Optimal approximate conversions of odds ratios and hazard ratios to risk ratios

Odds ratios approximate risk ratios when the outcome under consideration is rare but can diverge substantially from risk ratios when the outcome is common. In this paper, we derive optimal analytic conversions of odds ratios and hazard ratios to risk ratios that are minimax for the bias ratio when outcome probabilities are specified to fall in any fixed interval. The results for hazard ratios are derived under a proportional hazard assumption for the exposure. For outcome probabilities specified to lie in symmetric intervals centered around 0.5, it is shown that the square-root transformation of the odds ratio is the optimal minimax conversion for the risk ratio. General results for any nonsymmetric interval are given both for odds ratio and for hazard ratio conversions. The results are principally useful when odds ratios or hazard ratios are reported in papers, and the reader does not have access to the data or to information about the overall outcome prevalence.     

Blowfly strike in England and Wales: the relationship between prevalence and farm and management factors

Abstract. In order to develop and evaluate control strategies for blowfly strike, a greater understanding of the epidemiology is essential. A postal survey of sheep farmers yielded information about ten farm and management factors and their relationship to blowfly strike prevalence. The risk of a farm reporting at least one case of blowfly strike increased as flock size and stocking density increased (adjusted odds ratio of 1.13 for an increase in flock size of 100 sheep and 1.38 for an increase in stocking density of ten sheep per hectare). As farm altitude increased, the risk of blowfly strike decreased (adjusted odds ratio 0.67 for an increase in farm altitude of 100 m). The risk of high strike prevalence (more than 2% of sheep struck) decreased as both farm altitude and flock size increased. High strike prevalence was also associated with on-farm sheep carcase disposal (odds ratio 1.35). Farmers in the south-west of England were more likely to report at least one case of blowfly strike and high strike prevalence compared to all other regions.     

Covariate Adjustment and Ranking Methods to Identify Regions with High and Low Mortality Rates

Summary Identifying regions with the highest and lowest mortality rates and producing the corresponding color‐coded maps help epidemiologists identify promising areas for analytic etiological studies. Based on a two‐stage Poisson–Gamma model with covariates, we use information on known risk factors, such as smoking prevalence, to adjust mortality rates and reveal residual variation in relative risks that may reflect previously masked etiological associations. In addition to covariate adjustment, we study rankings based on standardized mortality ratios (SMRs), empirical Bayes (EB) estimates, and a posterior percentile ranking (PPR) method and indicate circumstances that warrant the more complex procedures in order to obtain a high probability of correctly classifying the regions with the upper 100γ% and lower 100γ% of relative risks for γ= 0.05, 0.1 , and 0.2. We also give analytic approximations to the probabilities of correctly classifying regions in the upper 100γ% of relative risks for these three ranking methods. Using data on mortality from heart disease, we found that adjustment for smoking prevalence has an important impact on which regions are classified as high and low risk. With such a common disease, all three ranking methods performed comparably. However, for diseases with smaller event counts, such as cancers, and wide variation in event counts among regions, EB and PPR methods outperform ranking based on SMRs.     

Extrapolation for Terrestrial Vertebrates

Risk assessment for terrestrial birds and mammals is most usually conducted for pesticides in standardized systems based on results of limited tests required for regulatory approval. Increasingly, attempts at risk assessment are being made for other chemicals. Typically for pesticides, dietary tests are extrapolated to a few representative species and risk factors derived as ratios against modeled environmental concentrations. There has been criticism of the validity of some of the standard tests, which makes even this simple approach difficult to justify. Attempts have been made to extrapolate from those values considered more reliable using statistical approaches. Relative sensitivity of test species has been determined. However, reliable extrapolation from laboratory to field remains elusive. Statistically derived values from test results probably generate extremely conservative estimates of environmental no-effect levels. Substantial information on the biology, distribution and food preference of species has, thus far, barely been applied to risk assessment. Other promising approaches, such as species differences in metabolic capacity, population dynamics models, and even sublethal effects on reproductive or behavioural endpoints, cannot in themselves provide simple risk factors either. A simple approach to generate approximate or relative risk factors is explored. While the accumulation of a set of circumstantial evidence might not be regarded as risk assessment in the normal sense, it might offer us a means to extrapolate to a reasonable understanding of likely effects in the field and contribute to a weight-of-evidence approach that informs risk management. It also focuses further studies to those areas and species within the environment most likely to be adversely affected     

