Mantel-Haenszel Procedure Research Articles

Lungworm infection and ovine visna–maedi: Real risk factor or a confounding variable?

Serum and fecal samples (n=2081) were taken from 74 commercial meat ovine flocks in Galicia (NW Spain) in order to establish the influence of lungworm infection with protostrongylids and/or Dictyocaulus filaria on visna–maedi (VM) seropositivity. VM was tested by an indirect ELISA test and lungworm larvae were detected by the modified Baermann–Wetzel technique. In addition to lungworm infection (protostrongylids and D. filaria), flock, management and individual factors were included in univariate and multivariate statistical tests.Overall VM seroprevalence was 18.9% and the prevalence by protostrongylids and D. filaria was 11.6% and 10.7%, respectively. Protostrongylid infection was considered as a risk factor for VM with univariate statistical analysis (1.49 significant OR), but neglected in a logistic multivariate regression analysis when the age of the animals was introduced as variable. As VM, protostrongylid infection is a chronic disease with a cumulative pattern over time, so that these two infections are closely related to the age of the animal. A Mantel–Haenszel procedure was applied and lungworm infection was detected as a confounding variable with no real effect on visna–maedi infection.

Detection of Differential Item Functioning

This study analyzes Differential Item Functioning (DIF) with three combined decision rules and compares the results with the variation of the Mantel-Haenszel procedure (vaMH) proposed by Mazor, Clauser, and Hambleton (1994). One decision rule combines the Mantel-Haenszel procedure (MH) with the Breslow-Day test of trend in odds ratio heterogeneity (BDT), having performed the Bonferroni adjustment, as Randall Penfield proposed. The second uses both MH and BDT without the Bonferroni adjustment. The third combines MH with the Breslow-Day test for homogeneity of the odds ratio without the Bonferroni adjustment. The three decision rules yielded satisfactory results, showed similar power, and none of them detected DIF erroneously. The second rule proved to be the most powerful in the presence of nonuniform DIF. Only in the presence of uniform DIF with the smallest difference of difficulty parameters, was there evidence of vaMH’s superiority.

Real and Artificial Differential Item Functioning

The literature in modern test theory on procedures for identifying items with differential item functioning (DIF) among two groups of persons includes the Mantel–Haenszel (MH) procedure. Generally, it is not recognized explicitly that if there is real DIF in some items which favor one group, then as an artifact of this procedure, artificial DIF that favors the other group is induced in the other items. Using the Rasch model for dichotomous responses as a theoretical basis, this article proves that the source of artificial DIF in the MH procedure is that estimates of the person locations are substituted for their unknown values. The article then demonstrates that the formalization of artificial DIF implies mathematically (a) a particular sequential, iterative procedure for detecting items with real DIF and for identifying a set of items that may have no DIF and (b) a resolution of the items with real DIF for quantifying the DIF on the same metric as the items showing no DIF and provides expected value curves and tests of fit for the item for each of the groups. Finally, because the source of artificial DIF in the MH procedure is the substitution of a parameter with its estimate, it is suggested that all procedures that use the substitution of an estimate for a parameter are likely to induce artificial DIF.

Causes of gender DIF on an EFL language test: A multiple-data analysis over nine years

This study tracked gender differential item functioning (DIF) on the English subtest of the Korean College Scholastic Aptitude Test (KCSAT) over a nine-year period across three data points, using both the Mantel-Haenszel (MH) and item response theory likelihood ratio (IRT-LR) procedures. Further, the study identified two factors (i.e. reading strategy and perceived interest) that explained a portion of the variance in the magnitude of gender DIF via a series of multiple linear regression analyses. The results indicated (1) an interaction between item type and gender DIF and (2) a significant relationship between gender differences in the examinee’s perceived interest in test items and the magnitude of gender DIF. The study discusses the results based on previous DIF research and presents pedagogical implications and some avenues for further research.

Antenatal antibiotic exposure in preterm infants with necrotizing enterocolitis

To determine whether an association exists between antenatal antibiotic exposure and incidence of necrotizing enterocolitis (NEC) in low birth weight infants. A retrospective case-control study was performed on all infants with a diagnosis of NEC born at our institition between 1988 and 2006. Medical histories of all infants with a diagnosis of NEC ≥Bell's stage IIA and matched controls without NEC were reviewed. Maternal and neonatal characteristics were compared using the Mantel-Haenszel chi-square procedure, and logistic regression models were constructed to account for confounding. Clinical data for 97 matched pairs were analyzed. The adjusted odds ratio (OR) for antenatal exposure to ampicillin was significantly greater for infants who developed NEC (OR 2.3, 95% confidence interval 1.1, 4.8, P=0.003) than for control infants. Infants who developed NEC were more likely to have a history of in utero exposure to ampicillin in the immediate antepartum period than infants who did not develop NEC.

