On Graphically Checking Goodness-of-fit of Binary Logistic Regression Models

OBJECTIVES: This paper is concerned with checking goodness-of-fit of binary logistic regression models. For the practitioners of data analysis, the broad classes of procedures for checking goodness-of-fit available in the literature are described. The challenges of model checking in the context of binary logistic regression are reviewed. As a viable solution, a simple graphical procedure for checking goodness-of-fit is proposed. METHODS: The graphical procedure proposed relies on pieces of information available from any logistic analysis; the focus is on combining and presenting these in an informative way. RESULTS: The information gained using this approach is presented with three examples. In the discussion, the proposed method is put into context and compared with other graphical procedures for checking goodness-of-fit of binary logistic models available in the literature. CONCLUSION: A simple graphical method can significantly improve the understanding of any logistic regression analysis and help to prevent faulty conclusions.

LOGITCPRPLOT: Stata module to graph component-plus-residual plot for logistic regression

logitcprplot can be used after logistic regression for graphing a component-plus-residual plot (a.k.a. partial residual plot) for a given predictor, including a lowess, local polynomial, restricted cubic spline, fractional polynomial, penalized spline, regression spline, running line, or adaptive variable span running line smooth

RRLOGIT: Stata module to estimate logistic regression for randomized response data

rrlogit fits a maximum-likelihood logistic regression for randomized response data.

First trimester markers for pre-eclampsia: placental vs. non-placental protein serum levels

BACKGROUND/AIM: Parallel investigation, in a matched case-control study, of the association of different first-trimester markers with the risk of subsequent pre-eclampsia (PE). METHOD: The levels of different first trimester serum markers and fetal nuchal translucency thickness were compared between 52 cases of PE and 104 control women by non-parametric two-group comparisons and by calculating matched odds ratios. RESULTS: In univariable analysis increased concentrations of inhibin A and activin A were associated with subsequent PE (p < 0.02). Multivariable conditional logistic regression models revealed an association between increased risk of PE and increased inhibin A and translucency thickness and respectively reduced pregnancy-associated plasma protein A (PAPP-A) and placental lactogen . However, these associations varied with the gestational age at sample collection. For blood samples taken in pregnancy weeks 12 and 13 only, increased levels of activin A, inhibin A and nuchal translucency thickness, and lower levels of placenta growth factor and PAPP-A were associated with an increased risk of PE. CONCLUSIONS: Members of the inhibin family and to some extent PAPP-A and placental growth factor are superior to other serum markers, and the predictive value of these depends on the gestational age at blood sampling. The availability of a single, early pregnancy 'miracle' serum marker for PE risk assessment seems unlikely in the near future.

Influence of HIV-related immunodeficiency on the risk of hepatocellular carcinoma

OBJECTIVE: To investigate HIV-related immunodeficiency as a risk factor for hepatocellular carcinoma (HCC) among persons infected with HIV, while controlling for the effect of frequent coinfection with hepatitis C and B viruses. DESIGN: A case-control study nested in the Swiss HIV Cohort Study. METHODS: Twenty-six HCC patients were identified in the Swiss HIV Cohort Study or through linkage with Swiss Cancer Registries, and were individually matched to 251 controls according to Swiss HIV Cohort Study centre, sex, HIV-transmission category, age and year at enrollment. Odds ratios and corresponding confidence intervals were estimated by conditional logistic regression. RESULTS: All HCC patients were positive for hepatitis B surface antigen or antibodies against hepatitis C virus. HCC patients included 14 injection drug users (three positive for hepatitis B surface antigen and 13 for antibodies against hepatitis C virus) and 12 men having sex with men/heterosexual/other (11 positive for hepatitis B surface antigen, three for antibodies against hepatitis C virus), revealing a strong relationship between HIV transmission route and hepatitis viral type. Latest CD4+ cell count [Odds ratio (OR) per 100 cells/mul decrease = 1.33, 95% confidence interval (CI) 1.06-1.68] and CD4+ cell count percentage (OR per 10% decrease = 1.65, 95% CI 1.01-2.71) were significantly associated with HCC. The effects of CD4+ cell count were concentrated among men having sex with men/heterosexual/other rather than injecting drug users. Highly active antiretroviral therapy use was not significantly associated with HCC risk (OR for ever versus never = 0.59, 95% confidence interval 0.18-1.91). CONCLUSION: Lower CD4+ cell counts increased the risk for HCC among persons infected with HIV, an effect that was particularly evident for hepatitis B virus-related HCC arising in non-injecting drug users.

