Experience with multiple control groups in a large population-based case-control study on genetic and environmental risk factors.

We discuss the analytic and practical considerations in a large case-control study that had two control groups; the first control group consisting of partners of patients and the second obtained by random digit dialling (RDD). As an example of the evaluation of a general lifestyle factor, we present body mass index (BMI). Both control groups had lower BMIs than the patients. The distribution in the partner controls was closer to that of the patients, likely due to similar lifestyles. A statistical approach was used to pool the results of both analyses, wherein partners were analyzed with a matched analysis, while RDDs were analyzed without matching. Even with a matched analysis, the odds ratio with partner controls remained closer to unity than with RDD controls, which is probably due to unmeasured confounders in the comparison with the random controls as well as intermediary factors. However, when studying injuries as a risk factor, the odds ratio remained higher with partner control subjects than with RRD control subjects, even after taking the matching into account. Finally we used factor V Leiden as an example of a genetic risk factor. The frequencies of factor V Leiden were identical in both control groups, indicating that for the analyses of this genetic risk factor the two control groups could be combined in a single unmatched analysis. In conclusion, the effect measures with the two control groups were in the same direction, and of the same order of magnitude. Moreover, it was not always the same control group that produced the higher or lower estimates, and a matched analysis did not remedy the differences. Our experience with the intricacies of dealing with two control groups may be useful to others when thinking about an optimal research design or the best statistical approach.

The risk for breast cancer is not evidently increased in women with hyperprolactinemia.

The question has been raised whether hyperprolactinemia in humans is associated with an excess risk for breast cancer. We aimed to assess the risk of breast cancer in a previously defined large cohort of patients treated for idiopathic hyperprolactinemia or prolactinomas. Based on the pattern of drug prescriptions we identified 11,314 subjects in the PHARMO network with at least one dispensing of dopamine agonists between 1996 and 2006. Of these, 1,607 subjects were considered to have dopamine agonist-treated hyperprolactinemia based on the prescribing pattern. For the present analysis, we included only women (n = 1,342). Patients with breast cancer were identified by hospital discharge codes. Data on breast cancer incidence in the Netherlands were derived from the Dutch cancer registry. Standardized mortality ratio (SMR) was the measure of outcome to assess the association between hyperprolactinemia and breast cancer. The 1,342 patients accounted for a total of 6,576 person years. Eight patients with breast cancer during follow-up were identified. Indirect standardization with incidence proportions from the general Dutch population revealed a 7.47 expected cases. The calculated SMR for breast cancer risk in patients treated hyperprolactinemia was 1.07 (95% confidence interval 0.50-2.03). In conclusion, there is no clear evidence for increased breast cancer risk in female patients treated for either idiopathic hyperprolactinemia or prolactinomas. The uncertainty about the exact risk that is due to the relatively low number of breast cancer cases, should be overcome by pooling results in a future meta-analysis.

Mortality in acromegaly: a metaanalysis.

CONTEXT: Several studies have assessed mortality risk in patients treated for acromegaly. All studies found a mortality that was higher than expected for the general population, but most of these increases were not statistically significant. For this reason, it is not formally established whether mortality in acromegaly is different from the general population. OBJECTIVE: The objective of the study was to address the all-cause mortality risk in patients with acromegaly. DESIGN: The study was a metaanalysis. METHODS: Sixteen studies on mortality in patients with acromegaly were included. The principal outcome of the metaanalysis was the weighted average of the standardized mortality ratio (SMR) of all studies. In addition, we performed a subgroup analysis of studies in which more than 80% of the patients were treated by transsphenoidal approach. RESULTS: The weighted mean of the SMR from all 16 studies was 1.72 (95% confidence interval 1.62-1.83). In studies with transsphenoidal surgery as the primary therapy, the weighted mean of the SMR was 1.32 (95% confidence interval 1.12-1.56). CONCLUSIONS: This metaanalysis shows increased all-cause mortality in acromegalic patients, compared with the general population, even after transsphenoidal surgery.

Cardiovascular risk factors after bariatric surgery: Do patients gain more than expected from their substantial weight loss?