Angiogenin variants in Parkinson disease and amyotrophic lateral sclerosis

Several studies have suggested an increased frequency of variants in the gene encoding angiogenin (ANG) in patients with amyotrophic lateral sclerosis (ALS). Interestingly, a few ALS patients carrying ANG variants also showed signs of Parkinson disease (PD). Furthermore, relatives of ALS patients have an increased risk to develop PD, and the prevalence of concomitant motor neuron disease in PD is higher than expected based on chance occurrence. We therefore investigated whether ANG variants could predispose to both ALS and PD. We reviewed all previous studies on ANG in ALS and performed sequence experiments on additional samples, which allowed us to analyze data from 6,471 ALS patients and 7,668 controls from 15 centers (13 from Europe and 2 from the USA). We sequenced DNA samples from 3,146 PD patients from 6 centers (5 from Europe and 1 from the USA). Statistical analysis was performed using the variable threshold test, and the Mantel-Haenszel procedure was used to estimate odds ratios. Analysis of sequence data from 17,258 individuals demonstrated a significantly higher frequency of ANG variants in both ALS and PD patients compared to control subjects (p = 9.3 × 10(-6) for ALS and p = 4.3 × 10(-5) for PD). The odds ratio for any ANG variant in patients versus controls was 9.2 for ALS and 6.7 for PD. The data from this multicenter study demonstrate that there is a strong association between PD, ALS, and ANG variants. ANG is a genetic link between ALS and PD.

Clustering and meso-level variables in cross-sectional surveys: an example of food aid during the Bosnian crisis

BackgroundFocus groups, rapid assessment procedures, key informant interviews and institutional reviews of local health services provide valuable insights on health service resources and performance. A long-standing challenge of health planning is to combine this sort of qualitative evidence in a unified analysis with quantitative evidence from household surveys. A particular challenge in this regard is to take account of the neighbourhood or clustering effects, recognising that these can be informative or incidental.MethodsAn example of food aid and food sufficiency from the Bosnian emergency (1995-96) illustrates two Lamothe cluster-adjustments of the Mantel Haenszel (MH) procedure, one assuming a fixed odds ratio and the other allowing for informative clustering by not assuming a fixed odds ratio. We compared these with conventional generalised estimating equations and a generalised linear mixed (GLMM) model, using a Laplace adjustment.ResultsThe MH adjustment assuming incidental clustering generated a final model very similar to GEE. The adjustment that does not assume a fixed odds ratio produced a final multivariate model and effect sizes very similar to GLMM.DiscussionIn medium or large data sets with stratified last stage random sampling, the cluster adjusted MH is substantially more conservative than the naïve MH computation. In the example of food aid in the Bosnian crisis, the cluster adjusted MH that does not assume a fixed odds ratio produced similar results to the GLMM, which identified informative clustering.

Accent, listening assessment and the potential for a shared-L1 advantage: A DIF perspective

This paper reports on an investigation of the potential for a shared-L1 advantage on an academic English listening test featuring speakers with L2 accents. Two hundred and twelve second-language listeners (including 70 Mandarin Chinese L1 listeners and 60 Japanese L1 listeners) completed three versions of the University Test of English as a Second Language (UTESL) listening sub-test which featured an Australian English-accented speaker, a Japanese-accented speaker and a Mandarin Chinese-accented speaker. Differential item functioning (DIF) analyses were conducted on data from the tests which featured L2-accented speakers using two methods of DIF detection – the standardization procedure and the Mantel-Haenszel procedure – with candidates matched for ability on the test featuring the Australian English-accented speaker. Findings showed that Japanese L1 listeners were advantaged on a small number of items on the test featuring the Japanese-accented speaker, but these were balanced by items which favoured non-Japanese L1 listeners. By contrast, Mandarin Chinese L1 listeners were clearly advantaged across several items on the test featuring a Mandarin Chinese L1 speaker. The implications of these findings for claims of bias are discussed with reference to the role of speaker accent in the listening construct.

A New Approach for Differential Item Functioning Detection Using Mantel-Haenszel Methods. The GMHDIF Program

To date, the statistical software designed for assessing differential item functioning (DIF) with Mantel-Haenszel procedures has employed the following statistics: the Mantel-Haenszel chi-square statistic, the generalized Mantel-Haenszel test and the Mantel test. These statistics permit detecting DIF in dichotomous and polytomous items, although they limit the analysis to two groups. On the contrary, this article describes a new approach (and the related software) that, using the generalized Mantel-Haenszel statistic proposed by Landis, Heyman, and Koch (1978), permits DIF assessment in multiple groups, both for dichotomous and polytomous items. The program is free of charge and is available in the following languages: Spanish, English and Portuguese.

