Multivariate analysis of flow data in breast lesions and validation in a normal clinical setting.

PURPOSE: To improve differentiation between benign and malignant breast lesions by Doppler measurements and to validate results in a normal clinical setting in comparison to study conditions. MATERIALS AND METHODS: Doppler measurements of 458 patients were compared in benign and malignant tumors in a prospective study. In a multivariate analysis a diagnostic score was developed using a logistic regression model and stepwise selection. These results were compared with 272 patients who were examined under routine clinical conditions. RESULTS: Most measurements showed highly significant (p < 0.001) differences between benign and malignant tumors. For each measurement we considered two cut-points to define a diagnostic rule. Despite significant differences, none of the corresponding classification rules exceeded 90 % sensitivity and specificity. Multivariate analysis selected a model including age and the number of arteries and contralateral arteries. Although significant, the last factor barely improved diagnostic accuracy. Therefore, we deleted it from the multivariate model. Based on a simple model including age and the number of tumor arteries, we defined classification rules with high sensitivity and specificity. The RI measurement did not improve the discriminatory power of our score. In the validation study the sensitivity decreased from 89 - 98 % to 58 - 78 % with a specificity of 82 - 92 % vs. 83 - 86 %. CONCLUSION: Color Doppler can be used for breast cancer differentiation. However, in the clinical routine the sensitivity decreases considerably compared with optimized study conditions.

Reporting of prognostic studies of tumour markers: a review of published articles in relation to REMARK guidelines.

BACKGROUND: Poor reporting compromises the reliability and clinical value of prognostic tumour marker studies. We review articles to assess the reporting of patients and events using REMARK guidelines, at the time of guideline publication. METHODS: We sampled 50 prognostic tumour marker studies from higher impact cancer journals between 2006 and 2007. The inclusion criteria were cancer; focus on single biological tumour marker; survival analysis; multivariable analysis; and not gene array or proteomic data. Articles were assessed for the REMARK profile and other REMARK guideline items. We propose a reporting aid, the REMARK profile, motivated by the CONSORT flowchart. RESULTS: In 50 studies assessed for the REMARK profile, the number of eligible patients (56% of articles), excluded patients (54%) and patients in analyses (98%) was reported. Only 50% of articles reported the number of outcome events. In multivariable analyses, 54% and 30% of articles reported patient and event numbers for all variables. Of the studies, 66% used archival samples, indicating a potentially biased patient selection. Only 36% of studies reported clearly defined outcomes. CONCLUSIONS: Good reporting is critical for the interpretability and clinical applicability of prognostic studies. Current reporting of key information, such as the number of outcome events in all patients and subgroups, is poor. Use of the REMARK profile would greatly improve reporting and enhance prognostic research.

Adding local components to global functions for continuous covariates in multivariable regression modeling.

When global techniques, based on fractional polynomials (FPs), are employed for modeling potentially nonlinear effects of several continuous covariates on a response, accessible model equations are obtained. However, local features might be missed. Therefore, a procedure is introduced, which systematically checks model fits, obtained by the multivariable fractional polynomial (MFP) approach, for overlooked local features. Statistically significant local polynomials are then parsimoniously added. This approach, called MFP + L, is seen to result in an effective control of the Type I error with respect to the addition of local components in a small simulation study with univariate and multivariable settings. Prediction performance is compared with that of a penalized regression spline technique. In a setting unfavorable for FPs, the latter outperforms the MFP approach, if there is much information in the data. However, the addition of local features reduces this performance difference. There is only a small detrimental effect in settings where the MFP approach performs better. In an application example with children's respiratory health data, fits from the spline-based approach indicate many local features, but MFP + L adds only few significant features, which seem to have good support in the data. The proposed approach may be expected to be superior in settings with local features, but retains the good properties of the MFP approach in a large number of settings where global functions are sufficient.

Prognostic markers in cancer: the evolution of evidence from single studies to meta-analysis, and beyond.

