Innovative methods for the identification of predictive biomarker signatures in oncology: Application to bevacizumab

Current methods for subgroup analyses of data collected from randomized clinical trials (RCTs) may lead to false-positives from multiple testing, lack power to detect moderate but clinically meaningful differences, or be too simplistic in characterizing patients who may benefit from treatment. Herein, we present a general procedure based on a set of newly developed statistical methods for the identification and evaluation of complex multivariate predictors of treatment effect. Furthermore, we implemented this procedure to identify a subgroup of patients who may receive the largest benefit from bevacizumab treatment using a panel of 10 biomarkers measured at baseline in patients enrolled on two RCTs investigating bevacizumab in metastatic breast cancer. Data were collected from patients with human epidermal growth factor receptor 2 (HER2)-negative (AVADO) and HER2-positive (AVEREL) metastatic breast cancer. We first developed a classification rule based on an estimated individual scoring system, using data from the AVADO study only. The classification rule takes into consideration a panel of biomarkers, including vascular endothelial growth factor (VEGF)-A. We then classified the patients in the independent AVEREL study into patient groups according to “promising” or “not-promising” treatment benefit based on this rule and conducted a statistical analysis within these subgroups to compute point estimates, confidence intervals, and p-values for treatment effect and its interaction. In the group with promising treatment benefit in the AVEREL study, the estimated hazard ratio of bevacizumab versus placebo for progression-free survival was 0.687 (95% confidence interval [CI]: 0.462–1.024, p = 0.065), while in the not-promising group the hazard ratio (HR) was 1.152 (95% CI: 0.526–2.524, p = 0.723). Using the median level of VEGF-A from the AVEREL study to divide the study population, then the HR becomes 0.711 (95% CI: 0.435–1.163, p = 0.174) in the promising group and 0.828 (95% CI: 0.496–1.380, p = 0.468) in the not-promising group. Similar results were obtained with the median VEGF-A levels from the AVADO study (“promising” group: HR = 0.709, 95%CI: 0.444–1.133, p = 0.151; “not-promising” group: HR = 0.851, 95% CI: 0.497–1.458, p = 0.556). Our analysis shows it is feasible to employ statistical methods for empirically constructing and validating a scoring system based on a panel of biomarkers. This scoring system can be used to estimate the treatment effect for individual patients and identify a subgroup of patients who may benefit from treatment. The proposed procedure can provide a general framework to organize many statistical methods (existing or to be developed) into a coherent set of analyses for the development of personalized medicines and has the potential of broad applications.

Bevacizumab improves the overall and progression-free survival of patients with metastatic colorectal cancer treated with 5-fluorouracil-based regimens irrespective of baseline risk

BACKGROUND

Kohne et al. [Ann Oncol 2002;13:308-317] showed that four prognostic variables can be used to classify patients with metastatic colorectal cancer (CRC) treated with 5-fluorouracil (5-FU)/leucovorin (LV) into three risk groups with different overall survival (OS). This model was applied to data from phase II/III trials of first-line bevacizumab plus 5-FU/LV with/without irinotecan (IFL).

METHODS

Data on tumor sites, Eastern Cooperative Oncology Group performance status, alkaline phosphatase levels and white blood cell counts were used to classify patients into Kohne prognostic high-, intermediate- and low-risk groups. Median OS and progression-free survival (PFS) were calculated for patients receiving 5-FU/LV plus bevacizumab or placebo (n = 489) and IFL plus bevacizumab or placebo (n = 812).

RESULTS

Median OS was longer in 5-FU/LV/bevacizumab (11.2-22.6 months) than in the 5-FU/LV/placebo (5.7-17.5 months), and in the IFL/bevacizumab arm (14.3-22.5 months) than in the IFL/placebo arm (8.4-17.9 months) across the Kohne high-, intermediate- and low-risk groups. The addition of bevacizumab also extended median PFS across the Kohne risk groups compared with placebo.

CONCLUSIONS

Bevacizumab improves OS and PFS across the Kohne risk classification in patients with metastatic CRC. The Kohne model can be extended to patients treated with 5-FU/LV/bevacizumab, IFL and IFL/bevacizumab and to PFS data.

