Nathalie C. Støer

Comparison of estimators in nested case–control studies with multiple outcomes

Reuse of controls in a nested case–control (NCC) study has not been considered feasible since the controls are matched to their respective cases. However, in the last decade or so, methods have been developed that break the matching and allow for analyses where the controls are no longer tied to their cases. These methods can be divided into two groups; weighted partial likelihood (WPL) methods and full maximum likelihood methods. The weights in the WPL can be estimated in different ways and four estimation procedures are discussed. In addition, we address modifications needed to accommodate left truncation. A full likelihood approach is also presented and we suggest an aggregation technique to decrease the computation time. Furthermore, we generalize calibration for case-cohort designs to NCC studies. We consider a competing risks situation and compare WPL, full likelihood and calibration through simulations and analyses on a real data example.

Hyperemesis gravidarum and birth outcomes—a population‐based cohort study of 2.2 million births in the Norwegian Birth Registry

To study associations between hyperemesis gravidarum (HG) and birth outcomes.

Inverse probability weighting in nested case-control studies with additional matching—a simulation study

Nested case-control designs are inevitably less efficient than full cohort designs, and it is important to use available information as efficiently as possible. Reuse of controls by inverse probability weighting may be one way to obtain efficiency improvements, and it can be particularly advantageous when two or more endpoints are analyzed in the same cohort. The controls in a nested case-control design are often matched on additional factors than at risk status, and this should be taken into account when reusing controls. Although some studies have suggested methods for handling additional matching, a thorough investigation of how this affects parameter estimates and weights is lacking. Our aim is to provide such a discussion to help developing guidelines for practitioners. We demonstrate that it is important to adjust for the matching variables in regression analyses when the matching is broken. We present three types of estimators for the inverse sampling probabilities accounting for additional matching. One of these estimators was somewhat biased when the cases and controls were matched very closely. We investigated how additional matching affected estimates of interest, with varying degree of association between the matching variables and exposure/outcome. Strong associations introduced only a small bias when the matching variables were properly adjusted for. Sometimes, exposure variables, for example, blood samples, are analyzed in batches. Rather, strong batch effects had to be present before this introduced much bias when the matching was broken. All simulations are based on a study of prostate cancer and vitamin D.

Is hyperemesis gravidarum associated with placental weight and the placental weight-to-birth weight ratio? A population-based Norwegian cohort study.

INTRODUCTION Studies have suggested a link between placental weight, placental weight-to-birth weight ratio (PW/BW) and adult health. Hyperemesis gravidarum (HG) may also have implications for adult health. No studies on HG and placental characteristics have been identified. We therefore explored the relationship between HG, placental weight and the PW/BW-ratio in a population-based cohort. METHODS Singleton births to primiparous women between 1999 and 2009 with data on HG, placental weight and birth weight in the Medical Birth Registry of Norway (MBRN) comprised the study base (n = 200,390). HG was defined through ICD-10 code 021.0, 021.1 and 021.9. Gender and gestational age specific percentile curves for placenta weight and PW/BW ratio were used to define those below the 10th and above the 90th percentile of both outcomes. Associations between HG and dichotomous outcomes were studied by multiple logistic regression. Multiple linear regression was applied to study placental weight as a continuous variable. Male and female offspring were analyzed separately. RESULTS The prevalence of HG was 1.2%. Women with HG and female offspring had significantly higher risk of a PW/BW-ratio above the 90th percentile (OR = 1.17, 95% CI: 1.03-1.34). HG and PW/BW-ratio below the 10th percentile were inversely associated (OR = 0.70, 95% CI: 0.56-0.89). For male offspring no association was observed for HG and PW/BW-ratio below the 10th or above the 90th percentile. DISCUSSION/CONCLUSIONS We observed positive associations between HG and high PW/BW ratio limited to female offspring only. The high PW/BW-ratio suggests that there may be a possible link between HG and adult health.

