Meta-analysis without study-specific variance information: Heterogeneity case

The random effects model in meta-analysis is a standard statistical tool often used to analyze the effect sizes of the quantity of interest if there is heterogeneity between studies. In the special case considered here, meta-analytic data contain only the sample means in two treatment arms and the sample sizes, but no sample standard deviation. The statistical comparison between two arms for this case is not possible within the existing meta-analytic inference framework. Therefore, the main objective of this paper is to estimate the overall mean difference and associated variances, the between-study variance and the within-study variance, as specified as the important elements in the random effects model. These estimators are obtained using maximum likelihood estimation. The standard errors of the estimators and a quantification of the degree of heterogeneity are also investigated. A measure of heterogeneity is suggested which adjusts the original suggested measure of Higgins’ I2 for within study sample size. The performance of the proposed estimators is evaluated using simulations. It can be concluded that all estimated means converged to their associated true parameter values, and its standard errors tended to be small if the number of the studies involved in the meta-analysis was large. The proposed estimators could be favorably applied in a meta-analysis on comparing two surgeries for asymptomatic congenital lung malformations in young children.

Statistical methodology for estimating the mean difference in a meta-analysis without study-specific variance information

Statistical inference for analyzing the results from several independent studies on the same quantity of interest has been investigated frequently in recent decades. Typically, any meta-analytic inference requires that the quantity of interest is available from each study together with an estimate of its variability. The current work is motivated by a meta-analysis on comparing two treatments (thoracoscopic and open) of congenital lung malformations in young children. Quantities of interest include continuous end-points such as length of operation or number of chest tube days. As studies only report mean values (and no standard errors or confidence intervals), the question arises how meta-analytic inference can be developed. We suggest two methods to estimate study-specific variances in such a meta-analysis, where only sample means and sample sizes are available in the treatment arms. A general likelihood ratio test is derived for testing equality of variances in two groups. By means of simulation studies, the bias and estimated standard error of the overall mean difference from both methodologies are evaluated and compared with two existing approaches: complete study analysis only and partial variance information. The performance of the test is evaluated in terms of type I error. Additionally, we illustrate these methods in the meta-analysis on comparing thoracoscopic and open surgery for congenital lung malformations and in a meta-analysis on the change in renal function after kidney donation. Copyright © 2017 John Wiley & Sons, Ltd.

Inter-rater reliability of the QuIS as an assessment of the quality of staff-inpatient interactions

BACKGROUND

Recent studies of the quality of in-hospital care have used the Quality of Interaction Schedule (QuIS) to rate interactions observed between staff and inpatients in a variety of ward conditions. The QuIS was developed and evaluated in nursing and residential care. We set out to develop methodology for summarising information from inter-rater reliability studies of the QuIS in the acute hospital setting.

METHODS

Staff-inpatient interactions were rated by trained staff observing care delivered during two-hour observation periods. Anticipating the possibility of the quality of care varying depending on ward conditions, we selected wards and times of day to reflect the variety of daytime care delivered to patients. We estimated inter-rater reliability using weighted kappa, κ _w , combined over observation periods to produce an overall, summary estimate, [Formula: see text]. Weighting schemes putting different emphasis on the severity of misclassification between QuIS categories were compared, as were different methods of combining observation period specific estimates.

RESULTS

Estimated [Formula: see text] did not vary greatly depending on the weighting scheme employed, but we found simple averaging of estimates across observation periods to produce a higher value of inter-rater reliability due to over-weighting observation periods with fewest interactions.

CONCLUSIONS

We recommend that researchers evaluating the inter-rater reliability of the QuIS by observing staff-inpatient interactions during observation periods representing the variety of ward conditions in which care takes place, should summarise inter-rater reliability by κ _w , weighted according to our scheme A4. Observation period specific estimates should be combined into an overall, single summary statistic [Formula: see text], using a random effects approach, with [Formula: see text], to be interpreted as the mean of the distribution of κ _w across the variety of ward conditions. We draw attention to issues in the analysis and interpretation of inter-rater reliability studies incorporating distinct phases of data collection that may generalise more widely.