Standard errors for attributable risk for simple and complex sample designs

Adjusted attributable risk (AR) is the proportion of diseased individuals in a population that is due to an exposure. We consider estimates of adjusted AR based on odds ratios from logistic regression to adjust for confounding. Influence function methods used in survey sampling are applied to obtain simple and easily programmable expressions for estimating the variance of AR. These variance estimators can be applied to data from case-control, cross-sectional, and cohort studies with or without frequency or individual matching and for sample designs with subject samples that range from simple random samples to (sample) weighted multistage stratified cluster samples like those used in national household surveys. The variance estimation of AR is illustrated with: (i) a weighted stratified multistage clustered cross-sectional study of childhood asthma from the Third National Health and Examination Survey (NHANES III), and (ii) a frequency-matched case-control study of melanoma skin cancer.     

Effectiveness of helmets in skiers and snowboarders: case-control and case crossover study

Objective To determine the effect of helmets on the risk of head and neck injuries in skiers and snowboarders.Design Matched case-control and case crossover study.Setting 19 ski areas in Quebec, Canada, November 2001 to April 2002.Participants 1082 skiers and snowboarders (cases) with head and neck injuries reported by the ski patrol and 3295 skiers and snowboarders (controls) with non-head or non-neck injuries matched to cases at each hill.Main outcome measures Estimates of matched odds ratios for the effect of helmet use on the risk of any head or neck injury and for people requiring evacuation by ambulance.Results The adjusted odds ratio for helmet use in participants with any head injury was 0.71 (95% confidence interval 0.55 to 0.92), indicating a 29% reduction in the risk of head injury. For participants who required evacuation by ambulance for head injuries, the adjusted odds ratio for helmet use was 0.44 (0.24 to 0.81). Similar results occurred with the case crossover design (odds ratio 0.43, 0.09 to 1.83). The adjusted odds ratio for helmet use for participants with any neck injury was 0.62 (0.33 to 1.19) and for participants who required evacuation by ambulance for neck injuries it was 1.29 (0.41 to 4.04).Conclusions Helmets protect skiers and snowboarders against head injuries. We cannot rule out the possibility of an increased risk of neck injury with helmet use, but the estimates on which this assumption is based are imprecise.     

Relationships of Psychiatric Disorders with Overweight and Obesity in an Adult General Population**

Objective: To explore relationships of smoking and risk drinking status, nicotine and alcohol dependence, and anxiety, depressive, and somatoform disorders with overweight and obesity. Research Methods and Procedures: A probability sample was drawn that was representative for the adult general population, 18 to 64 years of age, in one region of Germany; the participation rate was 70.2%. After excluding those who were pregnant or had a current eating disorder according to the DSM‐IV, 4063 individuals remained. Overweight and obesity were defined according to the BMI that was assessed in the face‐to‐face in‐home standardized interview (Composite International Diagnostic Interview) on psychiatric disorders. Results: Men with a former nicotine dependence had higher odds of being overweight than men who never had a nicotine dependence (adjusted odds ratio, 1.5; confidence interval, 1.1 to 2.1). Men at current risk for drinking and current alcohol‐dependent or abusing men had lower odds of being overweight compared with men who never were alcohol dependent, abusing, or at risk for drinking (adjusted odds ratio, 0.3; confidence interval, 0.8 to 0.9). Effect sizes were small. No relationship of overweight with depressive, anxiety, or somatoform disorders was found in the multivariate analysis. Discussion: There is a relationship between being overweight and nicotine and alcohol dependence or abuse among men but not among women. Even though one reason for women to refrain from quitting smoking is the fear of weight gain, these results do not support this. This information could help convince women to try to quit smoking.     

