Cédric Mahé

Heterogeneity can be a major component of meta-analyses and by virtue of that fact warrants investigation. Classic analysis methods, such as meta-regression, are used to explore the sources of heterogeneity. However, it may be difficult to apply such a method in complex cases or in the absence of an a priori hypothesis. This paper presents a graphical method to identify trials, groups of trials or groups of patients that are sources of heterogeneity. The contribution of these trials to the overall result can also be evaluated with this method. Each trial is represented by a dot on a 2D graph. The X-axis represents the contribution of the trial to the overall Cochran Q-test for heterogeneity. The Y-axis represents the influence of the trial, defined as the standardized squared difference between the treatment effects estimated with and without the trial. This approach has been applied to data from the Meta-Analysis of Chemotherapy in Head and Neck Cancer (MACH-NC) comprising 10,850 patients in 65 randomized trials. The graphical method allowed us to identify trials that contributed considerably to the overall heterogeneity and had a strong influence on the overall result. It also provided useful information for the interpretation of heterogeneity in this meta-analysis. The proposed graphical method identifies trials that account for most of the heterogeneity without having to explore all possible sources of heterogeneity by subgroup analyses. This method can also be applied to identify types of patients that explain heterogeneity in the treatment effect.

BACKGROUND: Cervical-cancer screening strategies that involve the use of conventional cytology and require multiple visits have been impractical in developing countries. METHODS: We used computer-based models to assess the cost-effectiveness of a variety of cervical-cancer screening strategies in India, Kenya, Peru, South Africa, and Thailand. Primary data were combined with data from the literature to estimate age-specific incidence and mortality rates for cancer and the effectiveness of screening for and treatment of precancerous lesions. We assessed the direct medical, time, and program-related costs of strategies that differed according to screening test, targeted age and frequency, and number of clinic visits required. Single-visit strategies involved the assumption that screening and treatment could be provided in the same day. Outcomes included the lifetime risk of cancer, years of life saved, lifetime costs, and cost-effectiveness ratios (cost per year of life saved). RESULTS: The most cost-effective strategies were those that required the fewest visits, resulting in improved follow-up testing and treatment. Screening women once in their lifetime, at the age of 35 years, with a one-visit or two-visit screening strategy involving visual inspection of the cervix with acetic acid or DNA testing for human papillomavirus (HPV) in cervical cell samples, reduced the lifetime risk of cancer by approximately 25 to 36 percent, and cost less than 500 dollars per year of life saved. Relative cancer risk declined by an additional 40 percent with two screenings (at 35 and 40 years of age), resulting in a cost per year of life saved that was less than each country's per capita gross domestic product--a very cost-effective result, according to the Commission on Macroeconomics and Health. CONCLUSIONS: Cervical-cancer screening strategies incorporating visual inspection of the cervix with acetic acid or DNA testing for HPV in one or two clinical visits are cost-effective alternatives to conventional three-visit cytology-based screening programs in resource-poor settings.

OBJECTIVES: To describe the progression times of HIV-1 infection from seroconversion to AIDS and to death, and time from first developing AIDS to death in rural Uganda. Also, to describe the proportion of individuals within the cohort dying with AIDS and the CD4 lymphocyte count before death. DESIGN: A prospective, longitudinal, population-based cohort. METHODS: Since 1990, 107 HIV-prevalent cases, 168 incident cases and 235 HIV-seronegative controls have been recruited into a cohort in rural Uganda. Participants are recruited from the general population and they are reviewed routinely every 3 months and at other times when ill. RESULTS: The median time from seroconversion to death was 9.8 years. Age over 40 years at seroconversion was associated with more rapid progression (P < 0.001, log rank test). For the first 4 years of the study, HIV contributed little to the death rates in the HIV incident cases, but after 5 years, the contribution of HIV became greater and was particularly marked in the oldest age group. Survival rates in the cohort were similar to those in the general population. The median time from seroconversion to AIDS was 9.4 years and from AIDS to death was 9.2 months. Of those infected with HIV-1, 80% died with AIDS and 20% had a CD4 count < 10 x 106 cells/l. CONCLUSIONS: Survival with HIV-1 infection is similar in Africa to industrialized countries before the use of antiretroviral therapy; when they do die, many of those in Africa are severely immunosuppressed and most have clinical features of AIDS.