Statistical methodology for age-adjustment of the GH-2000 score detecting growth hormone misuse

Background

The GH-2000 score has been developed as a powerful and unique technique for the detection of growth hormone misuse by sportsmen and women. The score depends upon the measurement of two growth hormone (GH) sensitive markers, insulin-like growth factor-I (IGF-I) and the amino-terminal pro-peptide of type III collagen (P-III-NP). With the collection and establishment of an increasingly large database it has become apparent that the score shows a positive age effect in the male athlete population, which could potentially place older male athletes at a disadvantage.

Methods

We have used results from residual analysis of the general linear model to show that the residual of the GH-2000 score when regressed on the mean-age centred age is an appropriate way to proceed to correct this bias. As six GH-2000 scores are possible depending on the assays used for determining IGF-I and P-III-NP, methodology had to be explored for including six different age effects into a unique residual. Meta-analytic techniques have been utilized to find a summary age effect.

Results

The age-adjusted GH-2000 score, a form of residual, has similar mean and variance as the original GH-2000 score and, hence, the developed decision limits show negligible change when compared to the decision limits based on the original score. We also show that any further scale-transformation will not change the adjusted score. Hence the suggested adjustment is optimal for the given data. The summary age effect is homogeneous across the six scores, and so the generic adjustment of the GH-2000 score formula is justified.

Conclusions

A final revised GH-2000 score formula is provided which is independent of the age of the athlete under consideration.

Revisiting proportion estimators

Proportion estimators are quite frequently used in many application areas. The conventional proportion estimator (number of events divided by sample size) encounters a number of problems when the data are sparse as will be demonstrated in various settings. The problem of estimating its variance when sample sizes become small is rarely addressed in a satisfying framework. Specifically, we have in mind applications like the weighted risk difference in multicenter trials or stratifying risk ratio estimators (to adjust for potential confounders) in epidemiological studies. It is suggested to estimate p using the parametric family (see PDF for character) and p(1 - p) using (see PDF for character), where (see PDF for character). We investigate the estimation problem of choosing c 0 from various perspectives including minimizing the average mean squared error of (see PDF for character), average bias and average mean squared error of (see PDF for character). The optimal value of c for minimizing the average mean squared error of (see PDF for character) is found to be independent of n and equals c = 1. The optimal value of c for minimizing the average mean squared error of (see PDF for character) is found to be dependent of n with limiting value c = 0.833. This might justifiy to use a near-optimal value of c = 1 in practice which also turns out to be beneficial when constructing confidence intervals of the form (see PDF for character).

Novel markers to detect recombinant human insulin-like growth factor-I (rhIGF-I)/rhIGF binding protein-3 (rhIGFBP-3) misuse in athletes

Insulin-like growth factor-I (IGF-I) is abused by elite athletes for its metabolic and anabolic effects. We have previously shown that it is possible to detect IGF-I misuse by measuring serum IGF-I and procollagen type III amino-terminal propeptide (P-III-NP) but a pilot study suggested measuring IGF-II, IGF binding protein-2 (IGFBP-2) and acid-labile subunit (ALS) may improve the detection of IGF-I administration. The aim of the study was to assess this in a randomized controlled trial. Twenty-six female and 30 male recreational athletes were randomized to 28 days' treatment with placebo or recombinant human (rh)IGF-I/rhIGF binding protein-3 (IGFBP-3) complex (30 mg/day or 60 mg/day), followed by 56 days' washout. IGF-II, IGFBP-2 and ALS (women only) were measured using commercial immunoassays. IGFBP-2 increased and IGF-II decreased in response to both low and high dose rhIGF-I/rhIGFBP-3 in both women and men while ALS decreased in women in response to high dose rhIGF-I/rhIGFBP-3. Two days after discontinuing treatment, significant differences remained between the three treatment groups in IGFBP-2 and IGF-II, but not ALS. Thereafter there were no significant differences between the three treatment groups in any of the markers. Combining IGF-I with IGF-II and/or IGFBP-2 improved the performance of the test to detect rhIGF-I/rhIGFBP-3 administration in both women and men. Copyright © 2016 John Wiley & Sons, Ltd.

