in the size of a lymph node actually parallel changes in the tumor in bone. Given the insensitivity of standard radiographic techniques to assess disease in bone, radiographic changes in lymph nodes may not be paralleled by bone scintigraphy (22). Similarly, whether distinct trials should be designed for the 15% of patients with disease limited to soft tissue or disease limited to osseous sites is unknown.

Nontarget lesions. The defined response categories are not applicable to patients with disease limited to bone, which are nontarget lesions. Further, there are no accepted standard criteria to describe a favorable outcome for treated osseous metastases by bone scan.

Similarly, the definitions of progression are also not standardized, although most investigators accept the appearance of one or more new lesions and/or unequivocal progression of existing nontarget lesions. Consequently, changes in bone scan are best characterized as “improved,” “stable,” or “progressing.” The issue of “pseudoprogression,” the appearance of new lesions or an increase in intensity of existing lesions with successful therapy, is not addressed. In these cases, a patient who is actually benefiting from therapy may be recorded as having progression of disease (23).

Prostate or prostate bed (newly diagnosed or recurrent). Although the size of the prostate can be estimated using ultrasound or magnetic resonance imaging, in most cases only a proportion of the gland is involved by tumor and no dominant lesion is visible. By ultrasound, there are no reproducible findings that can distinguish tumor- from nontumor-bearing regions reliably, although discrete lesions can be seen in some instances. Magnetic resonance imaging with spectroscopy has the potential to improve detection and monitoring (24), but is still considered investigational at this time. Recurrent disease in the prostatic bed is rarely palpable on physical examination or visible on an imaging study. Whereas magnetic resonance imaging may show abnormalities that are not detected using other modalities, it is not used routinely and is rarely “measurable.” (25) ProstaScint scanning is a Food and Drug Administration–approved imaging modality that uses a radiolabeled antibody to an internal epitope of prostate-specific membrane antigen to detect sites of disease (26, 27). However, the images obtained do not qualify as target lesions under RECIST and cannot be used on a serial basis to assess changes in tumor growth. In clinical practice, the presence or absence of disease in the prostatic bed following radical surgery is “predicted” on the basis of factors such as the pathologic stage, margin status, the time from primary treatment to the documentation of recurrence, PSA doubling time, and current PSA level (28–31). Even the performance of a biopsy to confirm or exclude disease in the prostate bed is not reliable enough to be recommended on a routine basis, and even if it were positive, would not qualify as a target lesion under RECIST. Consequently, the prostate is a “nonmeasurable,” nontarget site whether the patient is newly diagnosed or has persistent or recurrent disease in the gland.

As the tumor within the prostate itself cannot be accurately measured, the terms complete response, partial response, progressive disease, or stable disease are not applicable to patients with localized disease in which the sole site of disease is the newly diagnosed primary tumor itself, or a tumor that has persisted in the prostate after radiation therapy or prostate bed after radical prostatectomy. The exception is the rare case (3% in the current series) in which it has met the criteria for a measurable target lesion at entry.

Tumor markers. Tumor markers alone are not used to assess outcome by RECIST, and patients with an increasing PSA as the sole manifestation of disease are not eligible for trials. Thus, for patients enrolled on trials on the basis of progressing target lesions and an increasing PSA to be classified in a complete response category requires a disappearance of the target lesion and normalization of the PSA. However, what is considered normal will vary depending on whether the patient has received treatment to the prostate, and if so, of what type. Take, for example, three patients with no osseous disease who have with multiple target lesions that have regressed completely. The first had progressed after radical prostatectomy, the second after radiation therapy, and the third never received treatment to the primary tumor. For the first patient, a normal PSA is one that is undetectable; for the second, <0.5 or 1.0 ng/mL; for the third, 2.5 or perhaps as high as 4.0 ng/mL.

Symptoms. Symptoms that are present are recorded and monitored, but are not used to assess response. Worsening of preexisting symptoms or the development of new ones can, at the investigator's discretion, be used to show disease progression once a patient is on a study, although it is often difficult to distinguish progressive disease from drug-related adverse events in this patient group.

