This preview shows page 1. Sign up to view the full content.
Unformatted text preview: Observational Study Designs: Cross-sectional and Case Control
PH 150A: Introduction to Epidemiology Lisa F. Barcellos Lisa Barcellos Lecture 5 February 1, 2010
1 Readings for this lecture From Gordis: From Gordis Chapter 10 2 Objectives Disease of the day ????? Introduce methods of a cross-sectional study Introduce cross-sectional Understand when to use, pros and cons Introduce methods of a case-control study Introduce case-control Understand when to use, pros and cons 3 Cumulative Incidence
Number of new cases of outcome in Number new cases population at risk during a specified period of time Number of persons at risk for developing the disease at the beginning of time the beginning period Assumes entire population at risk for entire observation time Includes statement about time period 4 Prevalence
Number of persons with disease Number with in a population at a specific time Total population at risk at that time A “snapshot” of presence of disease (existing and new cases of disease,ie. prevailing cases) in a population at a given time A true proportion (numerator contained in the denominator) Synonyms: prevalence proportion 5 Components of Analytic Epidemiological Studies Study question Defined study population Defined study Information on study outcome Information study Information on study exposure(s) Information study Assessment of the association between Assessment between exposure and outcome. Conclusion
6 Defined Study Group Target population: population to which we want to apply study results. Study Group or Sample: a subset of target population assumed to represent the target population (best if randomly selected). Study proposal or published results should define and describe the eligibility criteria, define eligibility intended sample, final study group.
7 Conducting a Cross-sectional Study Describes relationship between exposure and outcome as they exist in population at a single point in time. Begin with defined study group Measure exposure and outcome Classify and Compare 8 Disease of the day???? http://www.sph.umich.edu/news_events/findings/fall05/features/one.htm 9 More Clues…. “It destroys families. It ruins careers. It ages patients prematurely, it is debilitating, progressive and relentless in its downhill course, as tough and worthy an opponent as any a doctor might choose to combat.” It is now a major public health issue. Some would argue it’s the major public health issue of this century.
10 Peter Kramer, Against Depression, 2005 http://www.sph.umich.edu/news_events/findings/fall05/features/one.htm 11 Mood Disorders Mood disorders include major depressive disorder, Mood major dysthymic disorder, and bipolar disorder. dysthymic disorder and bipolar Approximately 20.9 million American adults, or about 9.5 percent of the U.S. population age 18 and older in a given year, have a mood disorder. The median age of onset for mood disorders is 30 years. Depressive disorders often co-occur with anxiety disorders and substance abuse
12 http://www.nimh.nih.gov/health/publications/the-numbers-count-mentaldisorders-in-america/index.shtml Depression Illness that involves the body, mood, and thoughts Believed to be caused by abnormalities in levels of neurotransmitters Major depression: 4 or more symptoms last >2 weeks Dysthymia: a chronic low-grade depression SAD: Seasonal Affective Disorder Bipolar Disorder: mood swings (depression/mania)
http://www.mentalhealth.samhsa.gov/publications/allpubs/ken98-0049/default.asp#how, http://www.surgeongeneral.gov/library/mentalhealth/home.html 13 Signs and Symptoms of Depression Persistent sad, anxious, or "empty" mood Feeling hopeless, guilty, worthless, helpless Loss of interest or pleasure in hobbies and activities that were once enjoyed, including sex Decreased energy, fatigue, being "slowed down" Difficulty concentrating, remembering, making decisions Insomnia, early-morning awakening, or oversleeping Decreased appetite and/or weight loss or overeating and weight gain Thoughts of death or suicide; suicide attempts Restlessness, irritability Persistent physical symptoms that do not respond to treatment, such as headaches, digestive disorders, and chronic pain
14 Risk Factors for Depression Female sex Advanced age Lower socioeconomic status Recent stressful life experience Chronic (long-term) medical conditions Underlying emotional or personality disorder Substance abuse (such as alcohol, sleeping medications, medications for panic or anxiety, cocaine) Family history of depression, especially in a close relative (such as parent, brother or sister, or child) Lack of social support Obesity
http://www.emedicinehealth.com/articles/10289-2.asp 15 Major Depression