In oncology, prognostic markers are clinical measures used to help elicit an individual patient's risk of a future outcome, such as recurrence of disease after primary treatment. They thus facilitate individual treatment choice and aid in patient counselling. Evidence-based results regarding prognostic markers are therefore very important to both clinicians and their patients. However, there is increasing awareness that prognostic marker studies have been neglected in the drive to improve medical research. Large protocol-driven, prospective studies are the ideal, with appropriate statistical analysis and clear, unbiased reporting of the methods used and the results obtained. Unfortunately, published prognostic studies rarely meet such standards, and systematic reviews and meta-analyses are often only able to draw attention to the paucity of good-quality evidence. We discuss how better-quality prognostic marker evidence can evolve over time from initial exploratory studies, to large protocol-driven primary studies, and then to meta-analysis or even beyond, to large prospectively planned pooled analyses and to the initiation of tumour banks. We highlight articles that facilitate each stage of this process, and that promote current guidelines aimed at improving the design, analysis, and reporting of prognostic marker research. We also outline why collaborative, multi-centre, and multi-disciplinary teams should be an essential part of future studies.

Investigation on the improvement of prediction by bootstrap model averaging.

OBJECTIVES: We illustrate a recently proposed two-step bootstrap model averaging (bootstrap MA) approach to cope with model selection uncertainty. The predictive performance is investigated in an example and in a simulation study. Results are compared to those derived from other model selection methods. METHODS: In the framework of the linear regression model we use the two-step bootstrap MA, which consists of a screening step to eliminate covariates thought to have no influence on the response, and a model-averaging step. We also apply the full model, variable selection using backward elimination based on Akaike's Information Criterion (AIC), the Bayes Information Criterion (BIC) and the bagging approach. The predictive performance is measured by the mean squared error (MSE) and the coverage of confidence intervals for the true response. RESULTS: We obtained similar results for all approaches in the example. In the simulation the MSE was reduced by all approaches in comparison to the full model. The smallest values are obtained for bootstrap MA. Only the bootstrap MA and the full model correctly estimated the nominal coverage. The backward elimination procedures led to substantial underestimation and bagging to an overestimation of the true coverage. The screening step of bootstrap MA eliminates most of the unimportant factors. CONCLUSION: The new bootstrap MA approach shows promising results for predictive performance. It increases practical usefulness by eliminating unimportant factors in the screening step.

Role of isolated locoregional recurrence of breast cancer: results of four prospective studies.

PURPOSE: We evaluated the effect on future prognosis of an isolated locoregional recurrence (ILRR) after the primary diagnosis of breast cancer. Using data from four prospective studies of the German Breast Cancer Study Group, we investigated factors influencing prognosis after ILRR and defined a simple classification of patients into groups with different prognoses. PATIENTS AND METHODS: From 1983 to 1989, 2,746 patients were recruited into four studies comparing different treatments in primary breast cancer. After a median follow-up time of 8 years, 337 patients developed an ILRR as the first event. The influence of ILRRs on disease progression was examined. The effects of different prognostic factors on progression-free survival (PFS) and overall survival after ILRR were analyzed after a median follow-up time of 4.5 years. RESULTS: ILRRs increased the risk with respect to distant recurrence and death. After ILRR, 185 events occurred with respect to the PFS end point, and 171 patients died. Primary nodal status, tumor grade, estrogen receptor status of the primary tumor, and length of the disease-free interval (DFI) until the time of the ILRR had a significant prognostic impact. CONCLUSION: Determinants of prognosis after the ILRR should be taken into account for designing future risk-adapted clinical studies for these patients. Risk strata can be defined by a simple classification scheme based on primary nodal status and DFI.

Quality of life in goserelin-treated versus cyclophosphamide + methotrexate + fluorouracil-treated premenopausal and perimenopausal patients with node-positive, early breast cancer: the Zoladex Early Breast Cancer Research Association Trialists Group.