Accuracy of DIF Estimates and Power in Unbalanced Designs Using the Mantel–Haenszel DIF Detection Procedure

This study examined how much improvement was attainable with respect to accuracy of differential item functioning (DIF) measures and DIF detection rates in the Mantel–Haenszel procedure when employing focal and reference groups with notably unbalanced sample sizes where the focal group has a fixed small sample which does not satisfy the minimum DIF sample size requirement specified by the testing programs, while the reference group sample size far exceeds the minimum requirement. Results showed equivalent or better results with such unbalanced but large samples than with some of the currently used minimum DIF sample size conditions. DIF investigation, therefore, does not necessarily need to cease when the focal group does not meet the minimum sample size requirement. Some analytic explanations and guidelines for DIF investigations with unbalanced sample sizes are also provided.

Disease‐related differential item functioning in the work instability scale for rheumatoid arthritis: Converging results from three methods

The 23-item Work Instability Scale for Rheumatoid Arthritis (RA-WIS) is a promising measure to assess risk for future work disability. Validated in both rheumatoid arthritis (RA) and osteoarthritis (OA), it has high potential for cross-disease applications. Our objective was to examine disease-related differential item functioning (DIF) in the RA-WIS. Workers with RA (n = 120) or OA (n = 130) were recruited from 3 sites and completed a questionnaire consisting of demographic and health- and work-related variables, including the RA-WIS (range 0-23, where 23 = highest work instability). Multiple DIF detection methods were applied for comparability: 1) Mantel-Haenszel and Breslow-Day procedures, 2) hierarchical 3-step sequential logistic regression procedure, and 3) a 1-parameter item response theory approach (Rasch analysis). Both tests of significance (chi-square and F tests) and effect size statistics (Δ(MH) , ΔR(2) ) were assessed to confirm items demonstrating uniform or nonuniform DIF. A 2-step purification procedure was applied to establish a DIF-free conditioning variable (total RA-WIS score) for DIF analyses. The resultant impact of disease-related DIF at the scale level was also evaluated. All 3 DIF detection methods converged to reveal 3 RA-WIS items as having significant disease-related uniform DIF. Two items ("difficulty opening doors" and "pressure on hand") were more likely affirmed in RA, while 1 item ("very stiff") was more likely affirmed in OA. Overall, only a marginal impact at the scale level was found due to a small proportion of scale items exhibiting DIF and the bidirectional nature of DIF effects. RA-WIS scores can be directly compared between RA and OA without significant concerns for DIF-related measurement bias.

Fishery biology of the swordfish (Xiphias gladius, Linnaeus 1758) caught by surface longliners based in Itajaí, southern Brazil

The Xiphias gladius (Linnaeus, 1758) data used in the present study were obtained from commercial fishing cruises of the pelagic longline fleet based in Itajaí, operating in the Southwest Atlantic. For the Covariance models, fishing effort (number of hooks) and catches (kg) were pooled and categorized into the following groups: (1) year (1997, 1998, 2001, 2002 and 2007); (2) lunar phase; and (3) seasons, fishing effort being the covariate. The highest mean swordfish catch (kg) occurred during the waxing crescent moon and during wintertime. The Mantel-Haenszel procedure was applied to compare the efficiency of the two hooks tested, and showed that the "J" hook type (J 9/0 10º offset), traditionally used by the fleet, catches 1.2-fold more Xiphias gladius than the circle hook (18/0 10º offset), used as an optional device to reduce sea turtle bycatch. The conversion of individual dressed weight (kg) into total length (cm) from 2000 to 2002, showed that sub-adults represented 66% of the total catch.

Meta-Analysis of Multiple Primary Prevention Trials of Cardiovascular Events Using Aspirin