BACKGROUND: While it has been established that even limited weight loss (5-10%) improves obesity-associated cardiovascular risk factors, it is not known if considerable weight loss following laparoscopic adjustable silicone gastric banding (LASGB) results in a cardiovascular risk profile that is comparable, worse, or even better than that of matched control subjects. METHODS: Cardiovascular risk factors were compared in three groups of 24 women each: an index group that had lost considerable weight following LASGB for morbid obesity (BMI>40 kg/m(2)), a control group with the same BMI that the index group achieved after weight loss, and a pre-weight loss group of women with a BMI above 40 kg/m(2). Anthropometric measures, fasting serum glucose, insulin, lipids, C-reactive protein, and homocysteine levels were determined and insulin sensitivity was estimated using a homeostasis model assessment index (HOMA-IR). RESULTS: After bariatric surgery, the index group had a BMI of 32.0+/-0.8 kg/m(2). This resulted in a significantly better cardiovascular risk profile than that of the pre-weight loss group (BMI 42.8+/-0.6 kg/m(2)). Unexpectedly, after weight loss, the index group had significantly lower systolic blood pressure, fasting serum insulin, and HOMA-IR than the BMI-matched (32.8+/-0.9 kg/m(2)) control group. Although not significant, diastolic blood pressure, LDL-cholesterol, and CRP levels were also lower. CONCLUSION: Considerable weight loss following bariatric surgery leads to a greater improvement in cardiovascular risk factors than might be expected from the weight loss.

[Generalizability of therapeutic trials]. / Generaliseerbaarheid van therapeutisch onderzoek.

The external validity of a therapeutic trial addresses whether the study results may be applicable to other patients beyond the original study population. The inclusion criteria of the original study are a good indicator of the extent to which new patients are comparable with the study population, based on key characteristics. If the patients included in the study are not representative of the target population, the external validity is limited. The external validity can also be limited by study aspects that are not related to the inclusion criteria, such as the treatment protocol, the treatment centre, or the country in which the original study was performed. In exceptional cases the external validity can involve patients that would have been excluded in the original study.

[The N-of-1 trial: the ideal study design that is underused]. / De N = 1-trial, de meest ideale onderzoeksopzet, die te weinig wordt gebruikt.

The randomised experiment in a single patient, the N-of-1 trial, is the best study design for demonstrating causality, for example between agent and effect. Despite this, this type of study is only encountered sporadically in medical journals. One reason for this is that even this type of design cannot definitively demonstrate causality, because different points in time are compared with one another. Moreover, the design is rather inefficient, since the results correspond with those from observation without randomisation, placebo control or blinding. Even so, the N-of-1 trial is the ultimate form of verification in, for example, the individualisation of treatment. For this reason, this form of study might be used more often.

Risk factors for extended-spectrum ß-lactamase-producing Escherichia coli urinary tract infection in the community in Denmark: a case-control study.

OBJECTIVE: To verify the role of proton pump inhibitors (PPI) and nitrofurantoin, which have appeared as novel risk factors for carriage of extended-spectrum ß-lactamase (ESBL) -producing Escherichia coli, as risk factors for ESBL E. coli urinary tract infection (UTI). We included known risk factors to ascertain whether our findings are comparable with those of previous studies. METHODS: Population-based case-control study including 339 cases with community-onset ESBL E. coli UTI in 2007-2012, 3390 non-ESBL E. coli UTI controls and 3390 population controls. We investigated potential risk factors by estimating ORs and 95% CIs adjusting for sex, age and co-morbidity. RESULTS: Comparing cases with non-ESBL E. coli UTI, PPI use yielded an OR of 1.6 (95% CI 1.2-2.0) and antibiotic exposure gave an OR of 1.4 (95% CI 1.1-1.8); these were driven by nitrofurantoin (OR 1.8; 95% CI 1.3-2.6) and macrolides (OR 1.7; 95% CI 1.2-2.3). Other risk factors included previous hospitalization with one or two and more than two hospitalizations versus none yielding ORs of 1.9 (95% CI 1.4-2.5) and 4.6 (95% CI 3.2-6.8), recent surgery (OR 2.0; 95% CI 1.5-2.8), renal disease (OR 2.2; 95% CI 1.4-3.4), chronic pulmonary disease (OR 1.4; 95% CI 1.0-2.0) and cancer (OR 1.5; 95% CI 1.1-2.1). Comparing cases with population controls, we found that most risk factors were also risk factors for non-ESBL UTI. CONCLUSIONS: ESBL E. coli UTI were associated with previous hospitalization and surgery. Nitrofurantoin and macrolides augmented the risk. PPIs had a moderate effect but may be important facilitators of ESBL carriage due to their widespread use.