PURPOSE: To compare quality of life (QoL) in premenopausal and perimenopausal patients with node-positive, early breast cancer treated with the endocrine agent goserelin (Zoladex; AstraZeneca Pharmaceuticals LP, Wilmington, DE) or cyclophosphamide + methotrexate + fluorouracil (CMF). PATIENTS AND METHODS: Patients from 86 centers worldwide were randomly assigned to receive either goserelin (3.6 mg every 28 days for 2 years; n = 514) or CMF (six 28-day cycles; n = 496), and were included in the QoL study. QoL was assessed using a self-administered patient questionnaire that consisted of 39 items from the Rotterdam Symptom Checklist, including dimensions evaluating physical and psychological symptom distress, activities of daily living, hormonal effects, and an assessment of overall QoL. RESULTS: Early benefits were noted during months 3 to 6 of treatment, for goserelin compared with CMF. Significant differences were found for changes in overall QoL (eg, 6.96 +/- 0.88 v 0.69 +/- 0.92 at 6 months; P <.0001) and for physical symptom distress, activity levels, and "effort to cope with illness" dimensions. At 1, 2, and 3 years, there were no significant differences in overall QoL or specific QoL dimensions. Scores for hormonal symptoms were worse with goserelin during the 2-year goserelin treatment period; however, this trend was reversed at 3 years. CONCLUSION: Goserelin offers improved overall QoL during the first 6 months of therapy compared with CMF chemotherapy in premenopausal and perimenopausal patients with early breast cancer. Coupled with equivalent efficacy in estrogen receptor-positive patients, these data support the use of goserelin as an alternative to CMF in premenopausal and perimenopausal patients with estrogen receptor-positive, node-positive early breast cancer.

Randomized 2 x 2 trial evaluating hormonal treatment and the duration of chemotherapy in node-positive breast cancer patients: an update based on 10 years' follow-up. German Breast Cancer Study Group.

PURPOSE: In 1984, the German Breast Cancer Study Group started a multicenter randomized trial to compare six versus three cycles of cyclophosphamide, methotrexate, and fluorouracil (CMF) starting perioperatively and to investigate the additional effect of tamoxifen as adjuvant treatment in node-positive breast cancer patients treated with mastectomy. PATIENTS AND METHODS: From 1984 to 1989, 473 patients were randomized from 41 institutions. After a median follow-up of approximately 10 years for overall survival (OS) and 9 years for event-free survival (EFS), the treatment groups were compared with respect to OS and EFS. Results based on a median follow-up of 56 months have been published earlier. RESULTS: Estimated cumulative locoregional incidence rate after 10 years was 19.9%; the corresponding rate of distant recurrences was 41.3%. Concerning duration of chemotherapy, we did not find any difference between six and three cycles of CMF (EFS: relative risk [RR] in multivariate analysis = 0.95; 95% confidence interval [CI], 0.74 to 1.21 OS: RR = 0.90; 95% CI, = 0.69 to 1.18). Treatment with tamoxifen resulted in an improvement in outcome (EFS: RR = 0.81; 95% CI, 0.61 to 1.07, OS: RR = 0.74; 95% CI, 0.55 to 1.0) although it proved not significant. Number of positive lymph nodes and progesterone receptor were the dominant prognostic factors. CONCLUSION: In this study, we observed some tendency in favor of hormonal treatment, which is in agreement with the literature. Concerning duration of chemotherapy, the results of this study provide further evidence that a reduction to three cycles of CMF is possible without increasing the risk of recurrence or death. For a definitive conclusion, however, further investigations are required.

Randomized 2 x 2 trial evaluating hormonal treatment and the duration of chemotherapy in node-positive breast cancer patients. German Breast Cancer Study Group.