Weight-of-Evidence (WOE): Quantitative Estimation of Probability of Impairment for Individual and Multiple Lines of Evidence

Environmental decision-making is complex and often based on multiple lines of evidence. Integrating the information from these multiple lines of evidence is rarely a simple process. We present a quantitative approach to the combination of multiple lines of evidence through calculation of weight-of-evidence, with reference conditions used to define a not impaired state. The approach is risk-based with measurement of risk computed as the probability of impairment. When data on reference conditions are available, there are a variety of methods for calculating this probability. Statistical theory and the use of odds ratios provide a method for combining the measures of risk from the different lines of evidence. The approach is illustrated using data from the Great Lakes to predict the risk at potentially contaminated sites.     

Estimating standardized risk differences from odds ratios

Holland (1989, Biometrics 45, 1009-1016) gave simple formulas for an "adjusted" risk difference based on the Mantel-Haenszel odds ratio estimator and its variance. This "adjusted" risk difference is, in general, inconsistent, but Holland's variance formula is an immediate corollary of a more general formula by Greenland (1987, Journal of Chronic Diseases 40, 1087-1094). We show how, under a large-stratum limiting model, one can derive consistent estimators of standardized risk differences from any consistent odds ratio estimator. We also show how one can derive nonparametric standardized estimators under a sparse-data limiting model.     

Passive smoking at work as a risk factor for coronary heart disease in Chinese women who have never smoked.

OBJECTIVE--To study whether passive smoking at work is a risk factor for coronary heart disease. DESIGN--Case-control study. SETTING--Xi''an, China. SUBJECTS--59 patients with coronary heart disease and 126 controls, all Chinese women with full time jobs, who had never smoked cigarettes. RESULTS--The crude odds ratio for passive smoking from husband was 2.12 (95% confidence interval 1.06 to 4.25) and at work was 2.45 (1.23 to 4.88). The final logistic regression model, with passive smoking from husband and at work as the base, included age, history of hypertension, type A personality, and total cholesterol and high density lipoprotein cholesterol concentrations; the adjusted odds ratios for passive smoking from husband and at work were 1.24 (0.56 to 2.72) and 1.85 (0.86 to 4.00) respectively. For passive smoking at work, statistically significant linear trends of increasing risks (for both crude and adjusted odds ratios) with increasing exposures (amount exposed daily, number of smokers, number of hours exposed daily, and cumulative exposure) were observed. When these exposure variables were analysed as continuous variables, the crude and adjusted odds ratios were also significant. CONCLUSION--Passive smoking at work is a risk factor for coronary heart disease. Urgent public health measures are needed to reduce smoking and to protect non-smokers from passive smoking in China.     

Reliability of abundance ratios between aeropalynological taxa as indicators of the climate in France

The abundance ratio between two aeropalynological taxa can sometimes give more reliable information about climate than a single taxon, especially in poorly standardized spectra of sediments, mosses or dusts. The reliabilities of 8385 ratios have been computed from 130 taxa observed in a net of 80 annual standardized aeropalynological spectra measured mostly in France. Ten average climatic factors express cold, heat and rain. The reliability of a pollen ratio is the relative frequency (between 0 and 100%) of the ranks, in the range of the climatic factor, where the ratio is more indicative than its two parent taxa. Only 169 out of the 8,385 ratios have a reliability of 80% or more, with one thousand ratios having a reliability of 50% or more. The most reliable ratios are from the less indicative taxa, whereas ubiquitous taxa give generally unreliable ratios. The accuracy and stability of estimates are approximately the same from the ratios as from the single taxa. The use of ratios foreshadows the use of sociological relations between taxa. Sociological processing will be used to transform the intermittent information about the environment given by the observed abundance of taxa into a gradual information given by their probable abundance.     

