Morten Frydenberg

Does Low Participation in Cohort Studies Induce Bias

Background: Participation rates in large cohort studies have decreased during the last 2 decades. The consequences of this trend for relative risk estimation are unknown. Methods: The impact of a low participation rate (30%) on the Danish National Birth Cohort was examined among 49,751 women from the source population, including 15,373 participants in the cohort study. On the basis of independent data collection, we estimated odds ratios (ORs) in the source population and among participants for 3 exposure-risk associations: (a) in vitro fertilization and preterm birth, (b) smoking during pregnancy and birth of a small-for-gestational-age infant, and (c) prepregnancy body mass index and antepartum stillbirth. The effect of nonparticipation was described by a relative odds ratio (ROR), calculated as the OR(participants)/OR(source population). Two methods for calculation of confidence intervals for the relative odds ratio also were assessed. Results: The effect of nonparticipation on the selected ORs was small. The relative ORs were close to one and the bias was never larger than 16%, although some of the confidence intervals were wide. The 2 methods for calculation of confidence intervals gave very similar results and a small simulation study showed that the coverage probabilities were close to the 95% nominal level. Conclusion: For the 3 chosen associations, the ORs were not biased by nonparticipation. The results are reassuring for studies based on the Danish cohort and similar cohorts of pregnant women. The methodology used to compute confidence intervals for the relative odds ratios performed well in the scenarios considered.

Does smoking during pregnancy affect sons' sperm counts?

Background. There has been an apparent decline in sperm density during the last 5 decades in Denmark, a country in which women have among the highest rates of smoking in Europe. We examined semen quality and sex hormones in men in relation to their mothers’ tobacco smoking during pregnancy. Methods. Male participants were selected from the population-based Danish Twin Registry and the Danish Civil Registration System as part of a study on hereditary and environmental determinants of semen quality. From November 1999 to May 2000 we collected one fresh semen and blood sample from each of 316 men. Data on prenatal tobacco exposure were obtained for 265 of these men from a questionnaire filled in by their mothers. Results. Adjusting for age, current smoking status and other factors, sperm density was 48% lower (95% confidence interval = −69% to −11) among sons of mothers who smoked more than 10 cigarettes per day during pregnancy. Total sperm counts and levels of inhibin-B were also reduced among this group, whereas follicular stimulating hormone levels were somewhat higher (16% increase; 95% confidence interval = −13% to 54%). These effects were not seen in the lower smoking category (1–10 cigarettes per day). Conclusions. High levels of smoking (>10 cigarettes per day) during pregnancy may be a partial explanation for the apparent secular decline and the geographic differences in sperm counts.

Obstetric complications and risk for severe psychopathology in childhood.

The purpose of the study was to assess the association of obstetric complications with risk for mental disorders resulting in hospitalization before the age of 15. Records from all births in Denmark from 1973 through 1993 were linked to records of all psychiatric hospitalizations. Diagnoses were grouped into seven broad categories. A reference population of 10% of births in Denmark from 1973 to 1990 was used for comparison. Obstetric complications were associated with the range of mental disorders occurring in childhood. The strongest predictors were a variable indicating the interaction of birth weight with speed of growth and the 5-minute Apgar score. There was no diagnostic group that stood out as different with respect to obstetric complications. These results are consistent with the hypothesis of the continuum of reproductive casualty.

Time to diagnosis and mortality in colorectal cancer: a cohort study in primary care

Background:The relationship between the diagnostic interval and mortality from colorectal cancer (CRC) is unclear. This association was examined by taking account of important confounding factors at the time of first presentation of symptoms in primary care.Methods:A total of 268 patients with CRC were included in a prospective, population-based study in a Danish county. The diagnostic interval was defined as the time from first presentation of symptoms until diagnosis. We analysed patients separately according to the general practitioners interpretation of symptoms. Logistic regression was used to estimate 3-year mortality odds ratios as a function of the diagnostic interval using restricted cubic splines and adjusting for tumour site, comorbidity, age, and sex.Results:In patients presenting with symptoms suggestive of cancer or any other serious illness, the risk of dying within 3 years decreased with diagnostic intervals up to 5 weeks and then increased (P=0.002). In patients presenting with vague symptoms, the association was reverse, although not statistically significant.Conclusion:Detecting cancer in primary care is two sided: aimed at expediting ill patients while preventing healthy people from going to hospital. This likely explains the counterintuitive findings; but it does not explain the increasing mortality with longer diagnostic intervals. Thus, this study provides evidence for the hypothesis that the length of the diagnostic interval affects mortality in CRC patients.

Morbidity and GH deficiency: a nationwide study.