Bevacizumab improves outcomes of patients with metastatic colorectal cancer (mCRC) treated with IFL with or without bevacizumab (BV) independent of baseline risk

3539

Background: It has been reported that patients with mCRC obtain benefit from BV independent of baseline risk factor status. Analysis of 3,825 patients treated with 5-FU/LV in 19 randomised trials conducted by Koehne et al. [Ann Oncol 2002;13:308–17] showed that four baseline prognostic variables can be used to categorise patients into three prognostic groups with different overall survival (OS). We have retrospectively applied this model to data from a phase III trial of first-line IFL with placebo or BV (AVF2107g). Methods: Data on ECOG performance status (PS), tumour site, alkaline phosphatase (AP) levels and white blood cell counts (WBC) were prospectively collected for the 813 patients randomised in AVF2107g. These data were retrospectively used to categorise patients in the IFL/placebo (n=411) and IFL/BV (n=401; data not available for 1 patient) arms into three risk groups: Low - ECOG PS 0/1, one tumour site; intermediate - ECOG PS 0/1, >1 tumour site, AP level <300U/L, or ECOG PS >1, low WBC count, one tumour site; high - ECOG PS 0/1, >1 tumour site, AP level >300U/L, or ECOG PS >1, high WBC count, >1 tumour site. Median OS and PFS in these groups were calculated. Results: OS (Table) and PFS were significantly longer in the IFL/BV groups compared to the IFL/placebo groups, as well as in the lower compared to the higher risk groups in both arms. Conclusions: The analysis supports previous analyses showing that adding BV to IFL improves OS independent of baseline patient risk. Furthermore, high-risk patients, who are less likely to benefit from IFL, benefit from IFL/BV. The model developed by Koehne et al may be applicable to regimens other than 5-FU/LV and validates the efficacy of BV first line in mCRC.

[Table: see text]

[Table: see text]

Assessing the safety of drugs in pregnancy: the role of prospective cohort studies

Since, for obvious reasons, systematic testing of the teratogenic properties of drugs in humans is not possible in the premarketing phase, the epidemiological approaches to postmarketing risk evaluation are of major importance. Cohort studies, with their prospective exposure assessment, their ability to study even exposure to drugs not commonly used in pregnancy, and their ability to monitor both adverse and beneficial fetal outcomes, seem to be the most promising study type from a methodological viewpoint. Although there are numerous cohort studies on the harmful effects of drug use in pregnant women, only a few have been able to demonstrate a risk in terms of an increase in the prevalence of malformations. Most studies with significant findings were those investigating the risk potential of one group of drugs, the anticonvulsants. The lack of cohort studies showing a risk for drug use in pregnancy, however, is not necessarily indicative of some methodological deficiency. Rather, it may suggest that, for the majority of drugs, their use in pregnancy is not associated with an increased risk of congenital malformations.

The PEGASUS project--a prospective cohort study for the investigation of drug use in pregnancy. PEGASUS Study Group

Since the thalidomide disaster, it is well accepted that drugs can have adverse effects on the developing human being. Although numerous studies show that medication during pregnancy is wide-spread, there still is a serious lack of comprehensive and valid data concerning the risks of drug use during pregnancy. One objective of the PEGASUS project, a population-based cohort study focusing on Munich, is to prospectively record information on drug exposure during pregnancy, to evaluate these data with regard to teratogenic properties, and to contribute to the quality assurance of medical treatment in pregnancy. The results of PEGASUS confirm that drug utilization during pregnancy is rather common--84% of the women use at least one preparation. The most frequent groups are minerals (mainly magnesium), iodide, and iron preparations. Although randomized studies have shown that periconceptional folic acid supplementation considerably reduces the risk of neural tube defects, only very few women in the PEGASUS project recorded folic acid intake during the critical period and in sufficient dosage. Moreover, only 42% of the pregnant women apply iodide prophylaxis in the recommended dosage.

