Abstract
Purpose: The purpose of this study was to evaluate the feasibility of incorporating a novel wound angiogenesis assay into a Phase I study of BMS-275291, a broad-spectrum matrix metalloproteinase inhibitor, and to determine whether the wound angiogenesis assay was able to detect the inhibition of angiogenesis in patients treated with BMS-275291.
Experimental Design: Before treatment began, a 4-mm skin biopsy was performed. The wound was imaged for 14 days. Treatment was started on day 0, and a separate 4-mm biopsy was performed 14 days later. The second wound was also imaged for 14 days. Wound angiogenesis was scored by two independent observers who were blinded to treatment status.
Results: The median times in days (95% confidence interval) to reach the target average vascular score (AVS) of 1.5 and 2.0 based on the data of Observer 1 were 3.7 (2.2–6.9) and 8.0 (5.0–10.0) pretreatment whereas on-treatment the values were 4.9 (3.7–8.0) and 9.3 (7.0–11.5), respectively. The delay in the median time to reach an AVS of 1.5 was 1.2 days or a 32% reduction when comparing pretreatment with on-treatment (P = 0.06). For the target AVS of 2.0 the delay in the median time pretreatment versus on-treatment was 1.3 days or a 16% reduction (P = 0.04).
Conclusions: The wound angiogenesis assay used in this study was practical, well tolerated, and reproducible. Delays in wound angiogenesis because of BMS-275291 were detectable with this assay. This technique warrants additional investigation in clinical trials of other antiangiogenic agents.
Introduction
Angiogenesis has been shown to be critically involved in tumor development and progression and predictive for prognosis or clinical stage in most solid tumors that have been assessed (1, 2, 3, 4). Angiogenesis is required for many normal physiological processes, particularly embryonic development, reproductive functions, and wound healing (5, 6, 7). As an anticancer strategy, targeting tumor-associated endothelial cell may have several advantages compared with targeting the tumor cell directly, an approach that that has been extensively reviewed (8, 9, 10). There is the potential for limited toxicity, because nontumor-related angiogenesis in the adult is largely limited to wound healing and reproductive functions. Drug delivery is also simplified because endothelial cells line the vascular space, whereas drug diffusion is not inhibited by high interstitial pressures. In addition, endothelia are genetically stable, which reduces the potential for drug resistance related to mutation or loss of drug targets, effectors, or transporters. Lastly, many traditional cytotoxic agents demonstrate increased antitumor activity with little or no additional toxicity when combined with antiangiogenesis therapies (11). For these reasons, many new angiogenesis inhibitors have entered clinical testing.
Angiogenesis is a complex and dynamic process involving the interplay of multiple factors. These factors include multiple cellular and tissue matrix components, and many soluble and tissue-bound growth factors as well as their endogenous inhibitors. The role of these factors has been reviewed extensively (12, 13, 14, 15). Many of these growth regulators act in a paracrine or autocrine fashion and are stored in nascent form in the tissue matrix. These factors are released and activated by proteases and metalloproteinases secreted by tumor cells, endothelia, and activated macrophages associated with tumor angiogenesis (16, 17, 18, 19, 20). In addition, many of these growth factors are capable of up-regulating other angiogenic growth factors or their receptors, resulting in positive feedback loops that promote angiogenesis (21, 22, 23, 24).
Digestion of the tissue matrix by a class of enzymes known as MMPs3 is required for invasion and migration of tumor cells, endothelia, and other tumor-associated host cells such as fibroblasts and macrophages. Importantly, tumor matrix is strikingly similar to the provisional matrix that is part of normal wound healing, which is rich in fibrin, fibronectin, vitronectin, and nonfibrillar forms of collagen (collagen IV; Ref. 25). Under physiological conditions, this provisional matrix is eventually replaced by a matrix composed predominantly of mature collagens, proteoglycans, and glycoproteins. However, in contrast to normal wound healing, the tumor-associated matrix persists in its provisional form and does not mature (25). For several human tumors, elevated expression of MMPs has been associated with increased local invasiveness and worsened prognosis (26, 27, 28, 29, 30, 31). In addition, MMPIs have been shown in vivo to inhibit tumor angiogenesis and tumor growth in several murine models (32, 33). The importance of MMPs in cancer has been reviewed previously (34, 35).