OBJECTIVE To estimate morbidity in Denmark in all patients with GH deficiency (GHD). DESIGN Morbidity was analyzed in 1794 GHD patients and 8014 controls matched on age and gender. All records in the GHD patients were studied and additional morbidity noted. Diagnoses and dates of admissions were identified in the National Patient Registry. Lag time until first admission was used as a measure of morbidity. Patients were divided into childhood onset (CO) and adult onset (AO), discriminated by an age cut-off of 18 years at onset of GHD. METHOD Sex- and cause-specific hazard ratios (HRs) in CO and AO GHD compared with controls. RESULTS Total morbidity was significantly increased in the GHD patients. HR for CO males: 3.1 (95% confidence interval (CI): 2.7-3.7), CO females: 3.2 (95% CI: 2.6-3.9), AO males: 2.9 (95% CI: 2.6-3.2), and AO females: 3.2 (95% CI: 2.8-3.6). In 18 out of 20 chapters from the International Classification of Diseases-10, a significantly increased morbidity was identified for at least one of the four subgroups of patients. Morbidity was significantly increased in all the four subgroups due to infectious, endocrine, pulmonary, urogenital, and neurological diseases; cancer; diseases of the eye, ear, and circulatory diseases; and traumas. Fractures were significantly increased in AO females, not in males. CONCLUSIONS Morbidity was significantly increased in the GHD patients. The increased morbidity was due to a variety of disorders, some of which can readily be explained by GHD and other pituitary deficiencies, while others cannot be easily explained.

Evidence of increasing mortality with longer diagnostic intervals for five common cancers: a cohort study in primary care.

BACKGROUND Early diagnosis is considered a key factor in improving the outcomes in cancer therapy; it remains unclear, however, whether long pre-diagnostic patient pathways influence clinical outcomes negatively. The aim of this study was to assess the association between the length of the diagnostic interval and the five-year mortality for the five most common cancers in Denmark while addressing known biases. METHODS A total of 1128 patients with colorectal, lung, melanoma skin, breast or prostate cancer were included in a prospective, population-based study in a Danish county. The diagnostic interval was defined as the time from the first presentation of symptoms in primary care till the date of diagnosis. Each type of cancer was analysed separately and combined, and all analyses were stratified according to the general practitioners (GPs) interpretation of the presenting symptoms. We used conditional logistic regression to estimate five-year mortality odds ratios as a function of the diagnostic interval using restricted cubic splines and adjusting for comorbidity, age, sex and type of cancer. RESULTS We found increasing mortality with longer diagnostic intervals among the approximately 40% of the patients who presented in primary care with symptoms suggestive of cancer or any other serious illness. In the same group, very short diagnostic intervals were also associated with increased mortality. Patients presenting with vague symptoms not directly related to cancer or any other serious illness had longer diagnostic intervals and the same survival probability as those who presented with cancer suspicious/serious symptoms. For the former, we found no statistically significant association between the length of the diagnostic interval and mortality. CONCLUSION In full coherence with clinical logic, the healthcare system instigates prompt investigation of seriously ill patients. This likely explains the counter-intuitive findings of high mortality with short diagnostic intervals; but it does not explain the increasing mortality with longer diagnostic intervals. Thus, the study provides further evidence for the hypothesis that the length of the diagnostic interval affects mortality negatively.

Selection by socioeconomic factors into the Danish National Birth Cohort

Background Low participation at recruitment to the Danish National Birth Cohort (DNBC) has raised concern about non-participation bias. Objective To study the socioeconomic pattern of participation to the DNBC. Methods Independently of the DNBC, we identified the DNBC source population in two geographical areas of Denmark by means of local birth registers with full coverage. Socioeconomic information came from national registers, and the source population consisted of 48,560 births including 15,290 participating women. For every socioeconomic characteristic, we estimated the prevalence ratio [prevalence (participants)/prevalence (source population)] which corresponds to the relative representation of the group (presented in percentages with 95% confidence intervals). Results The overall participation rate was 31%. Women outside the work force or with no further education than compulsory school were underrepresented in the DNBC by 62% (59%; 64%) and 43% (41%; 45%), respectively. Also, women were underrepresented by 18% (13%; 23%) if they were unemployed, by 22% (20%; 24%) if they were in the lowest income group, 38% (35%; 40%) if they received a high proportion of social benefits, and 28% (24%; 31%) if they were singles. Particularly women with low resources according to two socioeconomic factors were strongly underrepresented, typically by 50–67%. Conclusion Groups with low socioeconomic resources in terms of education, occupation, income and civil status are underrepresented in the DNBC compared to the background population. These discrepancies must be taken into account when results from the DNBC and other cohorts of pregnant women are interpreted—especially when descriptive results are presented.

