Restrict Inference to Disease Outcome That Can Be Ascertained Accurately

Alcohol Free Forever

Stop Alcohol Cravings

Get Instant Access

Many, perhaps most, diseases have a spectrum of severity, with the probability of manifesting symptoms and being detected increasing as the severity of the condition increases. Prostate cancer in older men, for example, is extremely common and usually asymptomatic, but the form of the disease that is more aggressive and metastasizes to bone results in clinical symptoms that more often lead

Table 9.4. Multivariate-Adjusted Odds Ratios* for Colorectal Adenomas by Quintile of Fruit and Vegetable Intake for Women and Men, Minnesota Cancer Prevention Research Unit Case-Control Study, 1991-1994

FOOD GROUP QUINTILE

Mean Intake (Servings/week)

Cases vs. Colonoscopy-Negative Controls

Cases vs. Community Controls

Women

Men

Women

Men

WOMEN

MEN

OR

95% CI

OR

95% CI

OR

95% CI

OR

95% CI

Fruits

1

3.3

2.1

1.00

1.00

1.00

1.00

2

7.4

5.9

0.95

0.52, 1.72

0.79

0.46, 1.36

0.65

0.34, 1.25

0.73

0.43, 1.23

3

11.2

9.6

0.91

0.50, 1.63

1.06

0.61, 1.84

0.78

0.40, 1.52

1.00

0.59, 1.68

4

15.8

14.7

1.10

0.59, 2.05

0.79

0.44, 1.43

0.61

0.30, 1.20

0.62

0.36, 1.06

5

27.5

26.9

1.34

0.66, 2.69

0.66

0.35, 1.24

0.68

0.32, 1.43

0.75

0.41, 1.35

p trend

0.54

0.16

0.29

0.44

Vegetables

1

10.1

8.8

1.00

1.00

1.00

1.00

2

17.6

15.1

1.12

0.62, 2.01

1.29

0.75, 2.23

1.08

0.56, 2.07

0.67

0.39, 1.13

3

23.8

20.2

1.16

0.62, 2.16

1.11

0.64, 1.93

0.86

0.44, 1.68

0.73

0.43, 1.26

4

31.6

27.1

2.26

1.23, 4.14

1.30

0.72, 2.34

1.34

0.69, 2.59

0.59

0.34, 1.03

5

51.4

44.7

1.70

0.87, 3.34

0.90

0.48, 1.69

1.40

0.67, 2.92

0.55

0.30, 0.98

p trend

0.10

0.69

0.24

0.16

(continued)

Table 9.4. Multivariate-Adjusted Odds Ratios* for Colorectal Adenomas by Quintile of Fruit and Vegetable Intake for Women and Men, Minnesota Cancer Prevention Research Unit Case-Control Study, 1991-1994 (continued)

FOOD GROUP QUINTILE

Mean Intake (Servings/week)