PURPOSE: In 1984, the German Breast Cancer Study Group (GBSG) started a multicenter randomized clinical trial to compare the effectiveness of three versus six cycles of 500 mg/m2 cyclophosphamide, 40 mg/m2 methotrexate, and 600 mg/m2 fluorouracil (CMF) on day 1 and 8 starting perioperatively with or without tamoxifen (TAM) (3 x 10 mg/d for 2 years). The aim of the trial was to compare recurrence-free and overall survival between the different treatment modalities. PATIENTS AND METHODS: During 5 years, 41 institutions randomized 473 patients (3 x CMF: 145; 3 x CMF + TAM: 93; 6 x CMF 144; 6 x CMF + TAM: 91). Until March 31, 1992, median follow-up time was 56 months with 197 events for disease-free survival and 116 deaths observed. This provides a power of approximately 80% to detect a potential treatment difference corresponding to a relative risk (RR) of 0.67 for recurrence-free survival. Treatment modalities and various patient characteristics were evaluated by means of a multivariate Cox regression analysis. RESULTS: No significant difference in recurrence-free survival was observed with respect to hormonal therapy (RR = 0.75 TAM v no TAM; 95% confidence interval [CI], 0.54 to 1.04; P = .08) as well as duration of chemotherapy (RR = 0.90 of 6 x CMF v 3 x CMF; 95% CI, 0.67 to 1.19; P = .45). Similar results were obtained for overall survival. The multivariate analysis revealed a significant prognostic impact of the number of positive lymph nodes and the progesterone receptor level on recurrence-free survival. Compliance with chemotherapy within the range of 85% to 115% of the target dose was achieved in 94% and 78% of the patients randomized to 3 x CMF and 6 x CMF, respectively. Sufficient compliance with TAM was reported for 141 patients (93%). CONCLUSION: At this stage of follow-up, six courses of CMF are not superior to three courses with respect to recurrence-free survival.

Tumor-cell number and viability as quality and efficacy parameters of autologous virus-modified cancer vaccines in patients with breast or ovarian cancer.

PURPOSE: We investigated quality and efficacy criteria of an autologous, physically and immunologically purified, Newcastle disease virus (NDV)-modified, irradiated tumor-cell vaccine (ATV-NDV) by analyzing three independent cohorts (a through c) of patients vaccinated between 1991 and 1995. MATERIALS AND METHODS: Included were 63 patients with primary breast cancer (a), 27 with metastatic pretreated breast cancer (b), and 31 with metastatic pretreated ovarian cancer (c). In addition to vaccine, cohorts b and c received nonspecific immunotherapy as supportive treatment. After cryoconservation and purification, the vaccines varied in applied numbers of viable cells and dead cell contaminations. We retrospectively hypothesized that an immunogenic vaccine should contain at least 1.5 x 10(6) viable tumor cells and viability should be at least 33%. Each cohort was thus divided into two groups; one that received vaccine type A (A), fulfilling both criteria; and the other type B (B), missing one or both criteria. RESULTS: Conventional prognostic factors were wall balanced between A and B in cohorts a and c. In cohort a, there was a benefit in survival (P = .026) and disease-free survival (P = .089) for A. In addition, in cohort a, the relative risk of dying in the group that received A as compared with B was 0.2 (univariate Cox model). There were also survival trends in favor of A versus B (P = .18 and P = .09, respectively) in cohorts b and c, with relative risks of 0.5 and 0.42, respectively. In cohort b, the survival benefit could not be ascribed to vaccine quality alone, because of prognostic imbalance in favor of A. CONCLUSION: In cohort c, like in cohort a, the survival benefit for A may be ascribed to the ATV-NDV vaccine quality, since prognostic factors were not biased. This could imply clinical effectivity in breast and ovarian cancer with ATV-NDV high-quality vaccine. Furthermore, the data provide clinically relevant information for standardization and quality control of autologous tumor-cell vaccines. A randomized study is urgently needed.

Goserelin versus cyclophosphamide, methotrexate, and fluorouracil as adjuvant therapy in premenopausal patients with node-positive breast cancer: The Zoladex Early Breast Cancer Research Association Study.