Response criteria by Response Evaluation Criteria in Solid Tumors—a state-by-state view of best overall response. RECIST require the assignment of a best overall response of either complete response, partial response, or stable or progressive disease on the basis of defined changes in target lesions and some nontarget lesions. As shown in Table 5, the difficulties of categorizing changes in the individual sites of disease into a complete response, partial response, or stable disease category renders the best overall response classifications of limited value. Bone disease is particularly problematic because most techniques to assess changes in bone metastases are too insensitive to detect change, be it favorable or unfavorable, on a serial basis. In cases where plain radiographs are abnormal, resolution of the osteoblastic change rarely occurs and improvements in a radionuclide bone scan may not occur for many months even if all disease in bone were eliminated by the treatment. The criteria do require confirming a “favorable” change in an imaging study within 4 weeks of its original documentation, but this interval is too short to show a meaningful change on a bone scan.

A second difficulty is the failure to incorporate contemporary uses of serial changes in the PSA to assess treatment effects. Associations with posttreatment PSA changes and survival have been reported for patients with an increasing PSA after local therapy (32, 33), as well as for those castrate metastatic disease in retrospective analyses (34–38). Equally important, RECIST does not address the significance of increasing values as an indication of disease progression. As a result, a patient with an increasing PSA could be considered as responding favorably if his scans were without evidence of progressive disease.

Discussion

The difficulties associated with drug development in prostate cancer have long been recognized. To address this, consensus groups of experts have established guidelines to conduct clinical trials addressing questions for patients in the states of clinical metastases castrate disease (18), and more recently for the state of an increasing PSA (19). As illustrated by these efforts, and a contemporary review of trials for patients with detectable metastases on imaging, fewer than half had disease manifestations that would be recognized as target lesions under RECIST. Even when documented, lesions were generally few in number and small in size, and >80% of the sites were in lymph nodes. The most common manifestations of metastatic prostate cancer, osseous metastases with an increasing PSA, are not adequately addressed. For these and for other prostate cancer clinical states, eligibility and outcome measures designed to assess tumor regression simply do not apply.

Prolongation of life is ultimately the primary objective of therapy directed toward prostate cancers representing any clinical state. It can be achieved by eliminating the disease surgically, with radiation or by medical means, or, in theory by slowing the disease to the point where the patient dies of other, nonprostate cancer–related causes. There are, however, other treatment objectives of importance to patients in addition to prolongation of life, illustrated by the therapies approved for use for patients with castration resistant disease. Among them are mitoxantrone and prednisone, strontium-89 (Metastron), and samarium-153 (Quadramet), all of which were approved to palliate pain (39–43). Zometa was approved on the basis of reducing the risk of skeletal complications (40), whereas mitoxantrone and prednisone were shown to reduce the constitutional symptoms of an increasing disease burden (39). Noteworthy is that none of these therapies were approved on the basis of trials using an end point of tumor regression, and none were approved on the basis of an effect on PSA. In fact, most had no effect on PSA, and the clinical benefits were shown without prolongation of life. Only docetaxel, recently approved by the Food and Drug Administration for the treatment of patients with castration-resistant disease, was shown to prolong life in a randomized prospective clinical trial (15).

These examples show the importance of tailoring drug development to what is routinely done in clinical practice: focusing the eligibility and outcomes on the clinical reason that a therapy is being offered and what it is expected to achieve. The outcome measures must account for the putative action of the drug, whether it is cytotoxic or cytostatic, potentially curative or palliative, and whether it affects the cancer cell directly or the interaction between the tumor and host. The short-term measures may be effects on PSA, on symptoms, or, depending on the context, more traditional measures of disease regression. Assuming that these milestones are achieved, the more critical issue becomes the durability of these effects with a focus on clinically meaningful measures of disease progression.

By separating the short-term indices of a favorable outcome that justify continuing treatment and the longer-term measures of the duration of benefit to show when a drug is no longer working, trials are not limited to studying drugs solely on the basis of their effects on PSA, tumor regression, or other parameters of uncertain clinical significance. Disease progression is indicated by any clinically significant parameter that reflects a disease manifestation that is worsening, including an objective in the size of a target lesion, the failure to palliate or to control of symptoms, the development of new symptoms for patients at high risk, the increase in the size of or development of new lesions on a bone scan, or an unfavorable change in PSA. The crucial issue is that the definition of disease progression be tailored specifically to the reason that an intervention is being offered and the clinical state or clinical scenario of the patient group in which it is being evaluated, as shown in Table 6.

To apply this approach clinically, all manifestations and potential sites of disease must be monitored at fixed time intervals. Failure at any site would be indicative of disease progression. The overall effects of an intervention are then assessed by examining either the proportion of patients who have not progressed at a fixed time point, the proportion who have not progressed to the next state, or an overall reduction in the hazard ratio of progression. The principle can be applied to single-arm phase II trials and phase II randomized trials. The overall risk/benefit for the intervention could then be estimated by relating the adverse events noted on the study to the duration of benefit that was observed. These will be considered on a state-by-state basis.

