July 25, 1994
A discussion regarding KS staging and its use in assessing treatment response must occur before analyzing results from the various studies that have taken place for KS over the past 10 years.
Before 1989, there was no widely used and validated staging classification for Kaposi's sarcoma. It was not until July, 1988, that members of the AIDS Clinical Trials Group (ACTG) Oncology Committee came up with a comprehensive proposal for uniform evaluation, staging and response criteria for AIDS-related KS (Krown 1989). This is now standard for all ACTG KS trials. It has been suggested that the ACTG criteria "be used in all clinical trials in HIV-related KS if we are to build a consensus" (Volberding 1989). These staging and response criteria are not universally employed in non-ACTG KS trials.
Careful study of trials that pre-date 1988 indicate that at least three different staging classifications were being used (Krigel 1983; Mitsuyasu 1984; Vadhan-Raj, 1986). The very first, and now obsolete, four-stage KS classification proposed by Krigel and colleagues (Krigel 1983), took into account tumor extent (localized vs. general) and observed rate of progression. This classification was intended to stage patients with AIDS-KS along with those with classic and African KS. Most patients with AIDS-KS were, however, assigned to stages III or IV. Moreover, it did not take into account variables of immune status which appear to be important in AIDS-KS treatment response.
Mitsuyasu and Groopman's (Mitsuyasu 1984) four-stage system, specifically designed for AIDS-KS, was the first to stratify patients into two subtypes, A or B. Type (A) indicated the absence, and type (B) the presence, of "B" constitutional symptoms (unexplained fever, night sweats, more than 10% involuntary weight loss, or diarrhea persisting for more than two weeks); or opportunistic infections (OIs). Using this classification, Mitsuyasu not only noted a survival advantage in patients in tumor stages I and III, but also saw a marked survival difference when comparing KS patients with subtype (A) vs. subtype (B) (Mitsuyasu 1986).
The ACTG system went further to classify patients based on the clinical extent/placement of tumor (T), immune status/CD4 cell count (I) and assessment of HIV-related systemic illness (e.g., OIs/"B" symptoms/Karnofsky status) (S). Based on the composite score of the three categories, (T), (I) and (S), patients are scored as good risk (0) or poor risk (1). Thus, a patient with gastrointestinal KS, a CD4 count of 100, and no prior OIs or "B" symptoms would be staged T1, I1, S0.
Although patients are staged at entry into ACTG KS trials, they are not followed prospectively to determine survival. Five years after the implementation of this ACTG staging system, the ACTG Oncology Committee is now attempting a retrospective analysis of approximately 227 patients from past KS trials to determine the validity of their staging system. This will be an arduous task for the ACTG Oncology Committee because there is scant long-term follow up or survival data on most KS trial patients. This classification must be looked at prospectively to determine if the staging system is effective in determining treatment outcome and survival (Steven Miles, personal communication).
A group of Italian researchers recently completed a retrospective analysis of KS patients (reclassifying and restaging patients) using a staging system similar to the ACTG's T-I-S classification. They noted that (I) immune status/CD4 count, and (S) assessment of HIV-related systemic illness are accurate prognostic indicators of survival, but the extent of tumor (T) at baseline is not a major prognostic factor (Lawrence Kaplan, personal communication).
Krown and colleagues contended that scoring of responses in earlier clinical trials was "somewhat vague." (Krown 1989). Thus, attempting to make true comparisons of the efficacy of treatments based on clinical data before 1989 might be misleading.
While a complete response (CR) has generally been regarded as the disappearance of all lesions, past studies have not always specified their method for documenting a CR. For example, tumor biopsies were only conducted "in some patients" in the 1984 recombinant alpha-2 interferon trial conducted by Groopman and colleagues (Groopman 1984). A partial response (PR), which has generally been defined as 50% or more reduction in the number or size of lesions, was often ambiguously worded in clinical trials from the early and mid 1980s. Volberding and colleagues (Volberding 1985), in their 1985 Vinblastine trial, never specified the criteria used for assessing the "significant" changes of lesion nodularity and pigmentation that would actually warrant a PR.
