Documents
Resources
Learning Center
Upload
Plans & pricing Sign in
Sign Out

EPB PHC 6000 EPIDEMIOLOGY FALL_

VIEWS: 5 PAGES: 37

									          Unit 7:
Effect Measure Modification
 And Intervention Studies
Unit 7 Learning Objectives:
1. Understand the concept of “effect
   measure modification”.
2. Employ methods to investigate effect
   measure modification on both additive
   and multiplicative scales.
3. Recognize the differences between
   observational and experimental studies.
4. Distinguish between therapeutic and
   preventive intervention studies.
5. Understand design features of
   randomized clinical trials.
Unit 7 Learning Objectives:
6. Recognize ethical issues in clinical trials.
7. Recognize the role of Institutional Review
   Boards in clinical trials.
8. Understand the use of random allocation,
   factorial designs, and cross-over designs
   in experimental epidemiology.
9. Understand the use of blinding (masking)
   in the conduct of experimental studies.
10.Understand the impact of non-
   participation, compliance, and attrition of
   subjects in experimental studies.
         Assigned Readings:

Textbook (Gordis):
  Chapter 15, pages 233-238
  (Interaction)
  Chapter 7, Randomized Trials
  Chapter 8, Randomized Trials: some
  further issues

Reduction in the incidence of type 2
  diabetes with lifestyle intervention
  or metformin. New England Journal
  of Medicine 2002; 346:393-403.
Introduction to Effect
Measure Modification
   Effect Measure Modification
Effect Measure Modification: The
magnitude or direction of an association
varies according to levels of a third factor.

Also called:
     •     “Effect Modification”
     •     “Interaction”

Note: Unlike confounding, effect measure
modification should be described and
reported, rather than controlled.
        Effect Measure Modification
Hypothesis: High alcohol consumption is associated with
  larynx cancer (cohort study)

          D+      D-
                                 RR = (30 / 200) / (15 / 315)
   E+     30     170     200
                                     RR = 3.15
   E-     15     300     315
          45     470     515

 •Persons with high alcohol consumption appear
 to be at 3.15 times higher risk of developing
 larynx cancer than persons without high alcohol
 consumption. However, is this elevated risk
 similar among smokers and non-smokers?
         Effect Measure Modification
      NON-SMOKERS                       SMOKERS
         D+      D-                      D+        D-
E+        4      49         53    E+     26       121      147
E-        6      150        156   E-      9       150      159
         10      198        208          35       271      306
RR = (4 / 53) / (6 / 156)          RR = (26 / 147) / (9 / 159)

        RR = 1.96                          RR = 3.12

     Does smoking modify the relationship between
     alcohol consumption and larynx cancer?
      Effect Measure Modification
    CRUDE           STRATA 1        STRATA 2

    RRCA = 3.15     RRNS = 1.96     RRSM = 3.12

Unlike the assessment of confounding, the crude
estimate is NOT USED to evaluate the presence of
effect measure modification.
Instead, the stratum-specific estimates are
compared directly to see if they are different
(heterogeneous).

This example suggests “risk-ratio heterogeneity.”
    Effect Measure Modification

Keep in mind that the presence of “effect
measure modification” depends on which
measure of effect is evaluated (e.g. risk
difference, risk ratio, etc.).

The RD is on an additive scale.
The RR is on a multiplicative scale.

Let’s look at RD and RR separately.
       Effect Measure Modification
                         Risk     Exp.      Risk     Exp.
Smoke Alcohol   Incid.   Diff.   Additive   Ratio   Multipl.
 No     No      0.038    -----     -----    1.00      -----
 No     Yes     0.075    0.037     -----    1.96      -----
 Yes    No      0.057    0.018     -----    1.47      -----
 Yes    Yes     0.177    0.138    0.094     4.60     2.89


Expected additive = (0.075 + 0.057) – 0.038 = 0.094

Expected multiplicative = 1.96 x 1.47 = 2.89
       Effect Measure Modification
                         Risk    Expect.    Risk    Expect.
Smoke Alcohol   Incid.   Diff.   Additive   Ratio   Multipl.
 No     No      0.038    -----    -----     1.00     -----
 No     Yes     0.075    0.037     -----    1.96      -----
 Yes    No      0.057    0.018     -----    1.47      -----
 Yes    Yes     0.177    0.138    0.094     4.60     2.89


