Compared to what? Finding controls for case-control studies

doi:10.1016/S0140-6736(05)66379-9

The Lancet

Volume 365, Issue 9468, 16–22 April 2005, Pages 1429-1433

https://doi.org/10.1016/S0140-6736(05)66379-9 Get rights and content

Summary

Use of control (comparison) groups is a powerful research tool. In case-control studies, controls estimate the frequency of an exposure in the population under study. Controls can be taken from known or unknown study populations. A known group consists of a defined population observed over a period, such as passengers on a cruise ship. When the study group is known, a sample of the population can be used as controls. If no population roster exists, then techniques such as random-digit dialling can be used. Sometimes, however, the study group is unknown, for example, motor-vehicle crash victims brought to an emergency department, who may come from far away. In this situation, hospital controls, neighbourhood controls, and friend, associate, or relative controls can be used. In general, one well-selected control group is better than two or more. When the number of cases is small, the ratio of controls to cases can be raised to improve the ability to find important differences. Although no ideal control group exists, readers need to think carefully about how representative the controls are. Poor choice of controls can lead to both wrong results and possible medical harm.

Section snippets

Aim of controls

Controls in a case-control study, which progresses backwards in time from outcome to exposure,³ indicate the background frequency of an exposure in individuals who are free of the disease in question. Controls do not need to be healthy; inclusion of sick people is sometimes appropriate. Indeed, exclusion of ill people as controls can distort (bias)⁴ the results.⁵ (Like healthy individuals, ill people can develop a different condition of interest.) The final point is key: controls in a

Where to find controls?

The investigator (and, ultimately, the reader) needs to determine the group of individuals from which cases and controls will be drawn. A known group¹¹ consists of a defined population observed over a period (figure 2). This group might consist of passengers and crew on a week-long cruise of the Caribbean or all individuals living in Sweden over a decade. Cases are those who develop the disorder of interest and controls are those in the same group without the condition. Thus, case-control

Controls from a known group

When possible, random samples of people without the disease can serve as controls. Investigation of an outbreak of food-borne illness on a cruise ship generally uses a case-control approach. Cases are those who develop gastroenteritis; controls are those on board who do not. The study seeks to identify food exposures that are more common among the cases than the controls. Moreover, no one who had not eaten the suspect food should have become ill. On the ship, probability sampling among those

Neighbourhood controls

Neighbourhood controls generally are drawn in a specified pattern from the block in which the case lives. As always, selection of controls should be independent of the exposure of interest. To avoid selection bias, interviewers are given a specific pattern of houses to approach. Researchers have used two approaches to identify houses of controls: a population register or door-to-door canvassing.20, 21 A useful aid for the former is the cross-reference (also termed crisscross or reverse-street)

How many control groups?

Some authors have argued for using two separate control groups; if results are consistent, then findings are deemed more credible.14, 31 For example, a case-control study of oestrogen therapy and endometrial cancer used both hospital and community controls.³² In the unhappy circumstance of disparate results, however, which result should be ignored?¹⁵ Another immediate disadvantage is the added cost in time and resources. For example, in the case-control study of endometrial cancer cited above,³²

How many controls per case?

Readers are sometimes surprised to discover large disparities between the numbers of cases and controls in a case-control report; clinicians intuitively expect similar group sizes.³⁴ This inequality reflects attempts by investigators to increase the ability of the study to find differences of importance, should they exist. In unmatched case-control studies, having roughly equal numbers of cases and controls is most efficient if costs are similar for cases and controls. However, sometimes the

What to look for in controls

The validity of case-control studies depends on selection of appropriate control groups. Choosing controls might seem deceptively simple but it can be treacherous. Controls should reflect the background frequency of the exposure in the population. Hence, they should be similar in all important respects to cases, except that they do not have the disorder in question. Their selection must be independent of exposure.

When the study group of potential controls is known, a good approach is to take

References (39)

DA Grimes
Clinical research in ancient Babylon: methodologic insights from the book of Daniel
Obstet Gynecol
(1995)
KF Schulz et al.
Case-control studies: research in reverse
Lancet
(2002)
DL Sackett
Bias in analytic research
J Chronic Dis
(1979)
DA Grimes et al.
Cohort studies: marching towards outcomes
Lancet
(2002)
E Funkhouser et al.
Alternative strategies for selecting population controls: comparison of random digit dialing and targeted telephone calls
Ann Epidemiol
(2000)
LJ Vernick et al.
Selection of neighborhood controls: logistics and fieldwork
J Chronic Dis
(1984)
DA Grimes et al.
Bias and causal associations in observational research
Lancet
(2002)
DA Grimes
Intrauterine devices and pelvic inflammatory disease: recent developments
Contraception
(1987)
MA Ibrahim et al.
The case control study: the problem and the prospect
J Chronic Dis
(1979)
BS Hulka et al.
Estrogen and endometrial cancer: cases and two control groups from North Carolina
Am J Obstet Gynecol
(1980)

KF Schulz et al.