On the basis of the central role of MMPs in tumor progression, especially in tumor angiogenesis, MMPIs represent rational anticancer agents. Many MMPIs have been developed and have undergone clinical testing alone or in combination with standard chemotherapy agents in variety of tumor types and disease settings (early versus advanced disease). Most of these studies have demonstrated that MMPIs have little efficacy, and the development of some of these agents has been halted (36, 37, 38, 39). A lack of measurable efficacy in patients highlights the need to develop reliable biomarkers that reflect activation or inhibition of MMPs in vivo.
BMS-275291 is a novel p.o. bioavailable MMPI that has been shown in vitro to inhibit MMP-1, MMP-2, MMP-8, MMP-9, MMP-13, and MMP-14 in the nanomolar range. BMS-275291 also has shown antitumor activity in xenograft models with the ability to reduce the size and number of metastases in a dose-dependent fashion (40). It was also able to inhibit endothelial migration into Matrigel plugs (40).
In contrast to most other MMPIs, which are hydroxamate-based, BMS-275291 has a mercaptoacyl group that binds the Zn2+ ion at the active site of the enzyme (41). This property confers several pharmacological advantages. In addition, BMS-275291 was rationally designed to avoid inhibition of the sheddases, a class of MMP-related enzymes involved in tumor necrosis factor α receptor homeostasis. Altered tumor necrosis factor α binding and activity are thought to be involved in the development of the inflammatory arthritis and tendonitis seen with other MMPIs (42).
Determining the recommended Phase II dose for a novel agent such as BMS-275291 is complex given the potential of this class of agents for minimal toxicity and their potential to produce tumor dormancy with disease stabilization instead of regression. In contrast to traditional cytotoxic agents, new agents that target angiogenesis may have minimal toxicity at therapeutic doses. Optimal dose selection is particularly important for this class of agents because excess dosing may cause loss of specificity and unnecessary toxicity. In addition, many of these agents target endothelial receptor signaling pathways. Many biological systems do not have monotonic dose response curves, often because of compensatory up-regulation of receptor levels in the setting of growth factor inhibition, and vice versa (43, 44). For example, IFN-α has been shown to have antiangiogenic properties in both animal models and the clinic, where it has been used to treat life-threatening pediatric hemangiomas (45). In mouse models, the optimal antiangiogenic dose, as measured by basic fibroblast growth factor and MMP-9 mRNA expression, was well below the maximal tolerated dose (46). For these reasons, to determine recommended doses for Phase II and III trials, it would be extremely helpful to have a measure of biological effect for an angiogenesis inhibitor that did not depend on the traditional end point of dose-limiting toxicity.
We have developed a novel and practical wound angiogenesis assay to evaluate angiogenesis inhibitors in the clinic. The assay is based on the significant mechanistic similarities between tumor angiogenesis and wound angiogenesis. The wound angiogenesis assay uses a small, full thickness punch biopsy that serves as a stimulus for angiogenesis. In healthy volunteers, we have shown that this assay is feasible, well tolerated, and generates a highly reproducible pattern of neovascularization at the wound periphery that can be measured over time (47). In many respects the assay is similar to the commonly used corneal pocket, chorioallantoic membrane, Matrigel plug, and dorsal flap window chamber models. The use of a dermal wound also has several practical and theoretical advantages for use in clinical trials. In contrast to most solid tumors, skin is a healthy and easily accessible tissue, and wound angiogenesis is a reproducible and physiological process. Use of a dermal wound permits essentially continuous observation, repeated assays, and minimizes intra- and intersubject variation because of tissue heterogeneity and variability in sample acquisition. In particular, the potential to perform this assay before treatment and again at any point during treatment allows each patient to serve as his or her own control. In addition, wound angiogenesis represents a potentially global readout for multiple pro- and antiangiogenesis factors and pathways. There is also recent evidence that antiangiogenesis agents may reduce the wound vascular response (48, 49, 50). Lastly, because altered wound healing and increased bleeding tendencies are theoretical risks associated with angiogenesis inhibitors, this assay may also provide an early signal for such potential toxicities.