Helicobacter pylori infection and severe hyperemesis gravidarum among immigrant women in Norway: a case-control study

OBJECTIVE To study associations between Helicobacter pylori exposure and severe hyperemesis gravidarum (HG) among immigrant women in Norway by exploring IgG seropositivity and H. pylori antigens in faeces. Additionally, we investigated whether cytotoxin-associated gene A product (CagA) and vacuolating cytotoxin A (VacA) seropositivity modulated this association. STUDY DESIGN An institution-based case-control study among immigrant women in Norway was performed at Ullevål and Akershus University Hospitals in September 2005-December 2007. Blood samples were used to explore IgG, CagA and VacA seropositivity, and faecal samples were used to explore the presence of antigens. Multiple logistic regressions were used to study associations between HG and H. pylori exposure. RESULTS The sample comprised 170 women: 62 cases and 108 controls. The observed proportion of IgG seropositive women did not differ between cases and controls. Neither IgG seropositivity nor CagA and VacA seropositivity were significantly associated with HG. For IgG positive and CagA and VacA negative women, the crude OR was 1.26 (95% CI: 0.57-2.82). For those being IgG positive and CagA and VacA positive, the crude OR was 0.82 (0.40-1.68). Adjustment for confounding factors, such as maternal age, body mass index and earlier HG, did not change the results. Additional adjustment for faecal antigens did not change the conclusions regarding these associations. Likewise, the crude OR for H. pylori antigens was not statistically significant. Adjustment for confounders and IgG seropositivity did not change this result. CONCLUSIONS This study did not find H. pylori exposure to be significantly associated with severe HG among immigrant women in Norway. This was regardless of whether H. pylori exposure was investigated by IgG seropositivity, CagA and VacA seropositivity or by the presence of H. pylori antigens in faeces. These results may indicate that the association between H. pylori and HG is weaker than previously expected, particularly in populations with high prevalence of H. pylori infection.

Hyperemesis gravidarum and maternal cancer risk, a scandinavian nested case-control study

Reproductive factors have been shown to influence cancer risk. Several pathological conditions during pregnancy have also been associated with subsequent altered cancer risk in the mother. Hyperemesis gravidarum (hyperemesis) is an early pregnancy condition characterized by severe nausea and vomiting resulting in weight loss and metabolic disturbances. Studies have reported associations between hyperemesis and cancer, but results are inconsistent. In this nested case‐control study we linked the population‐based medical birth registries and cancer registries in Norway, Sweden and Denmark in order to examine overall cancer risk and risk of specific cancer types in women with a history of hyperemesis, using conditional logistic regression. In total, 168,501 cases of cancer in addition to up to 10 cancer‐free controls per case were randomly sampled, matched on year of birth and birth registry (n = 1,721,626). Hyperemesis was defined through the International Classification of Diseases. Analyses were adjusted for potential confounders. Hyperemesis was inversely associated with overall cancer risk with adjusted relative risk (aRR) of 0.93 (95% CI: 0.88–0.99), with cancer in the lungs (aRR: 0.60, 95% CI: 0.44–0.81), cervix (aRR: 0.66, 95% CI: 0.49–0.91) and rectum (aRR: 0.48, 95% CI: 0.29–0.78). Thyroid cancer was positively associated with hyperemesis (aRR 1.45, 95% CI: 1.06–1.99) and risk increased with more than one hyperemetic pregnancy (aRR 1.80, 95% CI: 1.23–2.63). Hormonal factors, in particular human chorionic gonadotropin, are likely to be involved in mediating these effects. This study is the first to systematically address these associations and provides valuable knowledge on potential long‐term consequences of hyperemesis.

Menopausal hormone therapy and risk of melanoma: Do estrogens and progestins have a different role?

The association between use of menopausal hormone therapy (HT) and occurrence of skin malignant melanoma (SMM) is controversial. We investigated the issue in a nationwide cohort of 684,696 Norwegian women, aged 45–79 years, followed from 2004 to 2008. The study was based on linkage between Norwegian population registries. Multivariable Poisson regression models were used to estimate the effect of HT use, different HT types, routes of administration and doses of estrogen and progestin on the risk of SMM. During the median follow‐up of 4.8 years, 178,307 (26%) women used HT, and 1,476 incident SMM cases were identified. Current use of HT was associated with increased risk of SMM (rate ratios (RR) = 1.19; 95% confidence interval (CI) 1.03–1.37). Plain estrogen therapy was associated with an increased risk of SMM (RR 1.45; 95% CI 1.21–1.73), both for oral (RR 1.45; 95% CI 1.09–1.93) and vaginal (RR 1.44; 95% CI 1.14–1.84) formulations, while combined estrogen and progestin therapy (EPT) was not (RR 0.91; 95% CI 0.70–1.19). We performed a dose–response analysis of estrogen and progestin in women using tablets, and found that use of estrogens was associated with increased risk (RR 1.24; 95% CI 1.00–1.53 per 1 mg/day) and use of progestins with decreased risk (RR 0.71; 95% CI 0.57–0.89 per 10 mg/month) of SMM. In conclusion, estrogens were associated with increased risk of SMM, while combinations of estrogens and progestins were not. Our results suggest that estrogens and progestins might affect the risk of SMM in opposite ways.