Habitual Consumption of Soy Products and Risk of Nasopharyngeal Carcinoma in Chinese Adults: A Case-Control Study

Background and Objectives

Many studies have shown a negative association between the consumption of soy products and the risk of some cancers, but little is known about the effect of soy consumption on nasopharyngeal carcinoma. We assessed the association between the consumption of soy products on nasopharyngeal carcinoma risk in Chinese individuals.

Methods

This case-control study included 600 (448 males and 152 females) incident cases of nasopharyngeal carcinoma, and an equal number of controls, matched according to gender, age (± 3 y) and household type to the nasopharyngeal carcinoma cases. All subjects were recruited from hospitals in Guangzhou, China. A face-to-face interview was conducted with each study individual to collect general information and habitual dietary intake using a 78-item quantitative food-frequency questionnaire. Odds ratios and their 95% confidence intervals were estimated using conditional logistic regression analyses.

Results

The median intakes of soy foods (in protein) were 0.5/0.5, 1.4/1.7, 2.7/3.3 and 6.1/7.7 (male/female) g/d in the quartiles 1 to 4. Both univariate and multivariate analyses showed no significant association between the consumption of soy proteins or soy isoflavones and the risk of nasopharyngeal carcinoma. The adjusted odds ratios (95% confidence intervals) between extreme quartiles were 0.97 (0.66-1.45) for soy proteins and 0.97 (0.66-1.42) for total isoflavones. Null associations were also observed between intake of the individual isoflavones daidzein, genistein and glycitein and NPC risk, with adjusted odds ratios for the extreme quartiles ranging between 0.73 and 1.23.

Conclusion

Habitual consumption of soy products had no significant effect on the risk of nasopharyngeal carcinoma in Chinese adults with a relatively low intake.     

The value of constant surveillance in a risky environment

In risky environments, where threats are unpredictable and the quality of information about threats is variable, all individuals face two fundamental challenges: balancing vigilance against other activities, and determining when to respond to warning signals. The solution to both is to obtain continuous estimates of background risk, enabling vigilance to be concentrated during the riskiest periods and informing about the likely cost of ignoring warnings. Human surveillance organizations routinely produce such estimates, frequently derived from indirect cues. Here we show that vigilant individuals in an animal society (the pied babbler, Turdoides bicolor) perform a similar role. We ask (i) whether, in the absence of direct predator threats, pied babbler sentinels react to indirect information associated with increased risk and whether they communicate this information to group mates; (ii) whether group mates use this information to adjust their own vigilance, and whether this influences foraging success; and (iii) whether information provided by sentinels reduces the likelihood of inappropriate responses to alarm calls. Using playback experiments, we show that: (i) sentinels reacted to indirect predator cues (in the form of heterospecific alarm calls) by giving graded surveillance calls; (ii) foragers adjusted their vigilance in reaction to changes in surveillance calls, with substantial effects on foraging success; and (iii) foragers reduced their probability of responding to alarm calls when surveillance calls indicated lowered risk. These results demonstrate that identifying attacks as they occur is only part of vigilance: equally important is continuous surveillance providing information necessary for individuals to make decisions about their own vigilance and evasive action. Moreover, they suggest that a major benefit of group living is not only the increased likelihood of detecting threats, but a marked improvement in the quality of information available to each individual.     

Effects of cluster sampling on epidemiologic analysis in population-based case-control studies

We consider population-based case-control designs in which controls are selected by one of three cluster sampling plans from the entire population at risk. The effects of cluster sampling on classical epidemiologic procedures are investigated, and appropriately modified procedures are developed. In particular, modified procedures for testing the homogeneity of odds ratios across strata, and for estimating and testing a common odds ratio are presented. Simulations that use the data from the 1970 Health Interview Survey as a population suggest that classical procedures may be fairly robust in the presence of cluster sampling. A more extreme example based on a mixed multinomial model clearly demonstrates that the classical Mantel-Haenszel (1959, Journal of the National Cancer Institute 22, 719-748) and Woolf-Haldane tests of no exposure effect may have sizes exceeding nominal levels and confidence intervals with less than nominal coverage under an alternative hypothesis. Classical estimates of odds ratios may also be biased with non-self-weighting cluster samples. The modified procedures we propose remedy these defects.     