PURPOSE: Current adjuvant therapies have improved survival for premenopausal patients with breast cancer but may have short-term toxic effects and long-term effects associated with premature menopause. PATIENTS AND METHODS: The Zoladex Early Breast Cancer Research Association study assessed the efficacy and tolerability of goserelin (3.6 mg every 28 days for 2 years; n = 817) versus cyclophosphamide, methotrexate, and fluorouracil (CMF) chemotherapy (six 28-day cycles; n = 823) for adjuvant treatment in premenopausal patients with node-positive breast cancer. RESULTS: Analysis was performed when 684 events had been achieved, and the median follow-up was 6 years. A significant interaction between treatment and estrogen receptor (ER) status was found (P =.0016). In ER-positive patients (approximately 74%), goserelin was equivalent to CMF for disease-free survival (DFS) (hazard ratio [HR], 1.01; 95% confidence interval [CI], 0.84 to 1.20). In ER-negative patients, goserelin was inferior to CMF for DFS (HR, 1.76; 95% CI, 1.27 to 2.44). Amenorrhea occurred in more than 95% of goserelin patients by 6 months versus 58.6% of CMF patients. Menses returned in most goserelin patients after therapy stopped, whereas amenorrhea was generally permanent in CMF patients (22.6% v 76.9% amenorrheic at 3 years). Chemotherapy-related side effects such as nausea/vomiting, alopecia, and infection were higher with CMF than with goserelin during CMF treatment. Side effects related to estrogen suppression were initially higher with goserelin, but when goserelin treatment stopped, reduced to a level below that observed in the CMF group. CONCLUSION: Goserelin offers an effective, well-tolerated alternative to CMF in premenopausal patients with ER-positive and node-positive early breast cancer.

Building multivariable regression models with continuous covariates in clinical epidemiology--with an emphasis on fractional polynomials.

OBJECTIVES: In fitting regression models, data analysts must often choose a model based on several candidate predictor variables which may influence the outcome. Most analysts either assume a linear relationship for continuous predictors, or categorize them and postulate step functions. By contrast, we propose to model possible non-linearity in the relationship between the outcome and several continuous predictors by estimating smooth functions of the predictors. We aim to demonstrate that a structured approach based on fractional polynomials can give a broadly satisfactory practical solution to the problem of simultaneously identifying a subset of 'important' predictors and determining the functional relationship for continuous predictors. METHODS: We discuss the background, and motivate and describe the multivariable fractional polynomial (MFP) approach to model selection from data which include continuous and categorical predictors. We compare our results with those from other approaches in examples. We present a small simulation study to compare the functional form of the relationship obtained by fitting fractional polynomials and splines to a single predictor variable. RESULTS: We illustrate the advantages of the MFP approach over standard techniques of model construction in two real example datasets analyzed with logistic and Cox regression models, respectively. In the simulation study, fractional polynomial models had lower mean square error and more realistic behaviour than comparable spline models. CONCLUSIONS: In many practical situations, the MFP approach can satisfy the aim of finding models that fit the data well and also are simple, interpretable and potentially transportable to other settings.

Long-term prognosis of breast cancer patients with 10 or more positive lymph nodes treated with CMF.

