We consider the estimation of the parameters indexing a parametric model for the conditional distribution of a diagnostic marker given covariates and disease status. Such models are useful for the evaluation of whether and to what extent a marker’s ability to accurately detect or discard disease depends on patient characteristics. A frequent problem that complicates the estimation of the model parameters is that estimation must be conducted from observational studies. Often, in such studies not all patients undergo the gold standard assessment of disease. Furthermore, the decision as to whether a patient undergoes verification is not controlled by study design. In such scenarios, maximum likelihood estimators based on subjects with observed disease status are generally biased. In this paper, we propose estimators for the model parameters that adjust for selection to verification that may depend on measured patient characteristics and additonally adjust for an assumed degree of residual association. Such estimators may be used as part of a sensitivity analysis for plausible degrees of residual association. We describe a doubly robust estimator that has the attractive feature of being consistent if either a model for the probability of selection to verification or a model for the probability of disease among the verified subjects (but not necessarily both) is correct.

BACKGROUND

In this study we prospectively evaluated the relationships between plasma concentrations of dehydroepiandrosterone (DHEA) and dehydroepiandrosterone sulfate (DHEA-S) and subsequent myocardial infarction in women.

METHODS

Using case-control sampling, we selected participants from the Nurses’ Health Study cohort. Blood samples were collected from 1989 to 1990 when the women were 43 to 69 years old. During follow-up through June 1998, 239 women were diagnosed with myocardial infarction (fatal and nonfatal). We matched cases 1:2 by age, cigarette smoking status, fasting status, and month of blood collection and used conditional logistic regression to adjust for potential confounders, including anthropometric factors and dietary intake.

RESULTS

Baseline median (10th, 90th percentiles) concentrations of DHEA were 17.1 (4.3, 46.7) nmol/L among women who subsequently developed myocardial infarction and 16.6 (6.1, 37.9) among controls. The risk of myocardial infarction increased with plasma concentrations of DHEA and its sulfate. Women in the highest DHEA quartile had a rate ratio (RR) of 1.27 (95% CI 0.92–1.74, P for trend = 0.008) for myocardial infarction compared with those in the lowest quartile, after adjusting for covariates. The results did not vary significantly by menopausal status, postmenopausal estrogen therapy, fasting status, or age at time of blood collection. Similar relationships between concentrations of DHEA-S and risk were observed, with an RR of 1.58 (95% CI 1.13–2.21; P for trend = 0.06) for myocardial infarction in the highest vs lowest quartile.

CONCLUSIONS

We observed a modest positive relationship between plasma concentrations of DHEA and its sulfate and the risk of subsequent myocardial infarction among predominantly postmenopausal women.

Background

Cysteine is a glutathione precursor, but is also a homocysteine byproduct. We prospectively evaluated relationships between fasting plasma concentrations of total cysteine and total homocysteine, and subsequent myocardial infarction (MI) in women.

Methods

Among 32826 women who provided blood samples between 1989 and 1990, 239 were diagnosed with incident MI after blood collection, but before July 1998. Of these women, 144 had provided a post-fast sample. We matched controls to cases 2:1 by age, cigarette smoking status, and month and fasting status at the time of blood collection. We used conditional logistic regression to adjust for confounding.

Results

Fasting total cysteine was positively related to MI risk in matching factor-adjusted analyses [RR for highest vs. lowest quartile 3.50 (95% CI 1.44, 8.52)]. However, after controlling for conventional risk factors of MI, it was not independently associated with risk [RR for highest vs. lowest quartile 1.32 (95% CI 0.42, 4.12; P trend=0.10)]. Fasting homocysteine was positively associated with MI risk; the multivariable adjusted rate ratio (RR) for the highest versus the lowest quartile was 3.37 (95% CI 1.30, 8.70; P trend=0.014).

Conclusions

Fasting plasma concentration of total homocysteine, but not total cysteine was positively associated with MI risk.

Background

Vitamin B6 is widely involved in amino acid metabolism and is a modulator of several reactions important to cardiovascular health. We prospectively evaluated relationships between fasting plasma levels of vitamin B6, as pyridoxal phosphate (PLP), to subsequent myocardial infarction risk in women. We also evaluated the predictors of fasting plasma concentration of pyridoxal phosphate.

Methods

Participants were adult nurses who completed questionnaires, and updated exposures every 2 years since 1976. Subjects for this analysis were selected by a nested case control design. Blood samples were collected between 1989 and 1990. We restricted our analysis to those women who had provided fasting blood samples (≥10 hours since last meal). During follow-up through June 1998, 144 were diagnosed with myocardial infarction (fatal and non-fatal). Cases were matched 1:2 by age, cigarette smoking status, and month and fasting status at the time of blood collection. Conditional logistic regression was used to adjust for potential confounders, including anthropometric factors, dietary intake, and selected biomarkers. Linear regression was used to determine which variables predict fasting total PLP concentration among control women.

Results

Median age at blood collection was 63. Among controls, lower estimated creatinine clearance, plasma total homocysteine and body mass index were statistically significant predictors of higher plasma PLP, as were higher dietary vitamin B6, and folate intake (all P <0.05). Plasma levels of pyridoxal phosphate were inversely associated with risk of MI, the multivariable adjusted rate ratio (RR) between extreme quarters was 0.22 (95% CI 0.09,0.55; Ptrend=0.05). The effect of plasma PLP varied by age. Among women who were aged less than 60 at blood sampling, the RR (95%CI) comparing top vs. bottom quarter was 0.03 (0.002,0.48), whereas among older women the corresponding RR (95%CI) was 0.43 (0.15,1.25).

Conclusions

Fasting plasma concentration of pyridoxal phosphate was inversely associated with MI risk. Plasma PLP is positively correlated with dietary vitamin B6, and is inversely correlated with renal function and body mass index. Future studies are needed to better understand both dietary and non-dietary determinants of plasma and tissue vitamin B6 status, and how these can be optimized to prevent MI and other diseases.

Background

A family history of prostate cancer (PCa) or breast cancer (BCa) has been associated with the risk of PCa, but the risks were inconsistent in terms of the affected family members, and data in the PSA era are limited.

Methods

This study included a subcohort of the Health Professionals Follow-Up Study composed of a highly PSA screened population from 1986 to 2004 with 3,695 PCa cases identified. Questionnaires and a food frequency questionnaire were administered every other and every four years, respectively. Family history of PCa and BCa was ascertained in 1990, 1992, and 1996. All statistics were two-sided.

Results

A family history of PCa in both a father and brother(s) was associated with a 2.3-fold increased risk of PCa [95% confidence interval (CI)=1.76–3.12]. Men with a father or brother(s) with a PCa diagnosis at age<60 and ≥60 had 2.16- and 1.95-fold increased risk of PCa, respectively. A family history of PCa was related to early-onset PCa (<65 years: RR=2.25, 95% CI=1.95–2.60) and weakly to late-onset PCa (≥65 years: RR=1.67, 95% CI=1.52–1.85). History of BCa in a mother or a sister was associated with a 1.22-fold increased risk of PCa (95% CI=1.08–1.38).

Conclusion

A family history of PCa or BCa significantly increases PCa risk. These associations are evident in a population with widespread PSA screening.