Epidemiology of soil-transmitted helminth infections in Semarang,Central Java,Indonesia

Soil-transmitted helminth (STH) infections are endemic in Indonesia. However, prevalence data for many parts of the country are incomplete. The aim of this study was to determine human STH prevalence and knowledge and practices relating to STH risk behaviour, to provide a current view of the status of STH infection in rural communities in Central Java. A cross-sectional survey of 16 villages was conducted in Semarang, Central Java in 2015. Demographic and household data together with information about knowledge and practices relating to STH and hygiene were elicited through face-to-face interviews. Stool samples were collected and examined using the flotation method. Children (aged 2–12 years) also had their haemoglobin (Hb) levels, height and weight data collected, and BMI estimated. Data were analysed using univariate logistic regression analysis. A total of 6,466 individuals with a mean age of 33.5 years (range: 2–93) from 2,195 households were interviewed. The overall prevalence of STH was 33.8% with Ascaris lumbricoides (roundworm) the predominant nematode identified (prevalence = 26.0%). Hookworm and Trichuris trichiura (whipworm) were found in 7.9% and 1.8% of participants, respectively. Females were at increased odds of infection with A. lumbricoides (adjusted OR 1.14, 95% CI [1.02–1.29], p = 0.02). Adults in age groups 51–60 and over 60 years had the highest odds of being infected with hookworm (adjusted OR 3.01, 95% CI [1.84–4.91], p<0.001 and adjusted OR 3.79, 95% CI [2.30–6.26], p<0.001, respectively) compared to 6–12 year olds. Farmers also had higher odds of being infected with hookworm (adjusted OR 2.36, 95% CI [1.17–4.76], p = 0.02) compared to other occupation categories. Poverty (OR 2.14, 95% CI [1.77–2.58], p<0.001), overcrowding (OR 1.35, 95% CI [1.27–1.44], p<0.001), goat ownership (OR 1.61, 95% CI [1.10–2.41], p = 0.02) and the presence of dry floor space in the home (OR 0.73, 95% CI [0.58–0.91], p = 0.01) were all household factors significantly associated with an increased odds of infection. Infection with STH was not significantly associated with the gastrointestinal illness (p>0.05), BMI or Hb levels; however, one third of all 2–12 year olds surveyed were found to be anaemic (i.e. Hb concentrations below 110g/l or 115g/l for children under 5 and 5 years or older, respectively), with a greater proportion of school-age children at risk. Knowledge and behaviour related to hygiene and gastrointestinal diseases varied widely and were generally not associated with STH infection. The study revealed that STH infection remains endemic in Central Java despite ongoing deworming programs. Current control efforts would benefit from being re-evaluated to determine a more effective way forward.     

Increased Risk of Acute Pancreatitis in Patients with Type 2 Diabetes: An Observational Study Using a Japanese Hospital Database

Background

Increased risks of acute pancreatitis in patients with type 2 diabetes mellitus have been reported recently in several countries. We aimed to estimate the risks of acute pancreatitis in Japanese patients with diabetes mellitus.

Methods/Findings

We examined a large-scale hospital administrative database consisting of one million patients in 16 secondary medical care hospitals, from 2003 to 2010. The incidence rates of acute pancreatitis were estimated with cohort design; the odds ratios associated with diabetes mellitus and other comorbid risk factors were estimated with separate case-control analyses.In cohort analysis, the incidence of acute pancreatitis was higher in 14,707 diabetic patients than in 186,032 non-diabetic patients (4.75 vs. 1.65 per 1,000 patient-years) and increased in male patients and as age advanced. The adjusted odds ratio of acute pancreatitis in patients with diabetes mellitus was 1.86 (P<0.001) compared with non-diabetic patients in case-control analysis from 1,372 cases and 5,469 matched controls, which is consistent with the ones reported in previous studies. Alcoholism and gallstones were associated with a large increase in the risk of acute pancreatitis (adjusted odds ratio 13.40 and 14.29, respectively, P<0.001), although dyslipidemia was associated with significant risk reduction (adjusted odds ratio 0.62, P<0.001).