The purpose of this investigation was to study the long-term prognosis of breast cancer patients with 10 or more positive lymph nodes after conventional chemotherapy treatment with cyclophosphamide, methotrexate and 5-fluorouracil (CMF). Between 1984 and 1989, 1048 node-positive patients were treated with CMF in two separate trials conducted by the German Breast Cancer Study Group (GBSG). Subgroups either received radiotherapy or tamoxifen in addition. In this study, long-term prognosis in the subgroup of 141 patients with 10 or more positive lymph nodes was investigated. Univariate and multivariate Cox models were used to evaluate the effect of prognostic factors on event-free survival (EFS) and overall survival (OS). Both univariate and multivariate analyses revealed the progesterone receptor (PR) status as the dominating prognostic factor for both EFS and OS, resulting in a strongly increased risk of more than 2-fold for receptor-negative patients. A large number of positive lymph nodes also affected the prognosis for EFS. In univariate analysis, the degree of lymph node involvement (i.e. percentage of positive nodes out of all examined nodes), oestrogen status (ER) status, and tumour grade also showed significant effects. To conclude, the prognosis in the subgroup of patients with 10 or more positive lymph nodes is heterogeneous. Some surprisingly high survival rates have been observed in case series of breast cancer patients treated with high-dose chemotherapy which may be explained by patient selection. From the usual factors investigated in this study, the PR status showed the strongest effect, and, at least this factor should be taken into account in the design and analysis of trials for breast cancer patients with a poor prognosis.

The effect of oral contraceptive use on the prognosis of node positive breast cancer patients. German Breast Cancer Study Group.

The effect of oral contraceptive (OC) use as a risk factor for breast cancer was recently assessed in a large meta-analysis, but currently available data on the prognostic effect are still insufficient. We investigated the relationship between OC use and standard prognostic factors and the effect of OC use on recurrence-free survival (RFS) and overall survival (OS) in 422 premenopausal pT1a-3aN + M0 patients from two trials of the German Breast Cancer Study Group (GBSG). 137 patients (32.5%) were OC users. They were younger on average (mean age 41.5 years versus 45 years for non-OC users) and the percentage of patients with smaller tumours was higher in the group of OC users. Based on 163 events for RFS and 103 events for OS, no significant effect of OC use on RFS and OS could be demonstrated in univariate and multivariate analyses. In our study of node positive breast cancer cases, OC users were younger and had smaller tumours. This may be an effect of earlier detection of breast cancer, but OC users did not have a better prognosis, both before and after adjustment for tumour size and other prognostic factors.

Eight-year results of a prospective non-randomised study on therapy of small breast cancer. The German Breast Cancer Study Group (GBSG).

In this report, the results of the first controlled clinical trial on breast cancer in Germany, begun in 1983, are presented after a median follow-up of 8 years. Four-year results have been previously published. In pT1 N0 M0 breast cancer, mastectomy as the standard treatment was to be compared with tumorectomy plus radiotherapy to the remaining breast tissue. The study design, originally planned as a comprehensive cohort study including randomised and non-randomised patients, had to be changed into a prospective observation study due to the low randomisation rate. 1036 out of 1119 recruited patients were evaluable. After a median follow-up of 97 months, 237 events (local recurrence, regional recurrence, distant metastases, contralateral breast cancer or death of the patient without previous recurrence) occurred. With the exception of death without recurrence, the events were evenly distributed among the two treatment groups. The 8-year local recurrence rate of the whole patient population is 8.8%. Out of all prognostic factors examined, only tumour size and grade had a significant influence on recurrent disease. Event-free survival decreased in cases with 'uncertain' tumour margins, whereas the width of the margin has no influence on disease recurrence. Based on 151 deaths observed so far, there was no significant difference in overall survival between the two treatment groups. The 8-year results of this study are in accordance with the 4-year results reported previously and with those of other breast-conserving treatment trials. There was no significant difference between the two treatment groups with regard to event-free and overall survival. Incomplete tumorectomy had a negative influence on recurrence.

Survival analyses from the ZEBRA study. goserelin (Zoladex) versus CMF in premenopausal women with node-positive breast cancer.