[Levels of evidence are insufficient]. / Niveaus van bewijskracht schieten tekort.

Medical knowledge is based on various types of research, each with its own 'indication' and 'contraindication'. Although the randomised controlled trial is highly useful to quantify small differences in treatment effects, it is not able to establish all medical knowledge needed at the bedside. In most instances of non-therapeutic research, observational data are more useful. Evaluating all medical knowledge on the basis of a single hierarchy of level of evidence is therefore seriously flawed.

[The N-of-1 trial: the ideal study design that is underused]. / De N = 1-trial, de meest ideale onderzoeksopzet, die te weinig wordt gebruikt.

The randomised experiment in a single patient, the N-of-1 trial, is the best study design for demonstrating causality, for example between agent and effect. Despite this, this type of study is only encountered sporadically in medical journals. One reason for this is that even this type of design cannot definitively demonstrate causality, because different points in time are compared with one another. Moreover, the design is rather inefficient, since the results correspond with those from observation without randomisation, placebo control or blinding. Even so, the N-of-1 trial is the ultimate form of verification in, for example, the individualisation of treatment. For this reason, this form of study might be used more often.

[No prior informed consent required for 'no-risk' scientific research]. / Geen voorafgaande 'informed consent' vereist bij wetenschappelijk onderzoek zonder risico of belasting.

Recent research has shown that prior informed consent not only leads to a lower response but also distorts the results of the study. The concern about the protection of the individual may not be given preference over the importance of participation in a study, in particular for those studies that carry a very limited or non-existent risk or that are not burdensome to the patient.

[Physicians and scientific research: slight decline of the numbers of physicians with a doctoral degree]. / Artsen en wetenschappelijk onderzoek: lichte teruggang van het aantal gepromoveerde artsen.

OBJECTIVE: To establish whether the number of physicians interested in a career in academia (i.e. research) is declining. DESIGN: Descriptive. METHOD: The researchers analysed the pre- and post-doctoral careers of PhD students at 3 university medical centres (VU Amsterdam, Nijmegen and Maastricht) in 4 separate reference years (1989, 1994, 1999 and 2003), using information from doctoral dissertations and the Dutch medical address book. The researchers recorded the gender of the students and the timing of the doctorate in relation to specialist training, university education and employment, as applicable. RESULTS: The total number of dissertations produced at the 3 medical faculties in the 4 reference years increased gradually by nearly a factor of 2 (1989: 112; 1994: 152; 1999: 198; 2003: 213). In terms of absolute numbers, the number of dissertations authored by physicians increased from 1989 to 1994 and again in 1999 (64, 90 and 105), but decreased slightly in 2003 (96). The percentage of female physicians obtaining a doctorate doubled during this period (1989: 9/64 (14); 2003: 28/96 (29)). Increasingly, physicians prepared their dissertation before or during their training as specialists or general practitioners (1989: 15/64 (23%); 2003: 51/96 (53%)). Ofthe clinical specialists who had received their doctorate, approximately half continued to work in an academic setting after obtaining their degree. This percentage remained approximately the same in all reference years (1989: 13/26 (50); 1994:19/35 (54); 1999: 21/45 (47); 2003: 21/40 (53)). CONCLUSION: Although the number of physicians performing scientific research as part of their doctoral degree project declined slightly in 2003 following an initial rise, our data indicate no cause for major concern. One reason may be increased interest in Clinical Research Fellow programmes. However, the future of medical research would look brighter if young physicians with doctorates had better career prospects within academic centres. To follow the academic careers of clinicians in The Netherlands, a national registry is needed to collect the type of data analysed in this study continually.