We have used this assay as part of a Phase I study of the novel angiogenesis inhibitor, BMS-275291. The primary objective of this wound study was to determine whether the wound angiogenesis assay was able to detect the inhibition of wound angiogenesis in patients treated with BMS-27291, and whether the effect was dose-dependent.
Patients and Methods
Criteria for enrollment were the same as for the treatment protocol, which were standard for most Phase I studies. These included solid tumor malignancy for which there was not standard effective therapy; adequate hematologic, renal, and hepatic function; Karnofsky performance status ≤2; no comorbid conditions that would affect absorption of the medication or jeopardize patient safety or compliance; age ≥18; not pregnant; using adequate birth control if of child-bearing potential; and provision of informed consent. The wound angiogenesis protocol was part of the treatment protocol, which was approved by the Duke Institutional Review Board.
The patients on study were enrolled at one of five dose levels of BMS-275291 administered p.o. once daily (600, 900, 1200, 1800, and 2400 mg). The Phase I and pharmacokinetic portion of this study are reported separately.4 BMS-275291 contains a free sulfhydryl group and forms disulfide dimers with other sulfhydryl-containing molecules in vivo. Therefore, blood samples were collected into tubes containing methyl acrylate to react with free sulfhydryl groups of monomeric BMS-275291 and on-line extraction liquid chromatography/tandem mass spectrometry analytical methods were developed to quantitate both BMS-275291 and ‘total’ BMS-275291 (BMS-275291 plus any BMS-275291 recovered from reducible disulfides containing BMS-275291). The plasma concentration-time data for parent and total BMS-275291 were analyzed by noncompartmental methods using the program MENU/PKMENU. Cmin values reported here are for monomeric or “free” BMS275291.
Before treatment (day −14), a 4-mm punch biopsy wound (Fray Products Corp., Buffalo, NY) was made on the upper medial aspect of the forearm of the subject after local anesthesia with 1% Lidocaine. A topical antibiotic ointment (Bacitracin Ointment; USP, Clay-Park Labs, Inc., Bronx, NY) was applied to the wound before being covered with a nonocclusive bandage (Coverlet; Beiersdorf-Jobst, Inc., Rutherford College, NC).
The patients were assessed three times per week beginning on day −13. For each observation, the subject wound was filled with saline before the application of aqueous ointment (K-Y Jelly; Johnson and Johnson Medical Inc., Arlington, TX) and a sterile microscope slide. The wounds were videotaped at low (×25) and high (×80) magnifications with a color CCD camera (Model DXC 151A; Sony Corp., Tokyo, Japan) attached to a Zeiss operating microscope (Stemi DRC; Carl Zeiss Inc., Thornwood, NY). After imaging, additional ointment and a dressing were applied to minimize wound drying between assessments. At the end of the 14-day assessment period (day 0), treatment with BMS-275291 was initiated. After the patient had been on therapy for 14 days (day +14), when drug levels were presumed to be at steady state, a second 4-mm biopsy was performed using the contralateral arm of the patient. This second wound was also videotaped for 14 days (Fig. 1).
Wound Scoring Methods.
Videotaped wound images were analyzed and scored in sequence by two observers (A. C. L. and R. D. B.) who were blinded to the treatment status of the subjects. All of the images were scored using an ordinal 0–4 scoring system of wound vascularization (Fig. 2; Ref. 2) that was developed and refined in healthy volunteers (47):
0: no vessels evident, normal appearing skin;
1: slight erythema, small caliber irregularly distributed vessels;
2: small vascular loops aligned perpendicularly to the skin surface;
3: looping vessels with partial radial alignment; and
4: organized, radially aligned vessels parallel to the skin surface.
Each wound field was divided into 12 clock hours, each section scored, and an AVS was obtained for each time point. If a clock hour was not visible at a given time point, that score was assigned conservatively using interpolation from the value obtained previously.
Statistical Methods.