Conclusions

This observational study ascertained the elevated incidence rates and risk of acute pancreatitis in Japanese patients with diabetes. The risk estimates in Japanese patients with diabetes were in agreement with the ones reported in previous studies, and the elevated risk of acute pancreatitis in patients with diabetes would be generalized in different locations/populations.     

Population-based case-control study of isolated congenital cataract

BACKGROUND: The aim of this study was to detect possible etiological factors in the origin of isolated congenital cataracts. METHODS: The data set of the Hungarian Case-Control Surveillance of Congenital Abnormalities, 1980-2002, contains 111 cases of isolated congenital cataract and 111 matched control pairs without the defect, 37,837 population controls without any defects, and 22,744 malformed controls with other nonocular abnormalities. Exposure data and family history are based on prospective medical records, retrospective maternal information, and information obtained by regional nurses during a home visit with nonrespondent mothers. RESULTS: A positive family history indicated an autosomal-dominant origin in 10% of cases. Rubella infections occurred more frequently in case mothers than in control mothers before vaccination against rubella virus was instituted. A higher prevalence of influenza or common cold during pregnancy was found in the case group (55.9%) than in the population control group (18.5%; adjusted odds ratios [ORs], 5.8; 95% confidence interval (CI), 4.0-8.4) or in the malformed control group (21.7%; adjusted OR, 4.7; 95% CI, 3.2-6.9). The prevalence of acute infectious diseases of the respiratory system during pregnancy was also higher in the case group (26.1%) than in the population control group (9.1%; adjusted OR, 3.8; 95% CI, 2.5-5.8), or the malformed control group (9.3%; adjusted OR, 3.4; 95% CI, 2.3-5.3). The higher risk for isolated congenital cataract in cases of mothers with influenza or common cold and acute infectious diseases of the respiratory system during pregnancy was not found after administration of antifever therapy. CONCLUSIONS: Some isolated congenital cataracts are preventable by rubella vaccination and probably by influenza vaccination in the epidemic period. In addition, our results suggest that using antifever therapy for fever-related respiratory diseases may restrict the teratogenic risk of hyperthermia.     

Maternal Snuff Use and Smoking and the Risk of Oral Cleft Malformations - A Population-Based Cohort Study

Objective

To determine if maternal use of snuff (containing high levels of nicotine, low levels of nitrosamines and no combustion products) is associated with an increased risk of oral cleft malformations in the infant and whether cessation of snuff use or smoking before the antenatal booking influences the risk.

Method

A population-based cohort study was conducted on all live born infants, recorded in the Swedish Medical Birth Register from 1999 through 2009 (n = 1 086 213). Risks of oral clefts were evaluated by multivariate logistic regression analyses (using adjusted odds ratios, with 95% confidence intervals [CI]).

Results

Among 975 866 infants that had information on maternal tobacco use, 1761 cases of oral clefts were diagnosed. More than 50% of the mothers who used snuff or smoked three months prior pregnancy stopped using before the antenatal booking. Almost 8% of the mothers were smoking at the antenatal booking and 1,1% of the mothers used snuff. Compared with infants of non-tobacco users, the adjusted odds ratios (95% CI) of any oral cleft for infants of mothers who continued to use snuff or to smoke were 1.48 [1.00–2.21] and 1.19 [1.01–1.41], respectively. In contrast, in infants of mothers who stopped using snuff or stopped smoking before the antenatal booking, the corresponding risks were not increased (adjusted odds ratios [95% CI] were 0.71 [0.44–1.14] and 0.88 [0.73–1.05], respectively).

Conclusion

Maternal snuff use or smoking in early pregnancy is associated with an increased risk of oral clefts. Infants of mothers who stopped using snuff or stopped smoking before the antenatal booking had no increased risk of oral cleft malformations. Oral snuff or other sources of nicotine should not be recommended as an alternative for smoke-cessation during pregnancy.