Cases vs. Colonoscopy-Negative Controls

Cases vs. Community Controls

Women

Men

Women

Men

WOMEN

MEN

OR

95% CI

OR

95% CI

OR

95% CI

OR

95% CI

Juice

1

0.5

0.5

1.00

1.00

1.00

1.00

2

2.2

1.9

0.81

0.48, 1.39

1.53

0.86, 2.73

0.97

0.53, 1.78

1.16

0.67, 2.01

3

4.8

4.2

0.72

0.41, 1.27

1.24

0.73, 2.10

0.80

0.43, 1.51

0.83

0.51, 1.35

4

7.7

7.4

0.61

0.34, 1.09

0.88

0.52, 1.51

0.56

0.31, 1.03

0.75

0.45, 1.26

5

14.2

15.1

0.50

0.27, 0.92

0.98

0.55, 1.73

0.56

0.30, 1.06

0.97

0.56, 1.67

p trend

0.02

0.97

0.04

0.58

Total Fruits

and Vegetables

1

18.4

16.5

1.00

1.00

1.00

1.00

2

31.8

26.8

0.76

0.42, 1.38

0.80

0.46, 1.38

0.61

0.32, 1.18

0.76

0.45, 1.30

3

41.8

36.1

1.06

0.59, 1.92

1.05

0.60, 1.83

1.01

0.52, 1.94

0.95

0.56, 1.61

4

53.8

48.5

1.48

0.79, 2.78

0.82

0.44, 1.51

0.71

0.36, 1.38

0.46

0.27, 0.80

5

82.8

75.9

0.96

0.47, 1.96

0.61

0.31, 1.22

0.76

0.34, 1.66

0.60

0.32, 1.12

p trend

0.79

0.40

0.86

0.20

*Adjusted for age (continuous), energy intake (continuous), fat intake (continuous), body mass index (continuous), smoking status (never, current, former), alcohol status (non-drinker, former drinker, current drinkers consuming <1 drink/week, current drinkers consuming >1 drink/week), nonsteroidal antiinflammatory use (yes, no), multivitamin use (yes, no), and hormone replacement therapy use (yes, no in women only). OR, odds ratio; CI, confidence interval. Smith-Warner et al., 2002.

*Adjusted for age (continuous), energy intake (continuous), fat intake (continuous), body mass index (continuous), smoking status (never, current, former), alcohol status (non-drinker, former drinker, current drinkers consuming <1 drink/week, current drinkers consuming >1 drink/week), nonsteroidal antiinflammatory use (yes, no), multivitamin use (yes, no), and hormone replacement therapy use (yes, no in women only). OR, odds ratio; CI, confidence interval. Smith-Warner et al., 2002.

to an accurate diagnosis. Endometriosis, in which there is endometrial tissue located at abnormal anatomic locations in the abdominal cavity, is quite common and largely undetected, but the form of the disease that is more extensive appears to be more likely to produce symptoms that lead to diagnosis. Some of the confusion is semantic, i.e., whether disease is truly present when the requisite biologic changes have occurred but there are no overt symptoms and such symptoms may never arise. For studying etiology, however, the biologic entity itself is often of primary interest, so that variability in symptom occurrence, recognition, care-seeking, and ultimately disease diagnosis represent opportunities for misclassification. An accurate diagnosis often requires many steps, as discussed above, but the probability of all those events occurring is often highest when the disease is most severe in terms of stage or scope.

The most informative approach to the study of diseases with a spectrum of severity is to comprehensively identify all cases across the spectrum of disease, even the least severe, in order to determine empirically whether risk factors differ for more versus less severe variants of the condition. When comprehensive diagnosis is infeasible due to the invasiveness or expense of the methods for definitive diagnosis, e.g., laparoscopic surgery to diagnose endometriosis, compromise in the study design is required. Shifting interest to the study of severe endometriosis or aggressive prostate cancer is one strategy for conducting a valid study that is capable of comprehensive ascertainment, accepting the consequent inability to examine potential influences on less severe variants of the disease. That is, a more readily studied endpoint is substituted for the less feasibly studied one.

In a case-control study of prostate cancer, analyses were divided based on case aggressiveness to evaluate the potential implications of both selection bias (incomplete ascertainment of cases) as well as true biologic differences in etiology for more versus less aggressive tumors (Vogt et al., 2002). This multicenter case-control study was conducted in the late 1980s in Atlanta, Detroit, and 10 counties in New Jersey. Controls were chosen through random-digit dialing for men under age 65 and through the Health Care Financing Administration records for men age 65 and older. Among a much larger pool of participants, 209 cases and 228 controls had blood specimens analyzed for lycopenes and other specific types of carotenoids, antioxidants found in fruits and vegetables. For lycopenes (Table 9.5), the inverse association with prostate cancer risk was much more pronounced among aggressive cases as compared to nonaggressive cases (as defined in the footnote to the table), with odds ratios of 0.37 versus 0.79 in the highest compared to lowest quartile. For the other categories of carotenoids, however, differences were not striking or consistent in direction. Nevertheless, these data are helpful in considering the potential for a bias due to incomplete ascertainment of non-aggressive prostate cancer cases.