Visual inspection-based screening tests, such as visual inspection with 4% acetic acid (VIA) and with Lugol's iodine (VILI), have been proposed as alternatives to cytology in mass screening programs. To date, there is only limited information on the accuracy of these tests in detecting High-grade Squamous Intraepithelial Lesions (HSIL). Eleven cross-sectional studies involving 56,939 women aged 25-65 years were conducted in Burkina Faso, Congo, Guinea, India, Mali and Niger to evaluate the accuracy of VIA and VILI performed by health workers. A common protocol and questionnaire was used. For final diagnosis, all women were investigated with colposcopy and biopsies were taken when necessary. Data from the studies were pooled to calculate sensitivity, specificity and predictive values of the tests for the detection of HSIL. Of the screened women, 16.1% and 16.4% were positive on examination using, respectively, VIA and VILI; 1,063 were diagnosed with HSIL. The pooled sensitivity, specificity, positive and negative predictive values for VIA were 76.8% (95% CI: 74.2-79.4%), 85.5% (95% CI: 85.2-85.8%), 9.4% (95% CI:8.8-10.8%) and 99.5% (95% CI:99.4-99.6%), respectively. The values were 91.7% (95% CI: 89.7-93.4%), 85.4% (95% CI: 85.1-85.7%), 10.9% (95% CI: 10.2-11.6%) and 99.8% (95% CI:99.7-99.9%), respectively for VILI. The range of sensitivity and specificity for VIA was 56.1-93.9% and 74.2-93.8%, respectively, between studies and were 76.0-97.0 % and 73.0-91.3% for VILI. VILI had a significantly higher sensitivity than VIA in detecting HSIL, but specificity was similar. VILI appears to be a more accurate visual test for use in screening and treatment programs in low-resource settings.

OBJECTIVE: To determine the attributable risk of death due to catheter-related septicemia (CRS) in critically ill patients when taking into account severity of illness during the intensive-care unit (ICU) stay but before CRS. DESIGN: Pairwise-matched (1:2) exposed-unexposed study. SETTING: 10-bed medical-surgical ICU and an 18-bed medical ICU. PATIENTS: Patients admitted to either ICU between January 1, 1990, and December 31, 1995, were eligible. Exposed patients were defined as patients with CRS; unexposed controls were selected according to matching variables. METHODS: Matching variables were diagnosis at ICU admission, length of central catheterization before the infection, McCabe Score, Simplified Acute Physiologic Score (SAPS) II at admission, age, and gender. Severity scores (SAPS II, Organ System Failure Score, Organ Dysfunction and Infection Score, and Logistic Organ Dysfunction System) were calculated four times for each patient: the day of ICU admission, the day of CRS onset, and 3 and 7 days before CRS. Matching was successful for 38 exposed patients. Statistical analysis was based on nonparametric tests for epidemiological data and on Cox's models for the exposed-unexposed study, with adjustment on matching variables and prognostic factors of mortality. RESULTS: CRS complicated 1.17 per 100 ICU admissions during the study period. Twenty (53%) of the CRS cases were associated with septic shock. CRS was associated with a 28% increase in SAPS II. Crude ICU mortality rates from exposed and unexposed patients were 50% and 21%, respectively. CRS remained associated with mortality even when adjusted on other prognostic factors at ICU admission (relative risk [RR], 2.01; 95% confidence interval [CI95], 1.08-3.73; P=.03). However, after adjustment on severity scores calculated between ICU admission and 1 week before CRS, the increased mortality was no longer significant (RR, 1.41; CI95, 0.76-2.61; P=.27). CONCLUSION: CRS is associated with subsequent morbidity and mortality in the ICU, even when adjusted on severity factors at ICU admission. However, after adjustment on severity factors during the ICU stay and before the event, there was only a trend toward CRS-attributable mortality. The evolution of patient severity should be taken into account when evaluating excess mortality induced by nosocomial events in ICU patients.