Several meta-analyses have focused on determination of the effectiveness of aspirin (acetylsalicylic acid) in primary prevention of cardiovascular (CV) events. Despite these data, the role of aspirin in primary prevention continues to be investigated. Nine randomized trials have evaluated the benefits of aspirin for the primary prevention of CV events: the British Doctors' Trial (BMD), the Physicians' Health Study (PHS), the Thrombosis Prevention Trial (TPT), the Hypertension Optimal Treatment (HOT) study, the Primary Prevention Project (PPP), the Women's Health Study (WHS), the Aspirin for Asymptomatic Atherosclerosis Trial (AAAT), the Prevention of Progression of Arterial Disease and Diabetes (POPADAD) trial, and the Japanese Primary Prevention of Atherosclerosis With Aspirin for Diabetes (JPAD) trial. The combined sample consists of about 90,000 subjects divided approximately evenly between those taking aspirin and subjects not taking aspirin or taking placebo. A meta-analysis of these 9 trials assessed 6 CV end points: total coronary heart disease, nonfatal myocardial infarction (MI), total CV events, stroke, CV mortality, and all-cause mortality. No covariate adjustment was performed, and appropriate tests for treatment effect, heterogeneity, and study size bias were applied. The meta-analysis suggested superiority of aspirin for total CV events and nonfatal MI, (p <0.05 for each), with nonsignificant results for decreased risk for stroke, CV mortality, and all-cause mortality. There was no evidence of a statistical bias (p >0.05). In conclusion, aspirin decreased the risk for CV events and nonfatal MI in this large sample. Thus, primary prevention with aspirin decreased the risk for total CV events and nonfatal MI, but there were no significant differences in the incidences of stroke, CV mortality, all-cause mortality and total coronary heart disease.

Formulating the Rasch Differential Item Functioning Model Under the Marginal Maximum Likelihood Estimation Context and Its Comparison With Mantel–Haenszel Procedure in Short Test and Small Sample Conditions

This study elaborates the Rasch differential item functioning (DIF) model formulation under the marginal maximum likelihood estimation context. Also, the Rasch DIF model performance was examined and compared with the Mantel–Haenszel (MH) procedure in small sample and short test length conditions through simulations. The theoretically known relationship of the DIF estimators between the Rasch DIF model and the MH procedure was confirmed. In general, the MH method showed a conservative tendency for DIF detection rates compared with the Rasch DIF model approach. When there is DIF, the z test (when the standard error of the DIF estimator is estimated properly) and the likelihood ratio test in the Rasch DIF model approach showed higher DIF detection rates than the MH chi-square test for sample sizes of 100 to 300 per group and test lengths ranging from 4 to 39. In addition, this study discusses proposed Rasch DIF classification rules that accommodate statistical inference on the direction of DIF.

Identification of Differential Item Functioning in Assessment Booklet Designs With Structurally Missing Data

The Mantel—Haenszel procedure is a popular technique for determining items that may exhibit differential item functioning (DIF). Numerous studies have focused on the strengths and weaknesses of this procedure, but few have focused the performance of the Mantel—Haenszel method when structurally missing data are present as a result of test booklet design. The current research compares the Type I error rate, power rate, and accuracy of DIF statistics when using the Mantel—Haenszel methods to detect DIF in three different test booklet designs that produce sparse data. Data were simulated for three different total sample sizes, fully crossed with three different test lengths, and two focal to reference group ratios. In general, the Mantel—Haenszel procedure was not adversely affected by sparseness of data under any of the booklet designs considered when sample size was adequately large. As sample sizes decrease, both the amount of sparseness and features of the booklet design may influence the ability to detect DIF.

Is there an increased risk for drug treated attention deficit/hyperactivity disorder in children born after in vitro fertilization?

Background There is mounting evidence that children born after in vitro fertilization (IVF) run an increased risk of neurological complications and notably cerebral palsy. Whether developmental disturbances occur more often than expected is debated. Aim To investigate the risk for ADHD in children conceived after IVF. Methods Children conceived after IVF and born between 1982 and 2005 were identified from all IVF clinics in Sweden. Children who developed attention deficit/hyperactivity disorder (ADHD) were identified with the use of a register over all prescribed drugs in Sweden, using prescriptions for methylphenidate or atomixetine as indicators of ADHD. Maternal and neonatal characteristics were obtained by linkage with the Medical Birth Register and relevant confounders were adjusted for using Mantel–Haenszel procedures. We studied 28 158 children born after IVF and compared them with 2 417 886 children in the population. Results After adjustment for year of birth, maternal age, parity, smoking, BMI, and maternal education and after exclusion of women who did not cohabit, a weak but statistically significant association was found with an odds ratio = 1.18, 95% confidence interval 1.03–1.36. The effect was stronger in girls (OR = 1.40) than boys (OR = 1.11) but this difference could be random. After adjustment for length of involuntary childlessness, the OR decreased slightly and lost statistical significance. Conclusions The study suggests a weak association between IVF and drug treated ADHD.

Planning and analysis of cross‐over trials in infertility

The use of cross-over trials in investigating treatments for infertility is discussed. A simple possible approach to analyzing such trials using the Mantel-Haenszel procedure is explained. A more flexible approach based on the normal-binomial mixture model of Ezzet and Whitehead is examined. It is shown how this may be implemented in various statistical packages by applying it to two real examples of trials in infertility. An approach that may be used to compare designs via simulation is explained briefly. It is concluded that provided that cross-over trials in infertility are regarded as parallel group trials with extra information rather than as cross-over trials with missing information, their use may be more promising than that has previously been concluded.