A flexible ratio regression approach for zero-truncated capture-recapture counts

Capture-recapture methods are used to estimate the size of a population of interest which is only partially observed. In such studies, each member of the population carries a count of the number of times it has been identified during the observational period. In real-life applications, only positive counts are recorded, and we get a truncated at zero-observed distribution. We need to use the truncated count distribution to estimate the number of unobserved units. We consider ratios of neighboring count probabilities, estimated by ratios of observed frequencies, regardless of whether we have a zero-truncated or an untruncated distribution. Rocchetti et al. (2011) have shown that, for densities in the Katz family, these ratios can be modeled by a regression approach, and Rocchetti et al. (2014) have specialized the approach to the beta-binomial distribution. Once the regression model has been estimated, the unobserved frequency of zero counts can be simply derived. The guiding principle is that it is often easier to find an appropriate regression model than a proper model for the count distribution. However, a full analysis of the connection between the regression model and the associated count distribution has been missing. In this manuscript, we fill the gap and show that the regression model approach leads, under general conditions, to a valid count distribution; we also consider a wider class of regression models, based on fractional polynomials. The proposed approach is illustrated by analyzing various empirical applications, and by means of a simulation study.

Application of the Vertex Exchange Method to estimate a semi-parametric mixture model for the MIC density of Escherichia coli isolates tested for susceptibility against ampicillin

In the last decades, considerable attention has been paid to the collection of antimicrobial resistance data, with the aim of monitoring non-wild-type isolates. This monitoring is performed based on minimum inhibition concentration (MIC) values, which are collected through dilution experiments. We present a semi-parametric mixture model to estimate the entire MIC density on the continuous scale. The parametric first component is extended with a non-parametric second component and a new back-fitting algorithm, based on the Vertex Exchange Method, is proposed. Our data example shows how to estimate the MIC density for Escherichia coli tested for ampicillin and how to use this estimate for model-based classification. A simulation study was performed, showing the promising behavior of the new method, both in terms of density estimation as well as classification.

The Effects of Recombinant Human Insulin-Like Growth Factor-I/Insulin-Like Growth Factor Binding Protein-3 Administration on Body Composition and Physical Fitness in Recreational Athletes

CONTEXT

IGF-I is thought to mediate many of the anabolic actions of GH, and there are anecdotal reports that IGF-I is misused by elite athletes. There is no published evidence regarding the effects of IGF-I administration on athletic performance.

OBJECTIVE

The objective of the study was to investigate the effects of IGF-I administration on body composition and physical fitness in recreational athletes.

DESIGN AND SETTING

This was a randomized, double-blind, placebo-controlled recombinant human (rh) IGF-I/rhIGF binding protein (IGFBP)-3 administration study at Southampton General Hospital (Southampton, United Kingdom).

PARTICIPANTS

Fifty-six recreational athletes (30 men, 26 women) participated in the study.

INTERVENTION

Participants were randomly assigned to receive placebo, low-dose rhIGF-I/rhIGFBP-3 (30 mg/d), or high dose rhIGF-I/rhIGFBP-3 (60 mg/d) for 28 days. Body composition (assessed by dual energy x-ray absorptiometry) and cardiorespiratory fitness (assessed by incremental treadmill test) were measured before and immediately after treatment. Within-individual changes after treatment were analyzed using paired t tests.