The Zoladex Early Breast Cancer Research Association (ZEBRA) trial compared the efficacy and tolerability of goserelin (Zoladex) with cyclophosphamide, methotrexate and 5-fluorouracil (CMF) chemotherapy in pre-/perimenopausal women with node-positive early breast cancer. The results of disease-free survival (DFS) analyses have already been published. Here we present an update including data on overall survival (OS) from the ZEBRA trial at a median follow-up of 7.3 years. In patients with oestrogen receptor (ER)-positive tumours, non-inferiority of goserelin versus CMF for OS was shown; goserelin was again shown to be equivalent to CMF for DFS. This updated analysis has demonstrated that the two treatments are also equivalent for distant disease-free survival (DDFS). In patients with ER-negative disease, goserelin was inferior to CMF for DFS, DDFS and OS. This follow-up analysis confirms the previously reported outcomes from the ZEBRA trial and demonstrates that goserelin offers an effective alternative to CMF chemotherapy for adjuvant therapy of premenopausal patients with ER-positive, node-positive early breast cancer.

Radiation therapy after breast-conserving surgery; first results of a randomised clinical trial in patients with low risk of recurrence.

To study the role of radiotherapy and tamoxifen after breast-conserving surgery (BCS) in patients with a favourable prognosis, a clinical trial was initiated by the German Breast Cancer Study Group. Between 1991 and 1998, 361 patients (pT1pN0M0, aged 45-75 years, receptor positive, grade I-II) were randomised to radiotherapy (yes/no) and tamoxifen for 2 years (yes/no) in a 2x2 factorial design; the exclusion of seven centres (14 patients) left 347 patients in the analysis. After a median follow-up of 5.9 years, 77 events concerning event-free survival have been observed. Since a strong interactive effect between radiotherapy and tamoxifen has been established, the results are presented in terms of the treatment effects for all four treatment groups separately. Mainly due to the presence of local recurrences, the event rate was about three times higher in the group with BCS only than in the other three groups. No difference could be established between the four treatment groups for distant disease-free survival rates. It is concluded that even in patients with a favourable prognosis, the avoidance of radiotherapy and tamoxifen after BCS increases the rate of local recurrences substantially.

Development and validation of diagnostic scores for atopic dermatitis incorporating criteria of data quality and practical usefulness.

No objective classification criteria for atopic dermatitis (AD) exist. Therefore the diagnosis is usually based on many variables including anamnestic, clinical, and laboratory findings. The aim of this study was to develop and validate a diagnostic score to standardize the diagnosis of atopic skin diathesis for clinical and epidemiological studies. In two separate studies, each consisting of cases and controls, 19 atopic binary features were examined by two independent experienced physicians and these features were classified as "objective," respectively, "subjective." On the basis of these criteria and two additional laboratory measures, we developed a score with high discriminative ability, using the logistic regression model and backward elimination. Ignoring "subjective" variables and the two laboratory measures, two additional models were built that had a worse fit in the original data, but still yielded high estimates of sensitivity (approximately 90%) and specificity (approximately 96%). Using the same data as for the model building, it is well known that these estimates are too optimistic. The validation study allows us to obtain unbiased estimates of sensitivity and specificity for the different scores and to investigate the influence of data quality-here given by the assessment of the reproducibility of the features (objective and subjective)-on the usefulness of diagnostic scores. The results of the validation study show that we developed simple and easy-to-use scores offering a base for a broad practical use in epidemiological and clinical research. In addition, we demonstrate that the criteria classified as "subjective" have no influence on the case-control status in the validation study.

The use of fractional polynomials to model continuous risk variables in epidemiology.

BACKGROUND: The traditional method of analysing continuous or ordinal risk factors by categorization or linear models may be improved. METHODS: We propose an approach based on transformation and fractional polynomials which yields simple regression models with interpretable curves. We suggest a way of presenting the results from such models which involves tabulating the risks estimated from the model at convenient values of the risk factor. We discuss how to incorporate several continuous risk and confounding variables within a single model. The approach is exemplified with data from the Whitehall I study of British Civil Servants. We discuss the approach in relation to categorization and non-parametric regression models. RESULTS: We show that non-linear risk models fit the data better than linear models. We discuss the difficulties introduced by categorization and the advantages of the new approach. CONCLUSIONS: Our approach based on fractional polynomials should be considered as an important alternative to the traditional approaches for the analysis of continuous variables in epidemiological studies.