Seizures in patients with Alzheimer's disease or vascular dementia: a population-based nested case-control analysis

PURPOSE Patients with Alzheimer's disease (AD) have an increased risk of developing seizures or epilepsy. Little is known about the role of risk factors and about the risk of developing seizures/epilepsy in patients with vascular dementia (VD). The aim of this study was to assess incidence rates (IRs) of seizures/epilepsy in patients with AD, VD, or without dementia, and to identify potential risk factors of seizures or epilepsy. METHODS We conducted a follow-up study with a nested case-control analysis using the United Kingdom-based General Practice Research Database (GPRD). We identified patients aged ≥65 years with an incident diagnosis of AD or VD between 1998 and 2008 and a matched comparison group of dementia-free patients. Conditional logistic regression was used to estimate the odds ratio (OR) with a 95% confidence interval (CI) of developing seizures/epilepsy in patients with AD or VD, stratified by age at onset and duration of dementia as well as by use of antidementia drugs. KEY FINDINGS Among 7,086 cases with AD, 4,438 with VD, and 11,524 matched dementia-free patients, we identified 180 cases with an incident diagnosis of seizures/epilepsy. The IRs of epilepsy/seizures for patients with AD or VD were 5.6/1,000 person-years (py) (95% CI 4.6-6.9) and 7.5/1,000 py (95% CI 5.7-9.7), respectively, and 0.8/1,000 py (95% CI 0.6-1.1) in the dementia-free group. In the nested case-control analysis, patients with longer standing (≥3 years) AD had a slightly higher risk of developing seizures or epilepsy than those with a shorter disease duration, whereas in patients with VD the contrary was observed. SIGNIFICANCE Seizures or epilepsy were substantially more common in patients with AD and VD than in dementia-free patients. The role of disease duration as a risk factor for seizures/epilepsy seems to differ between AD and VD.

Use of diuretics and risk of incident gout: a population-based case-control study

OBJECTIVE Use of diuretics has been associated with an increased risk of gout. Data on different types of diuretics are scarce. We undertook this study to investigate the association between use of loop diuretics, thiazide or thiazide-like diuretics, and potassium-sparing agents and the risk of developing incident gout. METHODS We conducted a retrospective population-based case-control analysis using the General Practice Research Database established in the UK. We identified case patients who were diagnosed as having incident gout between 1990 and 2010. One control patient was matched to each case patient for age, sex, general practice, calendar time, and years of active history in the database. We used conditional logistic regression to calculate odds ratios (ORs) and 95% confidence intervals (95% CIs), and we adjusted for potential confounders. RESULTS We identified 91,530 incident cases of gout and the same number of matched controls. Compared to past use of diuretics from each respective drug class, adjusted ORs for current use of loop diuretics, thiazide diuretics, thiazide-like diuretics, and potassium-sparing diuretics were 2.64 (95% CI 2.47-2.83), 1.70 (95% CI 1.62-1.79), 2.30 (95% CI 1.95-2.70), and 1.06 (95% CI 0.91-1.23), respectively. Combined use of loop diuretics and thiazide diuretics was associated with the highest relative risk estimates of gout (adjusted OR 4.65 [95% CI 3.51-6.16]). Current use of calcium channel blockers or losartan slightly attenuated the risk of gout in patients who took diuretics. CONCLUSION Use of loop diuretics, thiazide diuretics, and thiazide-like diuretics was associated with an increased risk of incident gout, although use of potassium-sparing agents was not.