Major Depressive Disorder is the leading cause of disability in the U.S. for ages 15-44 Major depressive disorder affects approximately 14.8 million American adults, or about 6.7 percent of the U.S. population age 18 and older in a given year. While major depressive disorder can develop at any age, the median age at onset is 32.5 Major depressive disorder is more prevalent in women than in men. Ref: NIMH
16 Treatment Medication Psychotherapy Only 50% of those with depression are diagnosed and treated. 17 Research Question Are depression and obesity associated in the US 2001 Behavioral Risk Factor Surveillance Survey (BRFSS)? Hero et al. Depressive mood and obesity in US Hero et adults: comparison and moderation by sex, age and race. Int J Obesity, 15 November 2005. race. Obesity 18 1. Select a Study Group Behavioral Risk Factor Surveillance Survey (BRFSS): Random, probability sample of US adults carried out yearly by the federal Centers for Disease Control and Prevention (CDC) Uses Random digit dialing (RDD) and a computer assisted interview program. http://www.cdc.gov/brfss/about.htm To see how interviewers are trained, http://www.cdc.gov/brfss/training/interviewer/index .htm 19 Original 2001 BRFSS Sample Final Study Example 215,510 BRFSS Respondents 44,800 Excluded: 167,710 Note: 167,710 excluded due to depression question not included by state; 2738 refused or answered “don’t know” Characteristics of persons excluded: Higher BMI, Hispanics
20 2. Measure Exposure and Outcome Exposure (BMI): self-reported weight and height [BMI = weight / (height2)] height [BMI BMI < 25 normal weight, 25-29 overweight, ≥ 30 obese “During the past 30 days, for about how many days have you felt sad, blue or depressed?” Outcome (Depression) Both exposure and outcome measured at the same point in time in BRFSS
21 Cross-sectional Design
Begin Exposure Measure
Depressed All BRFSS Participants with Data > 14 Days = YES Classify/Compare Disease/Outcome
BMI < 25 BMI ≥ 25 BMI < 25 BMI ≥ 25 Depressed > 14 Days = NO Present 22 Association of Depression and Obesity in US Adults, all ages, both sexes and all races, 2001: Study Population
Depression Yes BMI < 25 Obesity a Unexposed Cases c Exposed Cases a+c All Cases No b Unexposed Controls d Exposed Controls b+d All Controls Totals a+b All Unexposed c+d All Exposed a + b + c+ d Total
23 BMI ≥ 25 BMI Totals Association of Depression and Obesity: Hypothetical Expected Distribution
Depression Yes BMI < 25 Obesity BMI ≥ 25 BMI a (300) c (300) a+c 600 No b (4,700) d (4,700) b+d 9,400 Totals a+b 5,000 c+d 5,000 a + b + c+ d 10,000
24 Totals Association of Depression and Obesity: Hypothetical Observed Distribution
Depression Yes BMI < 25 Obesity BMI ≥ 25 BMI a (150) c (450) a+c 600 No b (4,850) d (4,550) b+d 9,400 Totals a+b 5,000 c+d 5,000 a + b + c+ d 10,000
25 Totals Advantages of Cross-sectional Design Useful to describe health of populations, especially if based on a probability sample Associations found can be useful in generating hypotheses to study Efficient, quick, low subject burden because all measurements are made at the same time
26 Prevalence (per 100) of Depressive Mood in Men by Age and Level of Obesity (BMI) Significant OR (P<0.05) for BMI >= 30 compared to group with BMI < 25. 27 Disadvantages of Cross-sectional Design Can demonstrate associations between exposure and outcome, but cannot determine cause-effect. Cannot sort out existing (prevalent) outcomes from new (incident) ones. Like all studies, cross sectional studies can have problems in selecting subjects and making measurements.
28 Figure. Prevalence of early-life abuse by depression status. 29 Conducting a Case Control study Begin with cases of disease/condition Select controls Controls can be matched or not matched Measure exposure in the same way for both cases and controls Classify by exposure Compare exposure rates in cases and controls 30 Research Question: Is exposure to childhood abuse associated with major depression? Between 1995-97, a Boston population-based survey of women aged 36-45 years old was conducted. In 1999, 212 cases of depression and 542 controls were selected from the survey participants At that time, women completed a questionnaire about interpersonal relationships Wise et al., Adult onset of major depressive disorder in relation to early life violent victimization: a 31 case-control study Lancet 2001;358:881-87. 1. Select Cases Goal: Select cases representative of all people in the target population with the outcome Selected from a population-based study of >6000 women in Boston Case Definition: current major depression disorder using a structured clinical interview based on the DSM-4 (Diagnostic and Statistical Manual (Diagnostic based of Mental Disorders—DSM) criteria for depression (history of depression also measured)
32 Choose New or Existing Cases? Newly diagnosed (incident) cases describe factors associated with development of disease. Existing (prevalent) cases may tell as much about factors associated with survival as with development of disease. Most case control studies today use incident cases.