RESULTS

There were no significant changes in body fat mass or lean body mass in women or men after the administration of the rhIGF-I/rhIGFBP-3 complex. There was a significant increase in maximal oxygen consumption (VO2 max) after treatment. When women and men and low- and high-dose treatment groups were combined, mean VO2 max increased by approximately 7% (P = .001). No significant change in VO2 max was observed in the placebo group.

CONCLUSIONS

rhIGF-I/rhIGFBP-3 administration for 28 days improves aerobic performance in recreational athletes, but there are no effects on body composition.

Correspondence: Some general points regarding Ledberg and Wennberg, BMC Medical Research Methodology 2014 April 27;14:58

The purpose of this note is to contribute some general points on a recent paper by Ledberg and Wennberg (BMC Med Res Meth 14:58, 2014) which need to be rectified. They advocate the capture-removal estimator. First, we will discuss drawbacks of this estimator in comparison to the Lincoln-Petersen estimator. Second, we show that their evaluation of the Chao estimator is flawed. We conclude that some statements in Ledberg and Wennberg with respect to Chao’s estimator and removal estimation need to be taken with great caution.

The development of decision limits for the GH-2000 detection methodology using additional insulin-like growth factor-I and amino-terminal pro-peptide of type III collagen assays

The GH-2000 and GH-2004 projects have developed a method for detecting GH misuse based on measuring insulin-like growth factor-I (IGF-I) and the amino-terminal pro-peptide of type III collagen (P-III-NP). The objectives were to analyze more samples from elite athletes to improve the reliability of the decision limit estimates, to evaluate whether the existing decision limits needed revision, and to validate further non-radioisotopic assays for these markers. The study included 998 male and 931 female elite athletes. Blood samples were collected according to World Anti-Doping Agency (WADA) guidelines at various sporting events including the 2011 International Association of Athletics Federations (IAAF) World Athletics Championships in Daegu, South Korea. IGF-I was measured by the Immunotech A15729 IGF-I IRMA, the Immunodiagnostic Systems iSYS IGF-I assay and a recently developed mass spectrometry (LC-MS/MS) method. P-III-NP was measured by the Cisbio RIA-gnost P-III-P, Orion UniQ™ PIIINP RIA and Siemens ADVIA Centaur P-III-NP assays. The GH-2000 score decision limits were developed using existing statistical techniques. Decision limits were determined using a specificity of 99.99% and an allowance for uncertainty because of the finite sample size. The revised Immunotech IGF-I - Orion P-III-NP assay combination decision limit did not change significantly following the addition of the new samples. The new decision limits are applied to currently available non-radioisotopic assays to measure IGF-I and P-III-NP in elite athletes, which should allow wider flexibility to implement the GH-2000 marker test for GH misuse while providing some resilience against manufacturer withdrawal or change of assays.

Extending the Lincoln-Petersen estimator for multiple identifications in one source

The Lincoln-Petersen estimator is one of the most popular estimators used in capture-recapture studies. It was developed for a sampling situation in which two sources independently identify members of a target population. For each of the two sources, it is determined if a unit of the target population is identified or not. This leads to a 2 × 2 table with frequencies f11 ,f10 ,f01 ,f00 indicating the number of units identified by both sources, by the first but not the second source, by the second but not the first source and not identified by any of the two sources, respectively. However, f00 is unobserved so that the 2 × 2 table is incomplete and the Lincoln-Petersen estimator provides an estimate for f00 . In this paper, we consider a generalization of this situation for which one source provides not only a binary identification outcome but also a count outcome of how many times a unit has been identified. Using a truncated Poisson count model, truncating multiple identifications larger than two, we propose a maximum likelihood estimator of the Poisson parameter and, ultimately, of the population size. This estimator shows benefits, in comparison with Lincoln-Petersen's, in terms of bias and efficiency. It is possible to test the homogeneity assumption that is not testable in the Lincoln-Petersen framework. The approach is applied to surveillance data on syphilis from Izmir, Turkey.