The spectrum of severity may be based on the actual size of an anatomic change, e.g., tumor size, the exact location of the pathologic alteration, e.g.,

Table 9.5. Prostate Cancer Odds Ratios for All Cases and for Nonaggressive and Aggressive Cases from a U.S. Multicenter Case-Control Study, 1986-1989

Quartile of Serum Carotenoid

ODDS RATIO (REFERENCE)^

RANGE ODDS RATIO RANGE ODDS RATIO

TEST FOR

ODDS RATIO P VALUE

a-Carotene (u,g/dl) 0.0-1.4 1.5-3.2 3.3-4.7 4.8-29.4

Nonaggressive cases§ 1.00 1.08 1.30 0.95 0.91

Aggressive cases 1.00 1.65 1.01 1.91 0.17

^-Carotene (u,g/dl) 0.3-8.2 8.3-14.8 14.9-23.1 23.2-117.5

Nonaggressive cases 1.00 1.16 1.48 1.54 0.20

Aggressive cases 1.00 1.57 1.34 1.61 0.41

j3-Cryptoxanthin (u,g/dl) 0.3-4.9 5.0-7.1 7.2-11.1 11.2-44.3

Nonaggressive cases 1.00 1.09 1.27 0.93 0.82

Aggressive cases 1.00 2.03 1.57 1.22 0.84

RANGE

RANGE

Lutein/zeaxanthin (u,g/dl) 3.6-14.1 All cases

Nonaggressive cases Aggressive cases

Lycopene (u,g/dl)

All cases

Nonaggressive cases Aggressive cases

tQuartile outpoints were based on the distribution for each exposure among controls. ¿All models were adjusted for age, race, study center, and month of blood draw.

§Among cases, distributions by stage and grade, respectively, were as follows: 146 localized, 23 regional, and 21 distant; 75 well differentiated, 70 moderately differentiated, 38 poorly differentiated, and one undifferentiated. Stage and grade were combined to form categories of disease aggressiveness. After the exclusion of 33 subjects because of missing information on grade and/or stage, "nonaggressive" disease included 111 cases with well- or moderately differentiated grade and localized stage. "Aggressive" disease included 65 cases with poorly differentiated to undifferentiated grade and/or regional to distant stage. This system of categorization seeks to distinguish between disease that is more versus less likely to progress and become fatal (Gann et al., Cancer Res 1999;59:1225-30).

whether it results in recognizable symptoms, or other features of the disease process. It is quite possible that these peculiarities of size and location are of no etiologic significance and represent random variations in a single disease entity that extends into the range in which misclassification is more common. If that is true, a study of diagnosable disease should yield information that is applicable to undiagnosable disease under the assumption that the diagnosable fraction represents incomplete underascertainment of the total spectrum of disease that is present. If the subsets of disease that are and are not recognizable have different etiologies, then of course, the study of recognizable disease will still be valid for the endpoint it examines, and a strong case can be made that the more severe version of the disease is the one that is more important to understand and ultimately prevent.

Pregnancy loss illustrates these issues nicely, since there is a spectrum of severity, ranging from very early loss that is unrecognizable other than through daily hormone assays, to the medically and emotionally severe outcome of a fetal loss in the second trimester of pregnancy. Reliance on clinically recognized or medically treated miscarriage is vulnerable to incomplete ascertainment relative to the universe of pregnancy losses, in that women will vary in recognizing symptoms of pregnancy and pregnancy loss and may or may not seek medical care in response to an uncomplicated pregnancy loss. On average, the more advanced the pregnancy at the time of the loss, the more likely it is to result in recognizing symptoms and seeking medical care. Study of risk factors for pregnancy loss using medical records for case identification (Savitz et al., 1994) is capable of considering only a part of the spectrum of pregnancy loss. Early losses (before 6 weeks) are subject to grossly incomplete ascertainment, and a rising proportion of losses will be identified through medical care up to approximately 12-15 weeks' gestation, at which time the ascertainment is thought to become virtually complete. A study based on medically treated spontaneous abortion may be excellent for pregnancies that survived to 14-20 weeks' gestation, acceptable for pregnancies that reached 10-14 weeks' gestation, and completely inadequate for pregnancies that ended prior to 10 weeks' gestation.

Was this article helpful?

0 0
Gaining Weight 101

Gaining Weight 101

Find out why long exhausting workouts may do more harm than good. Most of the body-building workout and diet regimens out there are designed for the guys that gain muscle and fat easily. They focus on eating less and working out more in order to cut the excess fat from their bodies while adding needed muscle tone.

Get My Free Ebook


Post a comment