Variable-resolution cone-beam computerized tomography with enhancement filtration compared with intraoral photostimulable phosphor radiography in detection of transverse root fractures in an in vitro model

OBJECTIVES The aim of this study was to compare the diagnostic accuracy of an intraoral photostimulable storage phosphor (PSP) plate system and cone-beam computerized tomography scanning (CBCT) for detection of experimentally induced transverse root fractures and to evaluate differences between original images and images enhanced with high-pass filters. STUDY DESIGN Sixty-nine extracted human teeth, 34 with root fractures and 35 without, were examined under standardized conditions using an intraoral PSP system (Digora Optime; Soredex). The images were saved in original 8-bit format and in a version sharpened with a high-pass filter. The teeth were examined with CBCT (i-Cat; Imaging Sciences) in 2 resolutions: 0.125 mm and 0.25 mm voxel size. Original images were saved together with images enhanced with 2 high-pass filters, sharpen and angio-sharpen. Six observers scored the presence of a root fracture in all modalities in random order. Sensitivity, specificity, and accuracy [(true positives + true negatives)/all scores] were calculated for each modality and each observer. Differences were estimated by analyzing the binary data, assuming additive effects of observer and modality in a generalized linear model. RESULTS High-resolution original CBCT images had higher sensitivity (P < .05) than lower-resolution images and PSP images (0.125 mm resolution 87%, 0.25 mm resolution 72%, and PSP 74%). Angio-sharpen-filtered images for both CBCT resolutions had higher sensitivities (P < .02) than the original images (0.125 mm resolution: 95% vs. 87%; 0.25 mm resolution: 81% vs. 72%). There was no significant difference between the lower-resolution CBCT and PSP images. Only small differences in specificity were seen between modalities, and accuracy was higher for high-resolution CBCT than for the other modalities (P < .03). CONCLUSIONS High-resolution i-Cat CBCT images resulted in an increase in sensitivity without jeopardizing specificity for detection of transverse root fractures compared with lower-resolution CBCT images, which were not more accurate than periapical PSP images. The angio-sharpen high-pass filter improved sensitivity in the high-resolution CBCT images in this in vitro model.

Obstetric factors, urbanization and psychosis

BACKGROUND Epidemiologic evidence as early as the 1930s has suggested urbanization is linked to schizophrenia, either by place of admission, place of upbringing, or, more recently, place of birth. In the past decade, obstetric complications have been implicated in the etiology of schizophrenia. METHODS With appropriate protections for anonymity, the files of the Danish Medical Birth Register were linked with the files of the Danish Psychiatric Case Register. The linkage produced 132 cases of schizophrenia and 69 cases of affective psychosis, who were born in 1973 or later, who entered a Danish psychiatric hospital before 1994. Controls were drawn from a 10% sample of the Medical Birth Register. Analysis was by logistic regression. RESULTS The risk of hospitalization for schizophrenia was 4.20 times higher (95% CI=2.4-7.4) for those born in Copenhagen versus those born in rural areas of Denmark, and a linear relationship was demonstrated between urbanization of birthplace and risk. There was no difference in risk of hospitalization for affective psychosis for those born in Copenhagen versus rural areas. Obstetric complications had a moderate sized relationship to schizophrenia, but the relationship of urban birth to schizophrenia was unaffected by adjustment for obstetric complications. CONCLUSION Urban birth is a strong risk factor for schizophrenia, not mediated by obstetric complications, which deserves further exploration.

Mental illness in new neurological patients

Objective: To determine the prevalence of psychiatric disorders in new neurological inpatients and outpatients, and examine whether they are recognised, treated, or referred to psychiatric consultation. Methods: 198 consecutive patients referred for the first time to a neurologist were studied using a two phase design. ICD-10 psychiatric diagnoses were established by means of the SCAN (Schedules for Clinical Assessment in Neuropsychiatry). Results: The overall prevalence of current mental disorders was 55.1% (95% CI: 46.2 to 63.8), and 65.0% (95% CI: 56.1 to 73.0) had at least once in their life had a psychiatric disorder. The most frequent current diagnoses were somatoform disorders (33.8%, (95% CI: 25.9 to 42.7%)), followed by phobias (21.8%; 95% CI: 15.3 to 30.0), substance use disorders (13.3%; 95% CI: 8.3 to 20.6) and depression/dysthymia (14.4; 95% CI: 9.1 to 21.8). The psychiatric morbidity markedly declined with increasing age. Compared with 63.5% of the women, 46.4% of the men had a psychiatric disorder. Substance use disorders were more frequent in men than women (p=0.002). Patients with a psychiatric disorder were more frequently seen in the outpatients’ clinic than those without. The neurologists detected 14%–40% of the cases, 16.9% were in treatment, and only 4.6% were referred to mental health care. Conclusion: Psychiatric disorders, in particular somatoform disorders, are extremely common in neurological patients, especially in young and middle aged patients, outpatients, and women. The results call for more research on mental illness’ impact on care and outcome in neurological patients.