Drug use assessment and risk evaluation in pregnancy—the PEGASUS-project

Since the thalidomide tragedy it is well accepted that drugs can have adverse effects on the unborn child. Although numerous studies show that drug use during pregnancy is widespread, there still is a serious lack of comprehensive and valid data on the risks of drug use during pregnancy. One objective of the PEGASUS-project, which focuses on Munich, is to enlarge the knowledge on embryo- and fetotoxic properties of drugs by prospectively recording information on drug exposure during pregnancy and analysing these data with regard to untoward fetal outcome. First results of PEGASUS confirm that drug utilization during pregnancy is rather common-85% of women use at least one preparation. The most frequent groups are haematologicals, minerals, iodide, and vitamins. Randomized studies have shown that periconceptional folic acid supplementation considerably reduces the risk of neural tube defects. However, only very few women in the PEGASUS-project recorded folic acid intake during the critical period or in sufficient dosage.

[Cardiovascular abnormalities in Bavaria 1984-1991]

The study presents data on cardiovascular malformations in Bavarian livebirths, born between 1984 and 1991. Cases have been ascertained retrospectively by reviewing hospital records of all children being referred to a children's hospital up to 2 years of age. The classification scheme was based on abnormalities in developmental mechanisms. Among 984,570 livebirths, 7020 cases with structural congenital heart disease were identified. The birth prevalence was 7.1 per 1000 livebirths. Between 1984 and 1991, total prevalence increased from 5.9/10(3) to 8.0/10(3). Prevalence in males was 7.3/10(3) and in females 6.9/10(3). 78.1% of all heart defects were isolated, the remaining 21.9% were associated either with chromosomal abnormalities (9.6%), non-chromosomal syndromes (1.0%), or noncardiac malformations of other organ systems (11.3%). Total fatality rate was 12.0%, with two thirds of deaths occurring within a month of birth or the following month of life. Data were compared with those of the Baltimore-Washington Infant Study. This study presents for the first time regional data on birth prevalences of congenital heart defects in Germany. The classification scheme reduces the wide spectrum of phenotype cardiovascular defects to several pathogenetic groups. The defects in each group may be related to similar causal factors.

Perinatal mortality in Bavaria, Germany, after the Chernobyl reactor accident

As has been shown by the authors of a paper recently published in this journal, a deviation from a long-term trend in perinatal mortality within the former Federal Republic of Germany occurred in 1987, i.e. 1 year following the Chernobyl disaster. It is the aim of this study to make a comparison between the areas of the state Bavaria. Germany, with different fallout levels as well as between the observed and expected numbers of perinatal deaths relating to these areas. The expected numbers of perinatal deaths, defined as external standard, were derived from the remainder of the former FRG. Testing an a priori formulated hypothesis revealed no differences in the temporal development of perinatal mortality between the areas with different fallout levels and subsequent exposure. Including May 1986 into the analysis revealed a significant increase during the first 3 months after the accident, which is due to an excess in May alone. Since no elevated radiation risks for the last days in utero are known, the additional Chernobyl radiation exposure is not plausible as a causative agent. Further analyses on stillbirths showed an increase in Southern Bavaria during the first 2 years following the accident. Later on, the rates were comparable to the expected values again.

Birth prevalence of congenital malformations in Bavaria, Germany, after the Chernobyl accident

This study considers whether or not exposure to radioactive fallout from the Chernobyl accident led to an increased prevalence of congenital malformations in infants born in Bavaria, the German state with the highest levels of contamination after the accident. The odds ratios for major malformations after the accident relative to before were used as indicators for adverse health effects. Since measurements of caesium in soil showed that contamination was considerably higher in Southern Bavaria than in Northern Bavaria, the odds ratios were calculated for both regions separately. Analysis did not show a significant increase in any of the odds ratios of the selected malformations in Southern Bavaria as compared to Northern Bavaria. Consequently, this study provides no evidence that radiation from Chernobyl caused an increase in the birth prevalence of major congenital malformations.

[Social ecological analysis of birth prevalence and mortality of children with malformations]

Based on retrospective data collected from all over Bavaria, the influence of socioeconomic status defined for geographic areas rather than individuals and urbanisation on birth prevalence, fatality rate, and survival time of malformed infants was investigated. Urbanisation shows a negative effect on all of the outcome variables: the higher the degree of urbanisation, the higher the birth prevalence and fatality rate of malformed infants, whereas survival time decreases with increasing degree of urbanisation. Social status has a positive, but not statistically significant, effect on the outcome variables "fatality rate" and "survival time", both of which can be influenced by postnatal care and medical intervention. Increasing social status increases birth prevalence. This effect is not statistically significant.