Clustering of Mycobacterium tuberculosis cases in Acapulco: Spoligotyping and risk factors.

Recurrence and reinfection of tuberculosis have quite different implications for prevention. We identified 267 spoligotypes of Mycobacterium tuberculosis from consecutive tuberculosis patients in Acapulco, Mexico, to assess the level of clustering and risk factors for clustered strains. Point cluster analysis examined spatial clustering. Risk analysis relied on the Mantel Haenszel procedure to examine bivariate associations, then to develop risk profiles of combinations of risk factors. Supplementary analysis of the spoligotyping data used SpolTools. Spoligotyping identified 85 types, 50 of them previously unreported. The five most common spoligotypes accounted for 55% of tuberculosis cases. One cluster of 70 patients (26% of the series) produced a single spoligotype from the Manila Family (Clade EAI2). The high proportion (78%) of patients infected with cluster strains is compatible with recent transmission of TB in Acapulco. Geomatic analysis showed no spatial clustering; clustering was associated with a risk profile of uneducated cases who lived in single-room dwellings. The Manila emerging strain accounted for one in every four cases, confirming that one strain can predominate in a hyperendemic area.

Translationally optimal codons associate with aggregation‐prone sites in proteins

We analyze the relationship between codon usage bias and residue aggregation propensity in the genomes of four model organisms, Escherichia coli, yeast, fly, and mouse, as well as the archaeon Halobacterium species NRC-1. Using the Mantel-Haenszel procedure, we find that translationally optimal codons associate with aggregation-prone residues. Our results are qualitatively and quantitatively similar to those of an earlier study where we found an association between translationally optimal codons and buried residues. We also combine the aggregation-propensity data with solvent-accessibility data. Although the resulting data set is small, and hence statistical power low, results indicate that the association between optimal codons and aggregation-prone residues exists both at buried and at exposed sites. By comparing codon usage at different combinations of sites (exposed, aggregation-prone sites versus buried, non-aggregation-prone sites; buried, aggregation-prone sites versus exposed, non-aggregation-prone sites), we find that aggregation propensity and solvent accessibility seem to have independent effects of (on average) comparable magnitude on codon usage. Finally, in fly, we assess whether optimal codons associate with sites at which amino acid substitutions lead to an increase in aggregation propensity, and find only a very weak effect. These results suggest that optimal codons may be required to reduce the frequency of translation errors at aggregation-prone sites that coincide with certain functional sites, such as protein-protein interfaces. Alternatively, optimal codons may be required for rapid translation of aggregation-prone regions.

Blinded and uniform causes of death verification in cancer screening: A major influence on the outcome of a prostate cancer screening trial?

Background To assess the agreement between the causes of death assigned by a blinded and uniform review panel of the Rotterdam section of the European Randomised Study of Screening for Prostate Cancer and the official vital statistics and to explore the possible effect of the use of either of these two sources on the outcome of the screening trial. Methods A total of 670 deaths amongst men with prostate cancer, reviewed by the causes of death committee (CODC) up to 31st December 2006 were included in this study. The kappa statistics with confidence intervals (CI), sensitivity and specificity of the official statistics were determined, with the CODC considered the gold standard. The rate ratio (RR) and 95% confidence intervals (95% CI) for prostate cancer mortality, official statistics relative to CODC, were calculated following the Mantel–Haenszel procedure. Results The overall concordance and the kappa between official statistics and the CODC were 90.6% and 0.76 (0.71–0.82), remaining comparable when only the CODC category definitely prostate cancer was applied, with the sensitivity of official statistics increasing from 88.3% to 91.3% and specificity hardly changing (91.3% and 90.5%). High specificity and lower sensitivity is observed in the screening arm, whilst the opposite was seen in the control arm in men aged 55–69 and 70–74 years at entry. Considerable lower false positive rate was seen for both age groups in the screening arm (3.9% and 4.7%) compared to the control arm (8.4% and 14.3%). A statistically significant excess of prostate cancer death was observed for the official statistics in the age group 70–74 years, 1.53 (1.07–2.19), whilst it was not significant for men aged 55–69 at entry, 1.06 (0.83–1.36). Conclusion In the Rotterdam ERSPC section, official statistics tended to overreport prostate cancer as an underlying cause of death, particularly in the age group 70-plus in the control arm, which would overestimate the true effect in favour of screening.