Meta-analysis of the risk of gastric stump cancer: detection of high risk patient subsets for stomach cancer after remote partial gastrectomy for benign conditions.

Controversy about gastric cancer risk after partial gastrectomy exists, especially in the United States. Therefore, we performed a meta-analysis to determine overall relative risk and weighted mean relative risk for subsets of postgastrectomy patients, define possible high risk patients suitable for surveillance, and assess for publication bias which would overestimate risk. If 2 studies were excluded because of heterogeneity, overall relative risk (RR) for gastric stump cancer in 22 studies analyzed was 1.66 [95% confidence limits (CL), 1.54-1.79]. With these 2 studies included, the RR summarized with a random effects model to account for study heterogeneity was 1.46 (95% CL, 1.18-1.82). No obvious evidence of publication bias was detected. Patients 15 years or more postoperative had a weighted mean RR of 1.48 (95% CL, 1.31-1.67) and patients 5-14 years postoperative had a RR of 0.91 (95% CL, 0.71-1.17) (P = 0.026). Patients operated upon for gastric ulcer had a weighted mean RR of 2.12 (95% CL, 1.73-2.59) and patients with duodenal ulcers had a RR of 0.84 (95% CL, 0.66-1.05) (P = 0.001). The weighted mean RR for females was 1.79 (95% CL, 1.39-2.29) and for males 1.43 (95% CL, 1.27-1.62) (P = 0.074). For Billroth II gastrectomy the weighted mean RR was 1.60 (95% CL, 1.15-2.18) and for Billroth I gastrectomy 1.20 (95% CL, 1.01-1.42) (P = 0.220). Although differences in risk between subsets of postagastrectomy patients seen to exist, recommendations concerning endoscopic surveillance await further studies of cost-benefit analysis.

[Snippets of past medical research: worth saving?]. / Snippers uit een medisch onderzoeksverleden.

Medical researchers and physicians are not trained in the care of their written or digital past. Here, a scientific historian and a clinical epidemiologist reflect on possibilities for archiving the records of medical research in order to safeguard scientific legacies. In addition to the use of so-called witness seminars, which may suffer from interpretation by 'hindsight', archival material is necessary to understand and interpret the past. A particular problem is how to establish archives of day-to-day scientific undertakings that rely almost entirely on digital media for measurements, communication and publication. The recently developed conviction that good scientific practice encompasses an obligation to store all relevant information about medical research projects at the time of publication - for future replication or verification - might dovetail with the goals of medical historians, and thus might become a rich source of historical data in the future.

Female hormones and thrombosis.

Exogenous hormones are used by more than a hundred million women worldwide as oral contraceptives or for postmenopausal hormone replacement. Oral contraceptives increase the risk of venous thrombosis, of myocardial infarction, and of stroke. The risk is highest during the first year of use. The venous thrombotic risk of oral contraceptives is high among women with coagulation abnormalities and with so-called third-generation contraceptives (containing desogestrel or gestodene). The risk of myocardial infarction does not appear to depend on coagulation abnormalities or the type of oral contraceptive. Hormone replacement therapy increases the risk of venous thrombosis. This risk is also highest in the first year of use and among women with coagulation abnormalities. The risk becomes very high in women with a previous venous thrombosis. Randomized trials have not confirmed a beneficial effect of postmenopausal hormones on the occurrence of myocardial infarction.

Nonulcer dyspepsia in a Dutch working population and Helicobacter pylori. Ulcer history as an explanation of an apparent association.