Long Term Association between Serum 25-Hydroxyvitamin D and Mortality in a Cohort of 4379 Men

Objective A number of observational studies have shown an inverse association between circulating 25-hydroxyvitamin D and total mortality, but a reverse J-shaped association has also been reported. In a large nested case-control study, serum-25-hydroxyvitamin D (s-25(OH)D) was positively associated with incident prostate cancer. Based on the same study population, the primary aim of the present study was to investigate the association between s-25(OH)D and total mortality. Methods Men participating in population based health screenings during 1981–1991 and enrolled in a nested case-control study were followed throughout 2007 with respect to all-cause and cause-specific mortality. Hazard ratios (HR) with 95% confidence intervals (CI) were calculated using Cox proportional hazards regression. Results In men with prostate cancer (n = 2282), there was a significant inverse association between s-25(OH)D and total mortality after controlling for potential confounders (HR = 1.25 (95% CI 1.05–1.50), s-25(OH)D <50 nmol/l versus s-25(OH)D ≥50 nmol/l). The corresponding figure among controls (n = 2147) was HR = 1.15 (95% CI 0.88–1.50) and in the total study population HR = 1.19 (95% CI 1.03–1.38). For cause-specific deaths, we found no significant associations. Conclusions In this study population, s-25(OH)D was inversely associated with total mortality during more than two decades of follow-up, despite, as previous reported, high s-25(OH)D was associated with increased risk of prostate cancer.

Is the matched extreme case–control design more powerful than the nested case–control design?

For time-to-event data, the study sample is commonly selected using the nested case–control design in which controls are selected at the event time of each case. An alternative sampling strategy is to sample all controls at the same (pre-specified) time, which can either be at the last event time or further out in time. Such controls are the long-term survivors and may therefore constitute a more ‘extreme’ comparison group and be more informative than controls from the nested case–control design. We investigate this potential information gain by comparing the power of various ‘extreme’ case–control designs with that of the nested case–control design using simulation studies. We derive an expression for the theoretical average information in a nested and extreme case–control pair for the situation of a single binary exposure. Comparisons reveal that the efficiency of the extreme case–control design increases when the controls are sampled further out in time. In an application to a study of dementia, we identified Apolipoprotein E as a risk factor using a 1:1 extreme case–control design, which provided a hazard ratio estimate with a smaller standard error than that of a 2:1 nested case–control design.

Valid and efficient subgroup analyses using nested case-control data

Background It is not uncommon for investigators to conduct further analyses of subgroups, using data collected in a nested case-control design. Since the sampling of the participants is related to the outcome of interest, the data at hand are not a representative sample of the population, and subgroup analyses need to be carefully considered for their validity and interpretation. Methods We performed simulation studies, generating cohorts within the proportional hazards model framework and with covariate coefficients chosen to mimic realistic data and more extreme situations. From the cohorts we sampled nested case-control data and analysed the effect of a binary exposure on a time-to-event outcome in subgroups defined by a covariate (an independent risk factor, a confounder or an effect modifier) and compared the estimates with the corresponding subcohort estimates. Cohort analyses were performed with Cox regression, and nested case-control samples or restricted subsamples were analysed with both conditional logistic regression and weighted Cox regression. Results For all studied scenarios, the subgroup analyses provided unbiased estimates of the exposure coefficients, with conditional logistic regression being less efficient than the weighted Cox regression. Conclusions For the study of a subpopulation, analysis of the corresponding subgroup of individuals sampled in a nested case-control design provides an unbiased estimate of the effect of exposure, regardless of whether the variable used to define the subgroup is a confounder, effect modifier or independent risk factor. Weighted Cox regression provides more efficient estimates than conditional logistic regression.