G*Power (Erdfelder, Faul, & Buchner, 1996) was designed as a general stand-alone power analysis program for statistical tests commonly used in social and behavioral research. G*Power 3 is a major extension of, and improvement over, the previous versions. It runs on widely used computer platforms (i.e., Windows XP, Windows Vista, and Mac OS X 10.4) and covers many different statistical tests of thet, F, and χ2 test families. In addition, it includes power analyses forz tests and some exact tests. G*Power 3 provides improved effect size calculators and graphic options, supports both distribution-based and design-based input modes, and offers all types of power analyses in which users might be interested. Like its predecessors, G*Power 3 is free.

https://link.springer.com/content/pdf/10.3758/BF03193146.pdf

G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences

The extent of heterogeneity in a meta-analysis partly determines the difficulty in drawing overall conclusions. This extent may be measured by estimating a between-study variance, but interpretation is then specific to a particular treatment effect metric. A test for the existence of heterogeneity exists, but depends on the number of studies in the meta-analysis. We develop measures of the impact of heterogeneity on a meta-analysis, from mathematical criteria, that are independent of the number of studies and the treatment effect metric. We derive and propose three suitable statistics: H is the square root of the chi2 heterogeneity statistic divided by its degrees of freedom; R is the ratio of the standard error of the underlying mean from a random effects meta-analysis to the standard error of a fixed effect meta-analytic estimate, and I2 is a transformation of (H) that describes the proportion of total variation in study estimates that is due to heterogeneity. We discuss interpretation, interval estimates and other properties of these measures and examine them in five example data sets showing different amounts of heterogeneity. We conclude that H and I2, which can usually be calculated for published meta-analyses, are particularly useful summaries of the impact of heterogeneity. One or both should be presented in published meta-analyses in preference to the test for heterogeneity.

Quantifying heterogeneity in a meta‐analysis

: The results of observational studies are often disputed because of nonrandom treatment assignment. For example, patients at greater risk may be overrepresented in some treatment group. This paper discusses the central role of propensity scores and balancing scores in the analysis of observational studies. The propensity score is the (estimated) conditional probability of assignment to a particular treatment given a vector of observed covariates. Both large and small sample theory show that adjustment for the scalar propensity score is sufficient to remove bias due to all observed covariates. Applications include: matched sampling on the univariate propensity score which is equal percent bias reducing under more general conditions than required for discriminant matching, multivariate adjustment by subclassification on balancing scores where the same subclasses are used to estimate treatment effects for all outcome variables and in all subpopulations, and visual representation of multivariate adjustment by a two-dimensional plot. (Author)

/pdf/the-central-role-of-the-propensity-score-in-observational-hqdiv01vha.pdf

The central role of the propensity score in observational studies for causal effects

The quality assessment of non-randomized studies is an important component of a thorough meta-analysis of nonrandomized studies. Low quality studies can lead to a distortion of the summary effect estimate. Recent guidelines for the reporting of meta-analyses of observational studies recommend the assessment of the study quality (MOOSE) [1]. In principal, three categories of quality assessments tools are available: scales, simple checklists, or checklists with a summary judgment (for details see Sanderson et al. 2007 [2]). The results of the quality assessment can be used in several ways such as forming inclusion criteria for the meta-analysis, informing a sensitivity analysis or metaregression, weighting studies, or highlighting areas of methodological quality poorly addressed by the included studies [3]. It has been criticized that the use of summary scores involve inherent weighting of component items including items that may not be related to the validity of the study findings [2]. Sanderson et al. [2] recently identified overall 86 tools for assessing the quality of non-randomized studies. Their review "highlighted the lack of a single obvious candidate tool for assessing quality of observational epidemiological studies" [2]. In the field of randomized trials, it has been shown that the choice of quality scale can dramatically influence the interpretation of meta-analyses, and can even reverse conclusions regarding the effectiveness of an intervention [4]. Wells et al. [5] proposed a scale for assessing the quality of published non-randomized studies in meta-analyses,

/pdf/critical-evaluation-of-the-newcastle-ottawa-scale-for-the-1yvsntnojv.pdf

Critical evaluation of the Newcastle-Ottawa scale for the assessment of the quality of nonrandomized studies in meta-analyses

The number of computed tomographic (CT) studies performed is increasing rapidly. Because CT scans involve much higher doses of radiation than plain films, we are seeing a marked increase in radiation exposure in the general population. Epidemiologic studies indicate that the radiation dose from even two or three CT scans results in a detectable increase in the risk of cancer, especially in children. This article summarizes the facts about this form of radiation exposure and the implications for public health.

/pdf/computed-tomography-an-increasing-source-of-radiation-4b5ic1x47d.pdf

Computed Tomography — An Increasing Source of Radiation Exposure

The concepts that case-referent studies provide for the estimation of "relative risk" only if the illness is "rare", and that the rates and risks themselves are inestimable, are overly superficial and restrictve. The ratio of incidence densities (forces of morbidity)-and thereby the instantaneous risk-ratio-is estimable without any rarity-assumption. Long-term risk-ratio can be computed through the coupling of case-referent data on exposure rates for various age-categories with estimates, possibly from the study itself, or the corresponding age-specific incidence-densities for the exposed and nonexposed combined-but again, no rarity-assumption is involved. Such data also provide for the assessment of exposure-specific absolute incidence-rates and risks. Point estimation of the various parameters can be based on simple relationships among them, and in interval estimation it is sufficient simply to couple the point estimate with the value of the chi square statistic used in significance testing.