In this example, it appears that smoking modifies
(increases) both the risk difference and risk ratio
between alcohol consumption and larynx cancer.
       Effect Measure Modification
Hypothesis: Female gender is associated with depression
  (cohort study)
           D+       D-
   Fem     100     180    280 RR = (100 / 280) / (18 / 233)
   Male    18      215    233     RR = 4.62
           118     395    513


 • Females appear to be at 4.62 times higher risk
 of depression than males. However, is this
 elevated risk similar among young persons and
 older persons?
       Effect Measure Modification
       YOUNG                           OLD
       D+   D-                     D+      D-
F       6   48    54    F          94     132   226
M       6   144   150   M          12     71    83
       12   192   204              106    203   309
RD =                        RD =


RR =                    RR =
         Effect Measure Modification
         YOUNG                            OLD
        D+       D-                   D+      D-
 F        6      48      54    F      94     132        226
 M        6      144     150   M      12     71         83
         12      192     204          106    203        309
 RD = (6 / 54) - (6 / 150)      RD = (94 / 226) - (12 / 83)
RD = 0.111 – 0.040 = 0.071     RD = 0.416 – 0.145 = 0.271

 RR = (6 / 54) / (6 / 150)      RR = (94 / 226) / (12 / 83)
RR = 0.111 / 0.040 = 2.78       RR = 0.416 / 0.145 = 2.88
        Effect Measure Modification

                          Risk     Exp.      Risk     Exp.
Age     Gender   Incid.   Diff.   Additive   Ratio   Multipl.
Young   Male              -----     -----    1.00      -----
Young Female                        -----              -----
 Old    Male                        -----              -----
 Old    Female
      Effect Measure Modification
                        Risk     Exp.      Risk     Exp.
 Age Gender    Incid.   Diff.   Additive   Ratio   Multipl.
Young Male     0.040    -----    -----     1.00     -----
Young Female   0.111    0.071     -----    2.78      -----
Old    Male    0.145    0.105     -----    3.61      -----
Old   Female   0.416    0.376              10.40

 Expected additive =

 Expected multiplicative =
      Effect Measure Modification
                        Risk     Exp.      Risk     Exp.
 Age Gender    Incid.   Diff.   Additive   Ratio   Multipl.
Young Male     0.040    -----    -----     1.00     -----
Young Female   0.111    0.071     -----    2.78      -----
Old    Male    0.145    0.105     -----    3.61      -----
Old   Female   0.416    0.376    0.216     10.40    10.04


Expected additive = (0.111 + 0.145) – 0.040 = 0.216

Expected multiplicative = 2.78 x 3.61 = 10.04
       Effect Measure Modification
                         Risk     Exp.      Risk     Exp.
 Age Gender     Incid.   Diff.   Additive   Ratio   Multipl.
Young Male      0.040    -----    -----     1.00     -----
Young Female    0.111    0.071     -----    2.78      -----
 Old   Male     0.145    0.105     -----    3.61      -----
 Old   Female   0.416    0.376    0.216     10.40    10.04

In this example, older age modifies (increases) the
risk difference between gender and depression.
However, the risk ratio is not modified by older
age (no risk ratio heterogeneity).
         Effect Measure Modification
Hypothesis: Depression is associated with risk of hip
  fracture (cohort study)
            D+       D-
                                 RR = (40 / 220) / (30 / 245)
    E+      40      180    220
    E-      30      215    245        RR = 1.48

            70      395    465

  •Depressed persons appear to be at 1.48 times
  higher risk of hip fracture than non-depressed
  persons. However, is this elevated risk similar
  among persons with low and high body mass
  index (BMI)?
        Effect Measure Modification
        LOW BMI                   HIGH BMI
        D+   D-                  D+     D-
E+       6   50    56    E+      34    130   164
E-       6   144   150   E-      24    71    95
        12   194   206           58    201   259
 RD =                     RD =


 RR =                     RR =
         Effect Measure Modification
         LOW BMI                       HIGH BMI
        D+       D-                   D+        D-
 E+       6      50      56    E+     34       130      164
 E-       6      144     150   E-     24       71       95
         12      194     206          58       201      259
 RD = (6 / 56) - (6 / 150)      RD = (34 / 164) - (24 / 95)
RD = 0.107 – 0.040 = 0.067     RD = 0.207 – 0.253 = -0.045