A total of 291 observation time points were available for analysis on the 20 patients in this study. Each patient served as their own control, with pretreatment scores compared with those obtained on treatment. The sample size for the Phase I study was not determined based on the primary objective of this wound study. Therefore, these analyses are exploratory. Because observers were unblinded to patient and time sequence, the wound scores from the two observers were analyzed separately to minimize potential bias. The wound scores from Observer 1 were used as the primary data, and the scores from Observer 2 are presented for confirmation.
The Kaplan-Meier survival analysis method was used to estimate the time after biopsy to reach the target AVS, and the Prentice modification of the Wilcoxon signed-rank test for paired data with censoring (51) was used to test the change in the specified time until event scores pretreatment versus on-treatment. The target AVS of 1.5, 2, 2.5, and 3 were chosen a priori based on our experience with healthy volunteers. If the target AVS was reached between scheduled assessments, the time to that target AVS was determined using linear interpolation. Results related to the target AVS 2.5 and 3.0 are not presented because of the high percentage of censored observations, where the target AVS had not been reached by the time of the final measurement pretreatment or on-treatment of the subject. Spearman’s rank correlation coefficient was used to assess whether a positive or negative linear correlation existed between the change in time until target scores and the dose of BMS-275291, ignoring censoring. The Cox regression was considered as well when examining a possible dose-response relationship.
A longitudinal evaluation using the mixed-effects regression model was used to explore the contribution of predetermined predictor variables to determining the variability in the mean AVS including variables related to treatment. This method was selected because each patient had repeated observations, and the number of observations was unbalanced across patients. In this model, the raw AVS was the dependent variable, and all of the predictor variables had fixed effects except for the intercept, which varied with each subject and was used to account for the correlation among the observations from the same subject. The primary predictor variables considered were time after biopsy, age, gender, performance status, serum albumin, time to disease progression, and smoking history. Incorporating these variables, the pharmacological parameters, treatment (pretreatment versus on-treatment), Cmin, and dose, were also considered for their effect on the mean AVS in individual analyses. All of the treatment variables were considered continuous except for dose, which was analyzed as a three-tiered variable (low = 600 and 900 mg, medium = 1200 mg, and high = 1800 and 2400 mg). The main reason for using this categorized dose variable is to guard against the possibility of a model misspecification of having a linear dose-response relationship while trying to estimate the dose effect as accurately as possible. This is particularly important in our case because of the few patients in the study. The final model was determined by a stepwise-like selection with some low-order interaction terms examined and included. Variables were added or dropped from the model at a significance level of 0.05. All of the statistical testing was performed using SAS statistical software (SAS Institute Inc., Cary, NC), and all of the Ps are two-sided.
Results
Twenty-four patients enrolled in the treatment study of BMS-27291 at five dose levels. Twenty-two of these patients enrolled in the companion wound angiogenesis study. Two patients were removed from the wound angiogenesis portion of the study for disease progression before completing the on-treatment evaluations. Twenty patients completed all of the pretreatment and on-treatment evaluations, and were fully evaluable. The distribution of patients according to dose of BMS-275291 was as follows: 4 at 600 mg, 3 at 900 mg, 7 at 1200 mg, 3 at 1800 mg, and 3 at 2400 mg. The median age of the patients who completed the study was 52.5 years (range, 18–75 years). There were 16 men and 4 women. No subject withdrew from this study because of problems related to the skin biopsies or evaluations. There were no clinically significant alterations in wound healing observed.
The median time in days (95% confidence interval) required to reach the target AVS of 1.5 and 2.0 for Observer 1 are 3.7 (2.2–6.9) and 8.0 (5.0–10.0) pretreatment, whereas on-treatment the values were 4.9 (3.7–8.0) and 9.3 (7.0–11.5), respectively (Table 1). Kaplan-Meier curves representing these results are shown in Figs. 3 and 4. The delay in the median time to reach an AVS of 1.5 was 1.2 days or a 32% reduction, which approached statistical significance using the Prentice Wilcoxon test (P = 0.06) when comparing pretreatment with on-treatment. For the target AVS of 2.0 the delay in the median time pretreatment versus on-treatment was 1.3 days or a 16% reduction and reached statistical significance (P = 0.04).