Unequal group sizes in randomised trials: guarding against guessing

Lancet

(2002)

DJ Kearney et al.

A case-control study of endoscopy and mortality from adenocarcinoma of the esophagus or gastric cardia in persons with GERD

Gastrointest Endosc

(2003)

DJ Wiebe

Firearms in US homes as a risk factor for unintentional gunshot fatality

Accid Anal Prev

(2003)

AR Feinstein

Clinical epidemiology: the architecture of clinical research

(1985)

JJ Marbach et al.

The control group conundrum in chronic pain case/control studies

Clin J Pain

(1992)

KJ Rothman

Modern epidemiology

(1986)

JJ Schlesselman

Case-control studies: design, conduct, analysis

(1982)

AT Chan

Aspirin, non-steroidal anti-inflammatory drugs and colorectal neoplasia: future challenges in chemoprevention

Cancer Causes Control

(2003)

JA Baron

Epidemiology of non-steroidal anti-inflammatory drugs and cancer

Prog Exp Tumor Res

(2003)

Cited by (316)

Infrarenal EVAR for Penetrating Aortic Ulcer: A Comparative Study with Abdominal Aortic Aneurysm
2023, Annals of Vascular Surgery
Citation Excerpt :
Two patient groups were clustered: group 1 (G1) includes all consecutive patients treated with EVAR for iPAU; the control group (group 2, G2) includes patients treated with EVAR for iAAA in the same time period. The patient selection of G2 patient was obtained randomly with specific randomization function (Excel 2007 [Microsoft Corporation, Redmond, WA, USA]: RANDBETWEEN [bottom-top0111983]) comparing each iPAU to a minimum number of 4 AAA to increase the chance of detecting important differences.11 The 2 groups were compared in terms of patient demographics and pre-operative risk factors.
Endovascular aortic repair (EVAR), currently the preferred treatment for abdominal aortic aneurysm (AAA), has been described also for penetrating aortic ulcers (PAU) of the infrarenal aorta. However, data on its performance in this particular setting are still sparse in the literature. Aim of this study is to compare patient clinical characteristics, aorto-iliac features, and post-operative outcomes between infrarenal PAU and AAA treated by standard EVAR.
In this retrospective observational case-control multicenter study, the patients treated for infrarenal PAU (G1) with EVAR in 2 high-volume European centers from January 2014 to December 2019 were prospectively entered into a dedicated database and retrospectively analyzed. A 4-fold control group (G2) of infrarenal AAA patients, homogeneous for age and gender, was also considered. Preoperative clinical characteristics, aorto-iliac features (rupture, aortic maximum diameter, proximal neck diameter and length, aortic bifurcation diameter, distance between the lowest renal artery and the aortic bifurcation [RA-AoBi], severe aortic calcification), technical success, 30-day (morbidity, reintervention, complications, mortality) and follow-up outcomes (freedom from reintervention [FFR] and survival) were compared in the 2 groups (chi square/Fisher exact test, t-student test, Mann-Whitney test, logistic regression and Kaplan-Meier analysis).
Seventy-three patients (age 78 ± 7 years; male 84.9%) were included in G1 and 299 (age 78.4 ± 6.6 years; male 89.3%) in G2. At the time of diagnosis, G1 patients were more often symptomatic compared with G2 (odds ratio OR 10.21, 95% confidence interval CI 4.17–24.99, P < 0.001). At preoperative computed tomography angiography, G1 patients had more ruptures (OR 8.11, 95% CI 3.50–18.78, P < 0.001), smaller maximum diameter (OR 1.05, 95% CI 1.03–1.08, P < 0.001), longer and narrower proximal neck (OR 0.97, 95% CI 0.95–0.99, P = 0.020 and OR 1.47, 95% CI 1.32–1.64, P < 0.001, respectively) narrower aortic bifurcation (OR 1.34, 95% CI 1.24–1.45, P < 0.001), lower RA-AoBi (OR 1.09, 95% CI 1.07–1.12, P < 0.001), and more severe aortic calcification (OR 57, 95% CI 16–198, P = 0.001). Technical success (G1 98.6% vs G2 95.7% P = 0.320), 30-day morbidity (G1 2.7% vs G2 8.7% P = 0.133), reintervention (G1 2.7% vs G2 2.3% P = 0.691), complications (G1 6.8% vs G2 8% P = 0.737) and mortality (G1 1.4% vs 2% P = 0.720) were comparable in the 2 groups. The mean follow-up was 17.7 ± 16.4 months in G1 and 18.8 ± 15.1 in G2 (P = 0.576). Late FFR and survival were comparable in the 2 groups (1-year FFR: G1 94.8% vs G2 97.5%, P = 0.995; 1-year survival: G1 91.7% vs G2 92.3%, P = 0.960).
Infrarenal PAU are more often symptomatic with a higher rupture rate compared to infrarenal AAA. Despite some negative anatomical characteristics (narrower aortic bifurcation, lower RA-AoBi, extensive calcification), the results of EVAR are extremely satisfactory in this setting, suggesting that endovascular exclusion could be considered a valid treatment for infrarenal PAU.
Involvement of inflammatory lipoproteinemia with idiopathic adhesive capsulitis accompanying subclinical hypothyroidism