The ACTG's four separate response definitions for CR, PR, stable disease and progression were intended to ameliorate such ambiguity and vagueness. The "vagueness" is still present in some studies because: 1) making up a definition for response might make mediocre data look better2; 2) people do not want to invest the time to actually use the ACTG response criteria or attempt to come up with something better. This vagueness was embarrassingly evident at the FDA's Oncologic Drugs Advisory Committee hearing on Vestar's NDA for Daunoxome in 1993. Vestar claimed that its research showed a 51% PR with Daunoxome for patients who were refractory or intolerant to standard chemotherapy regimens. The FDA, in its separate analysis of the data, lowered the PR rate to 18%. Humiliated by this separate data analysis, Vestar was then chided for rating patients with "breakthrough" lesions as partial responders. Ad hoc committee advisor Susan E. Krown, carefully explained that with breakthrough lesions, this "response to treatment cannot meet the definition." (The Pink Sheet, 1993)
In order to implement this ACTG staging and response criteria, a patient evaluation form was drafted to document the extent of the patient's disease. This form is completed at baseline and with each follow-up visit. In addition, a complete physical examination, including a chest x-ray, a biopsy for KS confirmation and a series of photographs should take place before the evaluation process begins. The evaluation process that occurs while the patient is on drug can be subjective.
Originally, Krown and colleagues (Krown 1989) suggested that the "exact number" of lesions not be counted, but be classified (range counting) into one of four categories: 1 = 0; 2 = 10 or less; 3 = 10 to 50; 4 = more than 50. However, with the commencement of ACTG 163, the Oncology Committee mandated that all lesions must be counted on a patient's body with 50 lesions or less. If the patient has over 50 lesions, then a designated area of the patient's body was to be selected (i.e., the back) where all of those lesions will then be counted.
There are some physicians who prefer the older range counting method and feel that it is adequate because it is "impossible to count the exact number of lesions -- especially when there are more than 25 -- on a patient's body." (Alvin Friedman-Kien, personal communication) Others maintain that it is possible to count lesions at each clinic visit, especially when there are fewer than 50 -- "We do this all the time," according to Susan Krown.
This will be one of the many crucial issues that the FDA's Oncologic Drugs Advisory Committee (ODAC) must realize if they are to rationally evaluate KS NDAs. Besides the response rates, ODAC needs to also consider quality of life, complications and toxicity when determining the efficacy and usefulness of the KS agent.
In 1994, we now know that there are a variety of factors that help predict a patient's response to therapy. In a retrospective analysis of 688 patients with AIDS-KS at UCLA Medical Center from 1981 to 1990, Miles and colleagues were able to identify four baseline variables that best predicted survival: CD4 cell number, the volume of red cells (erythrocytes) in the blood, the number of KS lesions, and body mass index (Miles: unpublished data). Myskowski and colleagues have also documented the location of a patient's lesions is a predictor of a patient's response to therapy. They also documented that a KS patient's CD4/CD8 ratio can predict survival (Myskowski 1988). Using this ratio as a prognostic variable for patients with KS as their first AIDS defining illness, Myskowski and colleagues reported: patients with normal ratios (>1.7) had a median survival time of 49 months; those with ratios from 1.0-1.7 had 29 months; and those with ratios < 1.0 had 14 months.
Other factors that correlate with overall survival include past or present OIs, the presence of "B" constitutional symptoms and the extent of disease (Mitsuyasu 1986; Taylor 1986). Lesions which appear after the onset of an OI, or during an OI, follow a more aggressive course and may be difficult to treat (Buchbinder 1992; Donald Abrams, personal communication).
A host of other immunologic activation markers which include beta 2 microglobulin, neopterin, in vitro proliferative T-cell response, levels of endogenous interferon, and IL-6 level have all shown to have some prognostic value (Taylor 1986; Vadhan-Raj 1986; Krown 1991). Most recently, Martinez-Maza and colleagues (Martinez-Maza 1993) at UCLA documented that high levels of IL-6 were identified retrospectively in HIV positive patients 6 months to a year before they developed their first KS lesion.
With more conclusive pathogenesis research, it will be essential to measure cytokine levels (IL-6, IL-1, TNF, GM-CSF and alpha interferon) at baseline before initiating treatment of experimental KS therapies. If certain therapies are aimed at inhibiting various cytokines, which many do in vitro, then baseline and subsequent measurements from lesion biopsies may assist in detecting activity.