The median times to reach AVS of 2.5 and 3.0 were not statistically significant (P = 0.62 and P = 0.79, respectively). At these target AVS (2.5 and 3.0), the high percentage of censored observations (50% and 70%, respectively), where the target score had not been reached at the time of the final measurement of the subject, made these comparisons less powerful. Using the Prentice modification of the Wilcoxon signed-rank test, the significance of difference in the median time to reach the target scores of 1.5 and 2.0 for Observer 2 was 0.08 and 0.03, respectively. These results were in line with the above test parameters for Observer 1. The Spearman rank correlation coefficients between AVS measurements of Observer 1 and dose were 0.05 (P = 0.84) and −0.01 (P = 0.95) for the time to reach AVS of 1.5 and 2.0, respectively, indicating that there was likely no linear dose-response effect because of BMS-275291 in terms of the time to the target AVS scores. Similar results indicating an absence of dose response were found when using the Cox regression approach and grouping the doses as low (600 and 900 mg), medium (1200 mg), and high (1800 and 2400 mg). Results for Observer 2 were similar.
A mixed-effects regression model was selected to identify the predictors of the variability in the mean AVS including variables related to treatment (Table 2). Interpretation of the main effects of each variable was not straightforward because of the presence of significant interaction terms in the model. If an interaction term was significant, then both main effect terms are included in the model whether or not they are significant on its own.
On the basis of the final mixed-effects model results, the treatment variable, pretreatment versus on-treatment, appeared to be an important predictor for the variability in the measured mean AVS (P < 0.05). With pretreatment versus on-treatment included in the model, average Cmin had additional predictive power in determining the variability in the mean AVS (P = 0.003). When considered in the same manner as the average Cmin, the contribution of the dose of BMS-275291 to determining the variability in the mean AVS was of borderline statistical significance (P = 0.045), and this variable had a significant interaction with patient age. To account for the multiple tests performed in the model, this P is not considered significant.
In addition, the patient characteristics age and gender, as well as time after wounding, were also found to contribute significantly to the variability in the mean AVS outcome (P < 0.05). Fig. 5 is a graphical illustration of the treatment effect when the other model variables except gender and time are fixed. Gender appeared to modify the wound angiogenesis response to treatment. Under the final regression model, the treatment effect on the mean AVS for females was constant over time and estimated to be −0.307 (on- versus pretreatment, P = 0.015), whereas the effect for males was approximately −0.044 only (P = 0.44) after adjusting for all of the other model variables. However, with only 4 women enrolled in the study and a relatively small sample size (n = 20) this result should be interpreted only as exploratory. Additional studies are needed to pinpoint and confirm the effect size.
Performance status, serum albumin, time to disease progression, and smoking history did not seem to contribute significantly to the variability of the mean AVS. In the final regression model, observer was found to contribute significantly (P = 0.04) to the variability in the mean AVS. To additionally evaluate the magnitude of the observer effect, a univariate analysis with AVS as the dependent variable and observer as the predictor variable was performed. The mean difference in observer AVS scores pretreatment and on-treatment were −0.028 and −0.07, respectively. Although a statistical difference is present between the mean AVS of the observers, this discrepancy is on average less than the value of one clock hour scored differently and probably of no clinical importance.
Discussion
We have proposed a technique of monitoring wound angiogenesis that is well tolerated, practical, and highly reproducible in the clinical setting. This dermal wound assay was incorporated into the clinical evaluation of an p.o. bioavailable MMPI BMS-275291. From the data, a modest but significant delay in the normal wound angiogenesis process (the time to reach target AVS of 1.5 and 2.0) was observed using the Prentice-Wilcoxon test. In addition, treatment with BMS-275291 contributed significantly to the outcome of the mean AVS based on results from the mixed-effects regression model. The treatment effect was not found to be dose-dependent using either analysis technique although the study was not designed to test for a dose-response relationship. The lack of a dose effect can be explained by the few patients studied at each dose level. Furthermore, 19 of 20 patients had Cmin levels greater than the in vitro concentration at which 90% inhibition is achieved for MMP2 of 124 μg/ml, and 17 of 20 had Cmin levels greater than the in vitro concentration at which 90% inhibition is achieved for MMP9 of 261 μg/ml. Dosing for this patient study was guided by preceding pharmacokinetic studies of BMS-275291 in healthy volunteers. There was no correlation observed between the time to tumor progression and a delay in wound angiogenesis.