Incidence of and risk factors for severe hypoglycaemia in treated type 2 diabetes mellitus patients in the UK - a nested case-control analysis

AIMS To assess incidence rates (IRs) of and identify risk factors for incident severe hypoglycaemia in patients with type 2 diabetes newly treated with antidiabetic drugs. METHODS Using the UK-based General Practice Research Database, we performed a retrospective cohort study between 1994 and 2011 and a nested case-control analysis. Ten controls from the population at risk were matched to each case with a recorded severe hypoglycaemia during follow-up on general practice, years of history in the database and calendar time. Using multivariate conditional logistic regression analyses, we adjusted for potential confounders. RESULTS Of 130,761 patients with newly treated type 2 diabetes (mean age 61.7 ± 13.0 years), 690 (0.5%) had an incident episode of severe hypoglycaemia recorded [estimated IR 11.97 (95% confidence interval, CI, 11.11-12.90) per 10,000 person-years (PYs)]. The IR was markedly higher in insulin users [49.64 (95% CI, 44.08-55.89) per 10,000 PYs] than in patients not using insulin [8.03 (95% CI, 7.30-8.84) per 10,000 PYs]. Based on results of the nested case-control analysis increasing age [≥ 75 vs. 20-59 years; adjusted odds ratio (OR), 2.27; 95% CI, 1.65-3.12], cognitive impairment/dementia (adjusted OR, 2.00; 95% CI, 1.37-2.91), renal failure (adjusted OR, 1.34; 95% CI, 1.04-1.71), current use of sulphonylureas (adjusted OR, 4.45; 95% CI, 3.53-5.60) and current insulin use (adjusted OR, 11.83; 95% CI, 9.00-15.54) were all associated with an increased risk of severe hypoglycaemia. CONCLUSIONS Severe hypoglycaemia was recorded in 12 cases per 10,000 PYs. Risk factors for severe hypoglycaemia included increasing age, renal failure, cognitive impairment/dementia, and current use of insulin or sulphonylureas.

Poorly controlled type 2 diabetes mellitus is associated with a decreased risk of incident gout: a population-based case-control study

OBJECTIVE The aim of this study was to explore the risk of incident gout in patients with type 2 diabetes mellitus (T2DM) in association with diabetes duration, diabetes severity and antidiabetic drug treatment. METHODS We conducted a case-control study in patients with T2DM using the UK-based Clinical Practice Research Datalink (CPRD). We identified case patients aged ≥18 years with an incident diagnosis of gout between 1990 and 2012. We matched to each case patient one gout-free control patient. We used conditional logistic regression analysis to calculate adjusted ORs (adj. ORs) with 95% CIs and adjusted our analyses for important potential confounders. RESULTS The study encompassed 7536 T2DM cases with a first-time diagnosis of gout. Compared to a diabetes duration <1 year, prolonged diabetes duration (1-3, 3-6, 7-9 and ≥10 years) was associated with decreased adj. ORs of 0.91 (95% CI 0.79 to 1.04), 0.76 (95% CI 0.67 to 0.86), 0.70 (95% CI 0.61 to 0.86), and 0.58 (95% CI 0.51 to 0.66), respectively. Compared to a reference A1C level of <7%, the risk estimates of increasing A1C levels (7.0-7.9, 8.0-8.9 and ≥9%) steadily decreased with adj. ORs of 0.79 (95% CI 0.72 to 0.86), 0.63 (95% CI 0.55 to 0.72), and 0.46 (95% CI 0.40 to 0.53), respectively. Neither use of insulin, metformin, nor sulfonylureas was associated with an altered risk of incident gout. CONCLUSIONS Increased A1C levels, but not use of antidiabetic drugs, was associated with a decreased risk of incident gout among patients with T2DM.

Socioeconomic Status and Childhood Leukemia Incidence in Switzerland.