BACKGROUND: There is considerable debate about whether Helicobacter pylori infection is important in causing nonulcer dyspepsia. Many studies assessing this relationship have been performed in selected patient populations and included patients with a history of peptic ulcer. General population-based data with attention to ulcer history are needed to clarify this relationship. METHODS: A questionnaire on ulcer history and dyspeptic symptoms during the preceding 3-month period was obtained from apparently healthy employees who underwent a periodic medical examination in the Netherlands. In addition, serum samples were analyzed for anti-H pylori IgG antibodies. RESULTS: A total of 427 men and 73 women, aged 22 to 69 years, participated in the study. None of the women but 27 men (6%) had a previous diagnosis of peptic ulcer. Among 19 unoperated-on men with verified duodenal (17 subjects) and gastric (two subjects) ulcer, 89% were H pylori positive, while 74% had frequent dyspeptic symptoms in the 3 months before the study. Among the 400 men and 73 women without an ulcer history, the 3-month period prevalence of frequent dyspepsia was 13% and 21%, respectively. The rate of H pylori positivity was 25% in subjects with nonulcer dyspepsia and 29% in all others. The H pylori infection rate increased with age and with a lower occupational level but was independent of gender. In the male population, various differences in symptoms between H pylori-positive and H pylori-negative subjects could be detected when the 27 subjects with a history of ulcer were included, whereas these differences disappeared when these subjects were excluded. CONCLUSIONS: In the Dutch working population, nonulcer dyspepsia is not related to H pylori infection, whereas for duodenal ulcer the relationship is clear. The apparent association between dyspeptic symptoms and H pylori infection is entirely accounted for by subjects with an ulcer history.

Bleeding complications in oral anticoagulant therapy. An analysis of risk factors.

BACKGROUND: Insufficient data are available about the safety of oral anticoagulant therapy. The specialized organization of thrombosis services in the Netherlands can provide important information on the bleeding risk and various risk factors for bleeding in patients receiving oral anticoagulant therapy. METHODS: In a follow-up study over a 12-month period beginning in January 1988 on all patients treated by the Leiden Thrombosis Service, the frequency of bleeding complications was assessed. A Poisson regression model was used to assess the relative contribution to the bleeding risk of age, sex, target zone (intensity of anticoagulant effect aimed at), achieved intensity of anticoagulant therapy (International Normalized Ratio), and the type of coumarin derivative used. RESULTS: Six thousand eight hundred fourteen patients experienced 1003 bleeding complications (16.5 per 100 treatment-years), 162 of which were major bleeds (2.7 per 100 treatment-years). Bleeding increased significantly with age (32% increase for all bleeding, 46% for major bleeding for every 10-year increase in age in comparison with age < 40 years). Women had more minor bleeding complications than men, whereas both sexes experienced major bleeding in an equal frequency. There was no influence of target zone, while every one-point increase in International Normalized Ratio gave 42% more major bleeding (54% more regarding all bleeding). Use of acenocoumarol resulted in fewer bleeds (26% less regarding all bleeding and 46% less regarding major bleeding) than use of phenprocoumon. CONCLUSIONS: The risk of anticoagulant therapy in a routine, real-life situation is similar as in the setting of several well-organized clinical trials. The risk of bleeding complications rises significantly with age and with the achieved intensity of anticoagulation, and is dependent on the type of coumarin derivative that is used.

Usefulness of patient interview in bleeding disorders.

BACKGROUND: It is not known which questions in a medical interview are most informative for diagnosing mild bleeding disorders, and what the value is of the entire interview in screening for hemostatic disorders. METHODS: A questionnaire was sent to 222 patients with a proven bleeding disorder, to 134 patients suspected of a bleeding disorder but whose hemostasis proved normal, and to 341 healthy volunteers. A first comparison, between patients with a bleeding disorder and patients with bleeding complaints whose hemostasis proved normal, mimics the situation in a department of hematology where patients are referred because of complaints. The second comparison, between patients with a proven bleeding disorder and healthy volunteers, may serve as a model for the situation where the interview is used as a screening tool to detect patients with a bleeding disorder in a population where there is no prior suspicion, eg, before surgical intervention. For each question we calculated a univariate odds ratio, multivariate odds ratios, and a positive and negative likelihood ratio. With a receiver operating characteristic curve analysis we evaluated the value of a simple vs an elaborate interview. RESULTS: Ninety-two percent of the questionnaires were returned. For both comparisons the most informative questions were questions about bleeding disorders in the family and traumatic events, with the exception of delivery. Noninformative questions were frequent gumbleeds and blood in the urine. A receiver operating characteristic curve analysis revealed that a simple interview has a high discriminating power in a screening situation, whereas in a referred situation even an elaborate interview has a low performance. CONCLUSIONS: A simple interview is useful as a screening tool for the dentist or surgeon. In a specialized hematology center with referred patients, however, the interview is of little value in identifying patients with a bleeding disorder.