Note that one limitation of our Kaplan-Meier analyses of the time to target AVS scores is that a linear interpolation was used to estimate the time to a target score whenever the target had been reached at an observation time. Another limitation is the Kaplan-Meier approach did not use all of the observed data over time directly.
The mixed effects model, a longitudinal evaluation of the data, did not have these limitations. Fig. 5 is an illustration of the treatment effect when all of the other model variables except gender and time are fixed. Because of the model assumptions, study design, and available data, the following remarks are in order. First, the treatment effect is assumed to be constant over time for either gender, which may or may not be true. Because the sample size is small, we could not verify this assumption within this study. Second, the treatment effect shown in the graph for each of the two genders is the one solely attributable to gender after adjusting for all other factors. It should not be confused with overall treatment effect not adjusted for any covariates. Finally, because of the imbalance in gender and the small sample size, the estimated treatment effect attributable to gender is very preliminary (i.e., much larger treatment effect for females than that for males) and should be interpreted only as exploratory. Additional studies are needed to pinpoint and confirm the effect size.
Nonetheless, both methods produced significant treatment effect and were in support of each other’s conclusions.
Because many patients did not reach an AVS of 2.5 or 3.0 in the 14-day period different from in the normal volunteers, we plan to use a longer observation period in future studies, particularly for older and sicker patients. To better estimate the delay in the median time to target AVS scores, more frequent evaluations will also be considered.
This dermal wound assay is based on the molecular and mechanistic similarities between wound angiogenesis and tumor angiogenesism and has the theoretical advantage of providing a global readout of all known and unknown pro- and antiangiogenesis factors in vivo (25, 52). Most of the cells, cytokines, matrix components, and plasma proteins involved in tumor angiogenesis are the same as those involved in wound angiogenesis (8, 11, 53, 54). Recent observations also indicate common patterns of gene expression between the two processes (55). The wound angiogenesis assay was designed to allow the evaluation of physiological angiogenesis in vivo without the limitations and risks associated with repeated biopsies of solid tumors. Despite the similarities, the wound and tumor microenvironments are distinct, and the wound angiogenesis assay may provide evidence of drug target inhibition whether or not tumor angiogenesis is affected. This assay may also overcome the limitations associated with the measurement of systemic levels of single growth factors, such as vascular endothelial growth factor or basic fibroblast growth factor. Measurement of these growth factors is complicated by their binding to platelets and monocytes in the circulation, and to heparins and other glycoproteins in the tissue matrix (56, 57, 58, 59). Interpretation of systemic or urinary levels of these isolated growth factors is additionally complicated by their local production and storage, their paracrine mechanisms of action, and their complex and dynamic crossregulation (60, 61).
MMPs are a gene family of proteinases expressed by a variety of cells and have evolved to fulfill a variety of functions included in wound healing and angiogenesis both in the normal setting and in the pathologic setting. Wound healing, as measured by wound area over time or tensile strength, was not evaluated in this study. Despite the theoretical risk for BMS-275291 to cause an alteration in wound healing, we observed no clinically apparent adverse effects on wound healing during this study.
Most patients exhibited similar patterns of score progression that were reproduced by each of the blinded evaluators. These findings of delayed wound angiogenesis using a therapy that demonstrated tumor inhibition and angiogenesis inhibition preclinically provide evidence to the theory of shared mechanisms between tumor angiogenesis and wound angiogenesis. These results suggest that this technique warrants additional investigation in clinical trials of other antiangiogenic agents, combinations of antiangiogenic agents, and therapy combining antiangiogenic agents with cytotoxic drugs.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Supported in part by grants 1K23-CA85582 from the National Cancer Institute; MO1-RR-30 from the National Center for Research Resources, General Clinical Research Centers Program, NIH; and the Harrell-Horn Foundation and the Harrell-Carr fund of the Duke Comprehensive Cancer Center.