Socioeconomic status (SES) discrepancies exist for child and adult cancer morbidity and are a major public health concern. In this Swiss population-based matched case-control study on the etiology of childhood leukemia, we selected the cases from the Swiss Childhood Cancer Registry diagnosed since 1991 and the controls randomly from census. We assigned eight controls per case from the 1990 and 2000 census and matched them by the year of birth and gender. SES information for both cases and controls was obtained from census records by probabilistic record linkage. We investigated the association of SES with childhood leukemia in Switzerland, and explored whether it varied with different definitions of socioeconomic status (parental education, living condition, area-based SES), time period, and age. In conditional logistic regression analyses of 565 leukemia cases and 4433 controls, we found no consistent evidence for an association between SES and childhood leukemia. The odds ratio comparing the highest with the lowest SES category ranged from 0.95 (95% CI: 0.71-1.26; P trend = 0.73) for paternal education to 1.37 (1.00-1.89; P trend = 0.064) for maternal education. No effect modification was found for time period and age at diagnosis. Based on this population-based study, which avoided participation and reporting bias, we assume the potential association of socioeconomic status and childhood leukemia if existing to be small. This study did not find evidence that socioeconomic status, of Switzerland or comparable countries, is a relevant risk factor or strong confounder in etiological investigations on childhood leukemia.

Association of hormone therapy and incident gout: population-based case-control study

OBJECTIVE This study aims to assess the odds of developing incident gout in association with the use of postmenopausal estrogen-progestogen therapy, according to type, timing, duration, and route of administration of estrogen-progestogen therapy. METHODS We conducted a retrospective population-based case-control analysis using the United Kingdom-based Clinical Practice Research Datalink. We identified women (aged 45 y or older) who had a first-time diagnosis of gout recorded between 1990 and 2010. We matched one female control with each case on age, general practice, calendar time, and years of active history in the database. We used multivariate conditional logistic regression to calculate odds ratios (ORs) with 95% CIs (adjusted for confounders). RESULTS The adjusted OR for gout with current use of oral formulations of opposed estrogens (estrogen-progestogen) was 0.69 (95% CI, 0.56-0.86) compared with never use. Current use was associated with a decreased OR for gout in women without renal failure (adjusted OR, 0.71; 95% CI, 0.57-0.87) and hypertension (adjusted OR, 0.62; 95% CI, 0.44-0.87) compared with never use. Tibolone was associated with a decreased OR for gout (adjusted OR, 0.77; 95% CI, 0.63-0.95) compared with never use. Estrogens alone did not alter the OR for gout. CONCLUSIONS Current use of oral opposed estrogens, but not unopposed estrogens, is associated with a decreased OR for incident gout in women without renal failure and is more pronounced in women with hypertension. Use of tibolone is associated with a decreased OR for incident gout. The decreased OR for gout may be related to the progestogen component rather than the estrogen component.

Diabetes, use of antidiabetic drugs, and the risk of glioma

BACKGROUND Prior epidemiologic studies suggest inverse relations between diabetes and glioma risk, but the underlying mechanisms, including use of antidiabetic drugs, are unknown. METHODS We therefore performed a matched case-control analysis using the Clinical Practice Research Datalink (CPRD). We identified incident glioma cases diagnosed between 1995 and 2012 and matched each case with 10 controls on age, gender, calendar time, general practice, and years of active history in the CPRD. We performed conditional logistic regression to estimate odds ratios (ORs) with 95% CIs, adjusted for body mass index and smoking. RESULTS We identified 2005 cases and 20 050 controls. Diabetes was associated with decreased risk of glioma (OR = 0.74; 95% CI = 0.60-0.93), particularly glioblastoma (OR = 0.69; 95% CI = 0.51-0.94). Glioblastoma risk reduction was markedly pronounced among diabetic men (OR = 0.60; 95% CI = 0.40-0.90), most apparently for those with diabetes of long-term duration (OR for >5 vs 0 y = 0.46; 95% CI = 0.26-0.82) or poor glycemic control (OR for HbA1c ≥8 vs <6.5% = 0.20; 95% CI = 0.06-0.70). In contrast, the effect of diabetes on glioblastoma risk was absent among women (OR = 0.85; 95% CI = 0.53-1.36). No significant associations with glioma were found for use of metformin (OR for ≥30 vs 0 prescriptions = 0.72; 95% CI = 0.38-1.39), sulfonylureas (OR = 0.71; 95% CI = 0.39-1.30), or insulin (OR = 0.79; 95% CI = 0.37-1.69). CONCLUSIONS Antidiabetic treatment appears to be unrelated to glioma, but long-term diabetes duration and increased HbA1c both show decreased glioma risk. Stronger findings in men than women suggest low androgen levels concurrent with diabetes as a biologic mechanism.