Localized disease. The objectives of treatment for a patient with localized disease are to eliminate the cancer completely and to prevent its recurrence in the primary site and systemically. The probability of eliminating the cancer can be estimated using risk models or nomograms (44, 45). For the patient who has undergone a prostatectomy, the first indication of “success” is whether or not an undetectable PSA is achieved postoperatively. If it is, recurrence or progression is signified when the PSA becomes detectable, e.g., >0.2 or 0.4 ng mL, and the abnormality is confirmed and continues to increase. For radiation therapy–treated patients, a different definition of PSA failure is used (46). The longer-term objective is to prevent an increasing PSA (or PSA recurrence), which antedates radiographic and clinical progression in virtually all cases. Neither definition of biochemical relapse addresses whether the failure is local or systemic, and if the latter, whether the patient is likely to develop metastasis or die of disease; a more relevant intermediate end point might be the proportion of patients who fail with a rapid PSA-DT, which has been shown to associated with a higher risk of prostate cancer metastases and death (33). This requires prospective validation.

Rising prostate-specific antigen: noncastrate. For patients with an increasing PSA after surgery or radiation, the first question is whether the increase in PSA represents recurrent/persistent disease in the primary site, systemic disease, or both. Locally persistent/recurrent disease in the prostate or prostate bed with no metastasis may still be curable or controllable with additional therapy to the prostate or prostate bed. As is the case for the newly diagnosed patient with localized disease, the first indication of success of a salvage prostatectomy or salvage radiation therapy is whether the PSA becomes undetectable.

For those in whom the increasing PSA represents systemic disease, the primary clinical objective is to prevent a transition to a state of clinical metastases, a point in the disease where the risk of a prostate cancer–specific death is significantly increased (47). The short-term objective may be achieved by eliminating the disease, slowing tumor growth, or inhibiting the metastatic process without directly affecting tumor growth. Here again, the mechanism of the agent must be considered. Some agents, such as those that promote differentiation, may induce short-term PSA increases that do not signify treatment failures. Others, such as immunologic agents or antiangiogenic drugs, may alter the rate of increase but not induce a PSA decline. The effect on PSA may also be delayed. As it is unlikely that patients with increasing PSA values will remain on therapy or be observed until an end point of radiographic progression is reached, it has been suggested that only agents with activity in more advanced clinical states should be tested here (19). Ultimately, the clinically significant end point is metastasis-free survival or overall survival, which are a better gauge of whether the drug is changing the natural history of the disease and not simply modulating PSA levels independent of an effect on tumor growth (19).

Rising prostate-specific antigen: castrate. Based on maturing data showing the prognostic significance of PSA doubling times and metastatic progression and prostate cancer–specific mortality (31, 32, 47), an increasing number of patients are receiving androgen deprivation before overt metastases are documented on an imaging study. It follows that patients are treated with androgen deprivation when an increasing PSA is the sole manifestation of disease progress in a similar manner—with an increasing PSA despite castrate levels of testosterone. The result is an entirely new disease state—increasing PSA: castrate—whose natural history is not well described. The overall treatment objectives are similar to those described for patients with increasing PSA: noncastrate, recognizing that the anticipated survival times following the development of metastases in the setting of castrate levels of testosterone carries a more ominous prognosis (48).

Clinical metastases: noncastrate. Elimination of all disease is rarely possible for patients with clinically detectable metastases. Instead, shorter-term therapeutic objectives are focused on strategies to increase the degree and the duration of response to androgen deprivation, whereas longer-term objectives include preventing the development of resistance, delaying the progression of established lesions or the development of new ones, and preventing new symptoms of disease (49). Given the high response proportions observed with androgen deprivation alone, these trials tend to be large and carry considerable risk of not demonstrating a significant difference in outcome for the experimental treatment. Such has been the case in trials studying the question of whether combined androgen blockade in superior to testosterone lowering monotherapies (50).

Noteworthy, however, is that PSA-based diagnosis, treatment, and monitoring has resulted in a decrease in the proportion of men that are being diagnosed with detectable metastases. The majority of men with detectable metastases do not have disease in the primary site and lower overall tumor burdens than in the past. In these cases as well, the probability of achieving an undetectable PSA varies as a function of disease extent (11). It remains an open question whether a proportion of these patients may be curable as well.