The abbreviations used are: MMP, matrix metalloproteinase; MMPI, matrix metalloproteinase inhibitor; AVS, average vascular score; Cmin, minimum concentration; VS, vascular score.
N. A. Rizvi, J. S. Humphrey, E. A. Ness, M. Johnson, E. Gupta, K. Williams, D. J. Daly, D. Sonnichsen, D. Conway, J. Marshall, and H. Hurwitz. A Phase I study of oral BMS-275291, a novel, non-hydroxamate, sheddase-sparing matrix metalloproteinase inhibitor (MMPI), in patients with advanced or metastatic cancer, submitted for publication.
. | AVS = 1.5 . | AVS = 2.0 . | |
---|---|---|---|
. | Median time in days (95% Confidence interval) . | . | |
Observer 1 | |||
Pretreatment | 3.7 (2.2–6.9) | 8.0 (5.0–10.0) | |
On-treatment | 4.9 (3.7–8.0) | 9.3 (7.0–11.5) | |
P | 0.06 | 0.04 | |
Observer 2 | |||
Pretreatment | 2.9 (2.3–5.8) | 7.0 (3.9–8.5) | |
On-treatment | 4.1 (3.0–7.0) | 8.9 (7.0–9.8) | |
P | 0.08 | 0.03 |
. | AVS = 1.5 . | AVS = 2.0 . | |
---|---|---|---|
. | Median time in days (95% Confidence interval) . | . | |
Observer 1 | |||
Pretreatment | 3.7 (2.2–6.9) | 8.0 (5.0–10.0) | |
On-treatment | 4.9 (3.7–8.0) | 9.3 (7.0–11.5) | |
P | 0.06 | 0.04 | |
Observer 2 | |||
Pretreatment | 2.9 (2.3–5.8) | 7.0 (3.9–8.5) | |
On-treatment | 4.1 (3.0–7.0) | 8.9 (7.0–9.8) | |
P | 0.08 | 0.03 |
Variable . | Estimated regression coefficient (SE) . | P . |
---|---|---|
Intercept | 1.96 (0.25) | <0.0001 |
Time | 0.27 (0.04) | <0.0001 |
Age | −0.044 (0.009) | 0.0005 |
Female (vs. male) | −0.023 (0.08) | 0.79 |
Obs. 1 (vs. Obs. 2) | 0.004 (0.03) | 0.89 |
On treatment (vs. pretreatment) | −0.044 (0.06) | 0.44 |
Time2 | −0.02 (0.006) | 0.0014 |
Time3 | 0.0007 (0.0002) | 0.007 |
Age2 | 0.0004 (0.0001) | 0.0018 |
Time*Obs 1 (vs. time*Obs 2) | −0.008 (0.004) | 0.04 |
Female*on-treatment (vs. others) | −0.24 (0.09) | 0.015 |
Variable . | Estimated regression coefficient (SE) . | P . |
---|---|---|
Intercept | 1.96 (0.25) | <0.0001 |
Time | 0.27 (0.04) | <0.0001 |
Age | −0.044 (0.009) | 0.0005 |
Female (vs. male) | −0.023 (0.08) | 0.79 |
Obs. 1 (vs. Obs. 2) | 0.004 (0.03) | 0.89 |
On treatment (vs. pretreatment) | −0.044 (0.06) | 0.44 |
Time2 | −0.02 (0.006) | 0.0014 |
Time3 | 0.0007 (0.0002) | 0.007 |
Age2 | 0.0004 (0.0001) | 0.0018 |
Time*Obs 1 (vs. time*Obs 2) | −0.008 (0.004) | 0.04 |
Female*on-treatment (vs. others) | −0.24 (0.09) | 0.015 |
Acknowledgments
We thank the nursing staff of the Duke University Medical Center General Clinical Research Center for their assistance and the members of the Duke University Medical Center Phase I research group for their hard work in completing this project.