 RR = (6 / 56) / (6 / 150)      RR = (34 / 164) / (24 / 95)
RR = 0.107 / 0.040 = 2.68       RR = 0.207 / 0.253 = 0.82
       Effect Measure Modification
                          Risk     Exp.      Risk     Exp.
BMI    Depress   Incid.   Diff.   Additive   Ratio   Multipl.
Low      No               -----    -----     1.00     -----
Low      Yes                        -----              -----
High     No                         -----              -----
High     Yes
       Effect Measure Modification
                          Risk     Exp.      Risk     Exp.
BMI    Depress   Incid.   Diff.   Additive   Ratio   Multipl.
Low      No      0.040    -----    -----     1.00     -----
Low      Yes     0.107    0.067     -----    2.68      -----
High     No      0.253    0.213     -----    6.32      -----
High     Yes     0.207    0.167              5.18


 Expected additive =

 Expected multiplicative =
       Effect Measure Modification
                          Risk     Exp.      Risk     Exp.
BMI    Depress   Incid.   Diff.   Additive   Ratio   Multipl.
Low      No      0.040    -----    -----     1.00     -----
 Low     Yes     0.107    0.067     -----    2.68      -----
High     No      0.253    0.213     -----    6.32      -----
High     Yes     0.207    0.167    0.320     5.18     16.92


Expected additive = (0.107 + 0.253) – 0.040 = 0.320

Expected multiplicative = 2.68 x 6.32 = 16.92
       Effect Measure Modification
                          Risk     Exp.      Risk     Exp.
BMI    Depress   Incid.   Diff.   Additive   Ratio   Multipl.
Low      No      0.040    -----    -----     1.00     -----
Low      Yes     0.107    0.067     -----    2.68      -----
High     No      0.253    0.213     -----    6.32      -----
High     Yes     0.207    0.167    0.320     5.18     16.92


In this example, it appears that high BMI modifies
(decreases) both the risk difference and risk ratio
between depression and risk of hip fracture.
      Effect Measure Modification
Axioms:
1. The presence of effect measure modification
   should be assessed by “eyeballing” the
   stratum-specific estimates to see if they differ.

2. Unlike confounding, which is a nuisance effect,
   effect measure modification represents useful
   information that should be explored and
   reported.

3. In the presence of effect measure modification,
   calculation and reporting of an overall (crude)
   effect is of dubious value, and is potentially
   misleading.
      Effect Measure Modification
Axioms:
4. Statistical tests of homogeneity of the stratum-
   specific estimates can be performed, but these
   tests are often underpowered – “eyeballing”
   the stratum-specific estimates is a better
   approach.

5. Be careful in the number of subgroups in which
   effect measure modification is investigated –
   each additional investigation increases the
   likelihood of a type I error (chance finding in
   which the null      hypothesis is erroneously
   rejected).
      Effect Measure Modification
Axioms:
6. Although some authors define effect measure
   modification (interaction) as any effect greater
   than additive, this is inappropriate since the
   stratum-specific estimates can differ in a non-
   additive fashion.

7. Any third variable has the potential to be:
   a) Confounder and effect modifier
   b) Confounder and not an effect modifier
   c) Not a confounder and an effect modifier

  Thus, there is no relationship between
  confounding and effect measure modification.
Intervention Studies
 Observational vs.
Experimental Studies


      REVIEW
       Observational Studies
     Investigator observes the natural
      course of events.
    – Documents who is exposed or non-
      exposed
    – Documents who has or has not
      developed the outcome of interest
               Experimental
          (Intervention) Studies

   Investigator allocates the exposure
         Therapeutic (Secondary Prevention)
         Prevention (Primary Prevention)
   Follow subjects to document
    subsequent development of disease
              Experimental
         (Intervention) Studies

   Therapeutic Trials - almost always
       conducted among individuals (e.g.
    clinical trial)
    Prevention Studies - may be conducted
    among individuals (e.g. field trial) or
    among entire populations (community
    trial)
             Therapeutic
             Clinical Trial

   Participants have a disease or
      condition
   Therapies are tested for safety and
       effectiveness (secondary
    prevention)
      Preventive Field Trial

    Participants (individuals) are free
      from the condition of interest
    Potential preventive treatments
       are tested -- can include healthy
       individuals at usual risk, or
    persons recognized to be at high
    risk    (primary prevention)
          Preventive
        Community Trial
    Entire communities are randomly
       allocated to treatments of
    interest
   Example: Newburgh-Kingston
      dental caries study

								
To top