Clinical metastases: castrate. Patients who have progressed on hormones face a high likelihood of experiencing cancer-related morbidity and dying of their cancers. The short-term objective of treatment in these cases includes the control of disease-related symptoms. The longer-term objective is to prevent their occurrence or recurrence and to prolong life. With the recent approval of docetaxel for castrate metastatic disease on the basis of a survival benefit (15, 38), drug development efforts are now focused in two contexts: patients who have not received taxane-based chemotherapy (first-line therapy) and those who have (second-line therapy). Regardless of whether the drug is being explored as first-line or second-line therapy, an initial treatment effect could be assessed by using the proportion of patients who achieve a predetermined change in PSA, the proportion who show a significant reduction in soft tissue disease by standard criteria, or those who do not progress by bone scintigraphy (3, 51, 52). For those with symptoms, the proportion of patients who enjoy a palliative benefit is of interest, but equally important is the durability of these benefits. Accordingly, many trials consider a change in oncolytic therapy, such as the administration of radiation therapy to new symptoms of disease or a change in chemotherapy, as an indication that the treatment under study was no longer effective. This approach addresses the clinical reality that therapy is rarely stopped if the PSA is declining or stable even if a 50% benchmark is not reached. It is also and end point that can be validated prospectively in randomized comparisons of a first-line and second-line cytotoxic approach.

As a screening method for active treatments, many trials report the proportion of patients who achieve a defined posttherapy change (e.g., the proportion showing a 50% or 80% decline from baseline) as suggested by the Prostate Specific Antigen Working Group (18), while monitoring all sites of disease to ensure concordance (6). Whether a posttherapy decline in PSA can serve as a “surrogate” for clinical benefit, however, remains controversial (53–55). Recent work by our group has shown an association between a “stable” bone scan at 6 months after the start of therapy and survival (56). This observation, coupled with our observation of the prognostic significance of a “no increase” versus “increase” in PSA at 12 weeks (54), has led us to propose a composite end point that requires that the levels of PSA do not increase and that the bone scan does not progress for a minimum of 6 months as a potential intermediate end point for more definitive trials. The 6-month landmark represents the median time to progression observed in the docetaxel/estramustine arm of the Southwest Oncology Group study 99-16 (38), in which a survival benefit was shown relative to the control treatment. It also represents the median time to a change in chemotherapy for patients enrolled on first-line chemotherapy trials were subsequently enrolled on a second-line chemotherapy trial (Beekman), representing a change in disease status that was significant enough to require a change in systemic chemotherapy.

Composite end points, however, are not without pitfalls. An example is provided by a recent comparative trial of the endothelin receptor-A antagonist atrasentan, in which a protocol definition of progression mandated removal of patients from the study. The definition, the development of two or more lesions on a bone scan, did not address the lag that may occur between the elevations in PSA required for entry on study and changes in radionuclide bone scan. As a result, over half of the patients came off trial at 3 months in spite of stable or improving symptoms and PSA values. In fact, the continued analysis of the trial has shown that all the other end points, save the bone scan, favored the atrasentan arm (57).

Critical to the use of composite end points is tailoring each element to the class of drug being tested. For example, an increasing PSA may not signify a treatment failure for certain biological agents, such as those that induce differentiation or those that slow progression without cytotoxic effects. By the same token, patients should not be taken off study for a worsened bone scan 8 to 12 weeks following initiation of cytotoxic therapy in the context of a declining PSA or improving symptoms, as the scintigraphic findings may also represent flare phenomena. Similarly, a pain flare following treatment with certain hormonal agonist/antagonists or bone-seeking radiopharmaceuticals may indicate response rather than treatment failure. Ultimately, each element of the composite—be it PSA, scintigraphy, radiography, biological assay, or symptomatology—must be independently defined using the putative mechanism of the drug and the preclinical data as part of the trial design process.

Summary

The purpose of the phase II trial is to determine whether a treatment produces a sufficient effect in a sufficient proportion of patients for a sufficient period of time to justify further development. How activity is defined should be determined by the treatment objective for the patients to whom it is being offered. The manifestations of prostate cancer do not lend themselves to the “packaging” assessments of treatment effects into the response categories proposed by RECIST. More important than defining “response” is to define objective, meaningful, measures of disease progression at which point it can be stated more precisely that the therapy under evaluation is no longer effective. This methodology can be applied across the clinical spectrum of the disease, is independent of the specific intervention used, and circumvents the controversies of assessing response using changes in PSA and bone, the most common manifestations of the disease.