/pdf/estimability-and-estimation-in-case-referent-studies-1jqtb29dx5.pdf

Estimability and estimation in case-referent studies

BACKGROUND The outcome among patients with clinical stage I cancer that is detected on annual screening using spiral computed tomography (CT) is unknown. METHODS In a large collaborative study, we screened 31,567 asymptomatic persons at risk for lung cancer using low-dose CT from 1993 through 2005, and from 1994 through 2005, 27,456 repeated screenings were performed 7 to 18 months after the previous screening. We estimated the 10-year lung-cancer-specific survival rate among participants with clinical stage I lung cancer that was detected on CT screening and diagnosed by biopsy, regardless of the type of treatment received, and among those who underwent surgical resection of clinical stage I cancer within 1 month. A pathology panel reviewed the surgical specimens obtained from participants who underwent resection. RESULTS Screening resulted in a diagnosis of lung cancer in 484 participants. Of these participants, 412 (85%) had clinical stage I lung cancer, and the estimated 10-year survival rate was 88% in this subgroup (95% confidence interval [CI], 84 to 91). Among the 302 participants with clinical stage I cancer who underwent surgical resection within 1 month after diagnosis, the survival rate was 92% (95% CI, 88 to 95). The 8 participants with clinical stage I cancer who did not receive treatment died within 5 years after diagnosis. CONCLUSIONS Annual spiral CT screening can detect lung cancer that is curable.

Survival of patients with stage I lung cancer detected on CT screening.

Miettinen, O. S. (Harvard School of Public Health, Boston. Mass. 02115). Proportion of disease caused or prevented by a given exposure, trait or intervention. Am J Epidemiol 99: 325-332, 1974.—The structures of two epidemiologic parameters are explored. One, the \"etiologic fraction.\" relates to markers of increased risk, and it is the proportion of disease attributable to the marker and/or to factors associated with it. The other, the \"prevented fraction,\" is the equivalent of this for a marker of reduced risk. It is shown that both parameters depend—in different ways—on the frequency of the marker among cases of the disease, and on the \"standardized morbidity ratio\" for those with the marker. Point estimation of these parameters is often straight-forward, particularly in case-control studies.

Proportion of disease caused or prevented by a given exposure, trait or intervention

In this paper, we examine comparative analysis of rates with a view to each of the usual comparative parameters-rate difference (RD), rate ratio (RR) and odds ratio (OR)-and with particular reference to first principles. For RD and RR we show the prevailing statistical practices to be rather poor. We stress the need for restricted estimation of variance in the chi-square function underlying interval estimation (and also point estimation and hypothesis testing). For RR analysis we propose a chi-square formulation analogous to that for RD and, thus, one which obviates the present practice of log transformation and its associated use of Taylor series approximation of the variance. As for OR analysis, we emphasize that the chi-square function, introduced by Cornfield for unstratified data, and extended by Gart to the case of stratified analysis, is based on the efficient score and thus embodies its optimality properties. We provide simulation results to evince the better performance of the proposed (parameter-constrained) procedures over the traditional ones.

Comparative analysis of two rates

Objective In the Early Lung Cancer Action Project (ELCAP), we found not only solid but also part-solid and nonsolid nodules in patients at both baseline and repeat CT screening for lung cancer We report the frequency and significance of part-solid and nonsolid nodules in comparison with solid nodules Materials and methods We reviewed all instances of a positive finding in patients at baseline (from one to six noncalcified nodules) and annual repeat screenings (from one to six newly detected noncalcified nodules with interim growth) to classify each of the nodules as solid, part-solid, or nonsolid We defined a solid nodule as a nodule that completely obscures the entire lung parenchyma within it Part-solid nodules are those having sections that are solid in this sense, and nonsolid nodules are those with no solid parts Chi-square statistics were used to test for differences in the malignancy rates Results Among the 233 instances of positive results at baseline screening, 44 (19%) involved a part-solid or nonsolid largest nodule (16 part-solid and 28 nonsolid) Among these 44 cases of positive findings, malignancy was diagnosed in 15 (34%) as opposed to a 7% malignancy rate for solid nodules (p = 0000001) The malignancy rate for part-solid nodules was 63% (10/16), and the rate for nonsolid nodules was 18% (5/28) Even after standardizing for nodule size, the malignancy rate was significantly higher for part-solid nodules than for either solid ones (p = 0004) or nonsolid ones (p = 003) The malignancy type in the part-solid or nonsolid nodules was predominantly bronchioloalveolar carcinoma or adenocarcinoma with bronchioloalveolar features, contrasting with other subtypes of adenocarcinoma found in the solid nodules (p = 00001) At annual repeat screenings, only 30 instances of positive test results have been obtained; seven of these involved part-solid or nonsolid nodules Conclusion In CT screening for lung cancer, the detected nodule commonly is either only part-solid or nonsolid, but such a nodule is more likely to be malignant than a solid one, even when nodule size is taken into account

https://www.ajronline.org/doi/pdf/10.2214/ajr.178.5.1781053

Olli S. Miettinen

Papers

Estimability and estimation in case-referent studies

Survival of patients with stage I lung cancer detected on CT screening.

Proportion of disease caused or prevented by a given exposure, trait or intervention

Comparative analysis of two rates

CT Screening for Lung Cancer: Frequency and Significance of Part-Solid and Nonsolid Nodules