Examine Dose Response Gradients

If we are able to specify the etiologically relevant measure of the amount of exposure, and the amount of that exposure varies over the range that affects disease risk, then, by definition, more of that exposure will result in a greater probability of developing the disease. We will observe a dose-response gradient, in which increasing exposure results in an increasing risk of disease. The restrictions and uncertainties inherent in this evaluation should be recognized (Weiss, 1981). The critical aspect of the exposure that will yield increasing risk of disease may not be obvious, with the default approach based solely on some index of cumulative dose subject to uncertainty and error. What may be more important than the total amount of exposure is the form of the exposure, its biological availability, peak exposures, when it occurs, etc. A rather detailed understanding of the biologic process linking exposure and disease is required to quantify the relevant dose accurately. Furthermore, the shape of the dose-response function, if one is present at all, will vary across levels of exposure, potentially having a subthreshold range in which there is no response with increasing dose as well as ranges in which the maximum response has been attained and dose no longer matters. If the variation in the exposure levels that are available to study are all below or above the range in which disease risk responds to varying exposure, then no dose-response gradient will be found. Nonetheless, the potential value in identifying gradients in disease risk in relation to varying levels of exposure is always worthy of careful evaluation. When such a gradient is observed, it is informative and can support a causal hypothesis, but when it is absent, a causal association is by no means negated.

The hypothesized etiologic relationship under study should include at least a general specification of the type of exposure that would be expected to increase the risk of disease. The hypothesis and available data need to be scrutinized carefully for clues to the aspects of exposure that would be expected to generate stronger relations with disease. Total amount of exposure, generally measured as intensity integrated over time, is commonly used. Even in the absence of any measures of intensity, the duration of exposure may be relevant. More complex measures such as maximum intensity or average intensity over a given time period may also be considered.

For example, assume we are evaluating the relationship between oral contraceptive use and the development of breast cancer. If the hypothesis suggests that dose be defined by cumulative amount of unopposed estrogen, we would examine specific formulations of oral contraception and the duration of time over which those formulations were used. If the hypothesis concerns the suppression of ovulation, then the total number of months of use might be relevant. If the hypothesis concerned a permanent change in breast tissue brought about by oral contraceptive use prior to first childbirth, we would construct a different dose measure, one that is specific to the interval between becoming sexually active and first pregnancy.

In a study of oral contraceptives and breast cancer in young women (< 45 years of age), Brinton et al. (1995) examined several such dose measures. A mul-

tisite case-control study enrolled 1648 breast cancer cases and 1505 controls through random-digit dialing with complete data for analysis of oral contraceptive use. To examine varying combinations of duration of use and timing of use (by calendar time and by age), relative risks were calculated for a number of different dose measures (Table 8.7). Except for some indication of increased risk with more recent use, no striking patterns relative to duration and timing of oral contraceptive use were identified.

There is no universally applicable optimal measure because the specific features of the etiologic hypothesis lead to different ideal exposure measures. Multiple hypotheses of interest can be evaluated in the same study, each hypothesis leading to a specific exposure measure. In many cases, the various indices of exposure will be highly correlated with one another. On the positive side, even if we specify the wrong index of exposure (relative to the etiologically effective

Table 8.7. Relative Risks and 95% Confidence Intervals of Breast Cancer by Combined Measures of Oral Contraceptive Use Patterns: Women Younger Than 45 Years of Age, Multicenter Case-Control Study, 1990-1992

Used 6 months to < 5 years Used 5-9 years Used > 10 years

No. of Years Since First Use

< 15

136

1.44 (1.1-

-1.9)

66

1.55

(1.0-2.3)

22

1.27

(0.7-2.4)

15-19

221

1.14 (0.9-

-1.4)

120

1.45

(1.1-2.0)

93

1.58

(1.1-2.2)

> 20

292

1.29 (1.0-

-1.6)

190

1.10

(0.8-1.4)

119

1.11

(0.8-1.5)

No. of Years Since Last Use

<5

80

1.66

(1.1-

2.4)

87

1.49

(1.0-2.1)

131

1.37 (1.0-1.8)

5-9

66

1.28

(0.9-

-1.9)

71

1.49

(1.0-2.2)

66

1.13 (0.8-1.7)

> 10

503

1.21

(0.9-

-1.5)

218

1.14

(0.9-1.5)

37

1.34 (0.8-2.3)

Age at First

Use, Years

< 18

79

1.04

(0.7-

-1.5)

80

1.55

(1.1-2.2)

75

1.47 (1.0-2.2)

18-21

342

1.32

(1.1-

-1.6)

224

1.21

(0.9-1.5)

122

1.12 (0.8-1.5)

> 22

228

1.29

(1.0-

-1.6)

72

1.21

(0.8-1.8)

37

1.68 (0.9-3.0)

*Adjusted for study site, age, race, number of births, and age at first birth. All risks relative to women with no use of oral contraceptives or use for less than 6 months (389 patients and 431 control subjects).

RR, relative risk; CI, confidence interval. Brinton et al., 1995.

*Adjusted for study site, age, race, number of births, and age at first birth. All risks relative to women with no use of oral contraceptives or use for less than 6 months (389 patients and 431 control subjects).

RR, relative risk; CI, confidence interval. Brinton et al., 1995.

measure), it will be sufficiently correlated with the correct one to observe a dose-response gradient. On the other hand, correlations among candidate indices of exposure make it difficult or sometimes impossible to isolate the critical aspect of exposure.

10 Ways To Fight Off Cancer

10 Ways To Fight Off Cancer

Learning About 10 Ways Fight Off Cancer Can Have Amazing Benefits For Your Life The Best Tips On How To Keep This Killer At Bay Discovering that you or a loved one has cancer can be utterly terrifying. All the same, once you comprehend the causes of cancer and learn how to reverse those causes, you or your loved one may have more than a fighting chance of beating out cancer.

Get My Free Ebook


Post a comment