Higher risk of venous thrombosis during early use of oral contraceptives in women with inherited clotting defects.

BACKGROUND: Results of recent studies show that the risk for venous thrombosis is highest during initial oral contraceptive use. This suggests a subgroup of females who are at immediate risk of thrombosis when exposed to oral contraceptives. OBJECTIVE: To determine whether women with inherited clotting defects who use oral contraceptives develop venous thrombosis at an earlier stage than do those without inherited clotting defects. METHODS: Analysis of the data from the Leiden Thrombophilia Study, a population-based case-control study with data on duration of oral contraceptive use and recently detected genetic coagulation disorders. Patients had a first episode of objectively proven deep vein thrombosis. Patients and controls were considered thrombophilic when they had protein C deficiency, protein S deficiency, antithrombin deficiency, factor V Leiden mutation, or prothrombin 20210 A mutation. RESULTS: Risk of developing deep vein thrombosis was greatest in the first 6 months and the first year of oral contraceptive use. Compared with prolonged use, the risk of developing deep vein thrombosis was 3-fold higher in the first 6 months of use (95% confidence interval [CI], 0.6-14.8) and 2-fold higher in the first year of use (95% CI, 0.6-6.1). Patients who developed venous thrombosis in the early periods of use were more often thrombophilic. Among women with thrombophilia, the risk of developing deep vein thrombosis during the first 6 months of oral contraceptive use (compared with prolonged use) was increased 19-fold (95% CI, 1.9-175.7), and in the first year of use, it was increased 11-fold (95% CI, 2.1-57.3). CONCLUSIONS: Women with inherited clotting defects who use oral contraceptives develop venous thrombosis not only more often but also sooner than do those without inherited clotting defects. Venous thrombosis in the first period of oral contraceptive use might indicate the presence of an inherited clotting defect.

Risk of venous thrombosis with use of current low-dose oral contraceptives is not explained by diagnostic suspicion and referral bias.

BACKGROUND: The magnitude of the relative risk of venous thrombosis caused by low-dose oral contraceptive use is still debated because previous studies might have been affected by diagnostic suspicion and referral bias. METHODS: We conducted a case-control study in which the effect of diagnostic suspicion and referral bias was excluded. The study was performed in 2 diagnostic centers to which patients with clinically suspected deep vein thrombosis of the leg were referred. History of oral contraceptive use was obtained before objective testing for thrombosis. Young females with an objective diagnosis of deep vein thrombosis were considered case patients, and those who were referred with the same clinical suspicion but who had no thrombosis served as control subjects. Participants were seen between September 1, 1982, and October 18, 1995: 185 consecutive patients and 591 controls aged 15 to 49 years with a first episode of venous thrombosis and without malignant neoplasms, pregnancy, or known inherited clotting defects. RESULTS: The overall odds ratio for oral contraceptive use was 3.2 (95% confidence interval [CI], 2.3-4.5); after adjustment for age, family history of venous thrombosis, calendar time, and center, the odds ratio was 3.9 (95% CI, 2.6-5.7). In the idiopathic group (120 patients and 413 controls, excluding recent surgery, trauma, or immobilization), the odds ratio for oral contraceptive use was 3.8 (95% CI, 2.5-5.9); after adjustment, the odds ratio was 5.0 (95% CI, 3.1-8.2). CONCLUSIONS: In this study, in which patients and controls were subj ect to the same referral and diagnostic procedures, we found similar relative risk estimates for oral contraceptive use as in previous studies. We conclude that diagnostic suspicion and referral bias did not play an important role in previous studies and that the risk of venous thrombosis with use of current brands of oral contraceptives still exists.