2022, Journal of Shoulder and Elbow Surgery
Thyroid functional abnormalities are considered risk factors for idiopathic adhesive capsulitis (IAC) though that relationship remains uncertain. Although dyslipidemias are associated with IAC, no readily accessible study has reported associations between dyslipidemias and IAC patients with subclinical hypothyroidism. The purposes of this study were to investigate whether subclinical hypothyroidism is an independently associated factor for IAC and to determine the differences in prevalence of dyslipidemias between two groups of persons with subclinical hypothyroidism: one composed of IAC patients and the other of individuals without IAC.
This case-control study included a case group of 412 IAC patients without intrinsic shoulder lesions, extrinsic causes, or medication for thyroid dysfunction. The control group comprised 1236 age- and sex-matched persons seeking general checkups at the authors' health promotion center during the same period as the case group. Control subjects had normal shoulder function and no previously diagnosed adhesive capsulitis, no medication for thyroid dysfunction, and no history of trauma or of shoulder surgery. The studied variables were age, gender, obesity, diabetes, dyslipidemias, subclinical hypothyroidism, hypothyroidism, and hyperthyroidism. A conditional logistic regression analysis evaluated the matched sets of subjects to determine odds ratios and 95% confidence intervals for the studied variables. The differences in the prevalence of dyslipidemias between IAC patients with subclinical hypothyroidism and individuals with subclinical hypothyroidism but without IAC were determined with generalized estimating equations, using covariates of age, sex, and diabetes. The P values were set at < 0.05.
Subclinical hypothyroidism (odds ratio, 2.10; 95% confidence interval, 1.36-3.15; P = .001) was significantly associated with IAC. Patients with IAC and subclinical hypothyroidism had a significantly higher prevalence of hyper–low-density lipoproteinemia, an inflammatory lipoproteinemia, than individuals with subclinical hypothyroidism but without IAC (P = .002).
Subclinical hypothyroidism is significantly associated with IAC. Hyper–low-density lipoproteinemia, an inflammatory lipoproteinemia, is involved in IAC accompanied by subclinical hypothyroidism.
Evaluating the Quality of Case-control Studies involving the Association between Tobacco Exposure and Diseases in a Chinese Population based on the Newcastle-Ottawa Scale and Post-hoc Power
2022, Biomedical and Environmental Sciences
Short- and long-term impact of cancer on employment and financial outcomes of adolescents and young adults (AYAs): a large population-based case-control registry study in the Netherlands
2022, ESMO Open
Adolescent and young adult (AYA) cancer survivors, 18-39 years at initial cancer diagnosis, often self-report negative consequences of cancer (treatment) for their career. Less is known, however, about the objective impact of cancer on employment and financial outcomes. This study examines the employment and financial outcomes of AYA cancer survivors with nationwide population-based registry data and compares the outcomes of AYAs with cancer with an age- and sex-matched control population at year of diagnosis, 1 year later (short-term) and 5 years later (long-term).
A total of 2527 AYAs, diagnosed in 2013 with any invasive tumor type and who survived for 5 years, were identified from the Netherlands Cancer Registry (clinical and demographic data) and linked to Statistics Netherlands (demographic, employment and financial data). AYAs were matched 1 : 4 with a control population based on age and sex (10 108 controls). Analyses included descriptive statistics, chi-square tests, independent samples t-tests, McNemar tests and logistic regression.
AYA cancer survivors were significantly less often employed compared with their controls 1 year (76.1% versus 79.5%, P < 0.001) and 5 years (79.3% versus 83.5%, P < 0.001) after diagnosis, and received more often disability benefits (9.9% versus 3.1% 1 year after diagnosis, P < 0.001; 11.2% versus 3.8% 5 years after diagnosis, P < 0.001). Unemployed AYAs were more often diagnosed with higher disease stages (P < 0.001), treated with chemotherapy (P < 0.001), radiotherapy (P < 0.001) or hormone therapy (P < 0.05) and less often with local surgery (P < 0.05) compared with employed AYAs 1 and 5 years after diagnosis.
Based on objective, nationwide, population-based registry data, AYAs’ employment and financial outcomes are significantly affected compared with age- and sex-matched controls, both short and long-term after cancer diagnosis. Providing support regarding employment and financial outcomes from diagnosis onwards may help AYAs finding their way (back) into society.