A multinational case-control study on childhood brain tumours, anthropogenic factors, birth characteristics and prenatal exposures: A validation of interview data.

Little is known about the aetiology of childhood brain tumours. We investigated anthropometric factors (birth weight, length, maternal age), birth characteristics (e.g. vacuum extraction, preterm delivery, birth order) and exposures during pregnancy (e.g. maternal: smoking, working, dietary supplement intake) in relation to risk of brain tumour diagnosis among 7-19 year olds. The multinational case-control study in Denmark, Sweden, Norway and Switzerland (CEFALO) included interviews with 352 (participation rate=83.2%) eligible cases and 646 (71.1%) population-based controls. Interview data were complemented with data from birth registries and validated by assessing agreement (Cohen's Kappa). We used conditional logistic regression models matched on age, sex and geographical region (adjusted for maternal age and parental education) to explore associations between birth factors and childhood brain tumour risk. Agreement between interview and birth registry data ranged from moderate (Kappa=0.54; worked during pregnancy) to almost perfect (Kappa=0.98; birth weight). Neither anthropogenic factors nor birth characteristics were associated with childhood brain tumour risk. Maternal vitamin intake during pregnancy was indicative of a protective effect (OR 0.75, 95%-CI: 0.56-1.01). No association was seen for maternal smoking during pregnancy or working during pregnancy. We found little evidence that the considered birth factors were related to brain tumour risk among children and adolescents.

Nocardia infection in solid organ transplant recipients: a multicenter European case-control study.

BACKGROUND  Nocardiosis is a rare, life-threatening opportunistic infection, affecting 0.04% to 3.5% of patients after solid organ transplantation (SOT). The aim of this study was to identify risk factors for Nocardia infection after SOT and to describe the presentation of nocardiosis in these patients. METHODS  We performed a retrospective case-control study of adult patients diagnosed with nocardiosis after SOT between 2000 and 2014 in 36 European (France, Belgium, Switzerland, Netherlands, Spain) centers. Two control subjects per case were matched by institution, transplant date and transplanted organ. A multivariable analysis was performed using conditional logistic regression to identify risk factors for nocardiosis. RESULTS  One hundred and seventeen cases of nocardiosis and 234 control patients were included. Nocardiosis occurred at a median of 17.5 [range 2-244] months after transplantation. In multivariable analysis, high calcineurin inhibitor trough levels in the month before diagnosis (OR=6.11 [2.58-14.51]), use of tacrolimus (OR=2.65 [1.17-6.00]) and corticosteroid dose (OR=1.12 [1.03-1.22]) at the time of diagnosis, patient age (OR=1.04 [1.02-1.07]) and length of stay in intensive care unit after SOT (OR=1.04 [1.00-1.09]) were independently associated with development of nocardiosis; low-dose cotrimoxazole prophylaxis was not found to prevent nocardiosis. Nocardia farcinica was more frequently associated with brain, skin and subcutaneous tissue infections than were other Nocardia species. Among the 30 cases with central nervous system nocardiosis, 13 (43.3%) had no neurological symptoms. CONCLUSIONS  We identified five risk factors for nocardiosis after SOT. Low-dose cotrimoxazole was not found to prevent Nocardia infection. These findings may help improve management of transplant recipients.