33 Other Possible Sources of Cases Disease Registries Surveillance systems Clinic or hospital records Birth or death certificates 34 Select Controls Goal: select controls representative of the general population in terms of potential for exposure to the risk factor Example: From the same population-based survey, select women who answered no to all depression symptoms 35 Selection Process Mailed questionnaire to 907 women who were in the population based survey (295 cases and in (295 612 controls) Of these, 732 (81%) completed the survey: – 236 (80%) cases – 496 (81%) controls 496 controls 68 women reported being abused as an adult – Should they be excluded? Why? Should
36 Other Possible Sources of Controls Random sample of target population Difficult to achieve—Can this be done with random digit dialing (RDD)? Hospital controls Beware: Sick people may not be representative of general population in terms of exposure to abuse as a child.
37 Enrolling Participants Identify Approach Introduce study Obtain informed consent Document characteristics of “non-responders” Document or refusals if possible Measure exposure in cases and controls
38 Matching A process used to select controls to be similar to cases on one or more characteristics that might dilute (or hide) the true relationship between exposure and outcome (these characteristics referred to as confounding variables). Equalizes comparisons between cases and controls for the matching factor. Removes that factor from further study.
39 2. Measure Exposure Goal: measure same way for cases and controls Exposure to abuse: survey of interpersonal relationships; asked women to recall violence or abuse as children or adolescents. YES response to question(s) regarding fear of (or experience of) physical or sexual abuse = exposed to childhood abuse
40 Methods for Measuring Exposure in Case Control studies Interview/Questionnaire Abstraction of medical records Biological samples (biomarkers) Direct observation 41 3. Classify and Compare Exposure in Cases and Controls
Depression Yes Exposure to Child Abuse Yes a Unexposed Cases c Exposed Cases a+c All Cases No b Unexposed Controls d Exposed Controls b+d All Controls Totals a+b All Unexposed c+d All Exposed a + b + c+ d Total
42 No Totals Case-Control Design
Classify and Compare Exposure Abuse +
Abuse – Begin with Outcome
212 Cases with Depression Abuse +
Abuse – 452 Controls without Depression Past Present
43 Advantages: Case Control Study Efficient in time and money because you don’t have to wait for disease to occur Only feasible method for rare diseases Can evaluate a wide range of exposures Useful for initial, exploratory studies 44 Disadvantages: Case Control Study Difficult to assemble case group that reflects all cases Example: Very ill cases don’t participate so missing information on these subgroups. Remedy: Use population-based case ascertainment Enroll as many cases as possible Use proxy-respondents for ill or deceased cases Describe characteristics of non-responders
45 Disadvantages: Case Control Study (cont.) Difficult to assemble a control group from same underlying population as cases Example: In a study of coffee and pancreatic cancer, the investigators chose diseased controls with digestive problems (who had stopped drinking coffee) Remedy: Aim for controls who approximate the level of exposure in the target population. Use multiple control groups from different sources. Describe number and characteristics of nonresponders.
46 Disadvantages: Case Control Study (cont.) Necessary information on exposure may not be available. Example: Biomarkers of exposure may be changed by disease. Remedy: Seek evidence of recorded values from past in medical records Determine if there are stored (e.g. frozen) tissue samples from the past that can be analyzed For questionnaire data, be very clear about the timing of exposure (e.g. before disease onset)
47 Disadvantages: Case Control Study (cont.) Participants forget or report exposures inaccurately due to passage of time Remedy: Use time line/calendar to facilitate recall. Test questionnaire against records in past to determine accuracy of recall (Ideal, not always possible) Search pre-existing medical or other records for verification 48 Disadvantages: Case Control Study (cont.) Cases may report exposure differently from controls Example: Case may search her memory (recall bias) Example: (recall or interviewer may question cases differently from controls (interview bias) controls (interview Remedy: Compare recalled exposures with records from the past Compare responses of cases and controls on “dummy” Compare variables. Use multiple control groups--some with “other disease Use outcomes”. “Blind” interviewers to case-control status Train/monitor interviewers to use consistent methods
49 Disadvantages: Case Control Study (cont.) Exposure measured after disease Example: Low micronutrient intake a year before diagnosis of cancer might reflect poor appetite during the early stage of the disease process, rather than a nutrient deficiency causing the cancer. Remedy: Be explicit about period of interest Attempt to confirm exposure with records Do not interpret case-control findings as cause-effect relationships.
50 Summary-Case Control Advantages: Efficient, relatively cheap, quick. Only feasible method for rare outcomes. Can evaluate wide range of exposures. Disadvantages: Difficult to select cases and controls without error. Retrospectively assessed exposure. Not able to assess cause-effect relationships. Cannot tell you about development of disease.
51 Thank You.
See you in lecture on Wednesday! 52 ...
View Full Document
This note was uploaded on 02/21/2010 for the course PH 150A taught by Professor Adams during the Spring '08 term at University of California, Berkeley.
- Spring '08