Effectiveness of trivalent influenza vaccines against hospitalizations due to laboratory-confirmed influenza a in the elderly: Comparison of test-negative design with register-based designs
2022, Vaccine
Measuring influenza vaccine effectiveness (IVE) seasonally is important and has been conducted utilizing several observational study designs. The active test-negative design has been most widely used and the validity of passive register-based studies has been debated. We aimed to explore the potential differences, advantages, and weaknesses of different study designs in estimating influenza vaccine effectiveness.
We compared three study designs in estimating IVE against hospitalization in the elderly aged 65 years or more over three influenza seasons 2015/16, 2016/17 and 2017/18. Designs compared were active test-negative design (TND), register-based cohort design and register-based case-control design with different selection criteria for cases and controls.
Adjusted IVE estimates for the three consecutive seasons 2015–18 in active test-negative design were 82% (95% confidence interval 26, 96), 21% (-179, 77), 15% (-113, 66). For case-control design, estimates from different analyses ranged in 2015/16 from 47% (-16, 76) to 52% (-48, 84), in 2016/17 from 10% (-42, 43) to 29% (-20, 58), and in 2017/18 from −27% (-91, 15) to 1% (-40, 30). In the cohort design, the adjusted IVE estimates were 48% (-9, 75), 29% (1, 49), 13% (-21, 37) for the three seasons.
The register-based cohort design produced results more concordant with the active test-negative design than the case-control design. Furthermore, the register-based cohort design yielded most precise estimates with narrower confidence intervals. In Finland with the availability of near real-time nationwide register data, the register-based cohort design is the method of choice to continue the annual surveillance of influenza vaccine effectiveness.
Body sway and movement strategies for control of postural stability in people with spinocerebellar ataxia type 3: A cross-sectional study
2022, Clinical Biomechanics
Postural instability with an excessive body sway is a disabling manifestation in spinocerebellar ataxia type 3. Whether the larger body sway reflects distinct movement strategies for postural control remains uncertain. This study compared the control of postural stability of people with spinocerebellar ataxia type 3 with healthy subjects using body sway and movement strategy analyses derived from bi- and three-dimensional posturography.
Twenty-three patients (7 men, 16 women, 47 ± 11 years) and 102 healthy participants (34 men, 68 women; 44 ± 22 years) underwent posturography while standing with eyes open/closed tasks. Postural stability was assessed using elliptical area and average velocity of body sway. Spatial patterns (single-, double-, or multi-centered) were derived from the number of high-density regions in the three-dimensional statokinesigram.
Repeated measures two-way analysis-of-variance showed a vision-by-group interaction effect for area (F_1,122 = 28.831, P < 0.001, η² = 0.037) and velocity (F_1,123 = 59.367, P < 0.001, η² = 0.073); sway area and velocity were higher in spinocerebellar ataxia type 3 and increased under eyes-closed condition, with a higher increase in the spinocerebellar ataxia type 3. A main effect for group (F_1,123 = 11.702, P < 0.001, η² = 0.061) but not vision (F_1,123 = 2.257, P = 0.136, η² = 0.005) was found for the number of high-density regions. Spatial patterns were different between groups under trials with eyes closed (χ²_2,125 = 7.46, P = 0.023) but not open (χ²_2,125 = 2.026, P = 0.363), with a shift from single- to double- or multi-centered spatial patterns.
Compared to healthy subjects, a larger body displacement and velocity in spinocerebellar ataxia type 3, mainly under visual constraints, are not related to the predominance of either ankle or hip movement strategies.

View all citing articles on Scopus

View full text

SeriesCompared to what? Finding controls for case-control studies

Summary

Section snippets

Aim of controls

Where to find controls?

Controls from a known group

Neighbourhood controls

How many control groups?

How many controls per case?

What to look for in controls

Obstet Gynecol

Lancet

J Chronic Dis

Lancet

Ann Epidemiol

J Chronic Dis

Lancet

Contraception

J Chronic Dis

Am J Obstet Gynecol

Lancet

Gastrointest Endosc

Accid Anal Prev

Clinical epidemiology: the architecture of clinical research

The control group conundrum in chronic pain case/control studies

Clin J Pain

Modern epidemiology

Case-control studies: design, conduct, analysis

Aspirin, non-steroidal anti-inflammatory drugs and colorectal neoplasia: future challenges in chemoprevention

Cancer Causes Control

Epidemiology of non-steroidal anti-inflammatory drugs and cancer

Prog Exp Tumor Res

Series
Compared to what? Finding controls for case-control studies