Skip to main content
. 2018 Aug 28;9:916. doi: 10.3389/fphar.2018.00916

Table 4.

Association of positive, negative, and context-specific side effects with symptom relief.

Outcome = symptom relief

(1) (2) (3) (4)


Any {0,1} Percent of possible in category
Positive -1.100∗∗∗ -1.344∗∗∗ -2.345∗∗∗ -2.899∗∗∗
(-1.360 to -0.841) (-1.578 to -1.111) (-3.046 to -1.643) (-3.653 to -2.145)
Negative 0.174∗∗ 0.336∗∗∗ 2.311∗∗∗ 2.772∗∗∗
(0.015 to 0.334) (0.192 to 0.480) (1.461 to 3.161) (2.045 to 3.498)
Context-specific -0.339∗∗∗ -0.239∗∗∗ -0.781∗∗ -0.417
(-0.540 to -0.138) (-0.413 to -0.065) (-1.495 to -0.068) (-0.931 to 0.096)
Starting symptom level -0.660∗∗∗ -0.666∗∗∗
(-0.710 to -0.610) (-0.724 to -0.608)
Constant -2.307∗∗∗ 1.894∗∗∗ -3.098∗∗∗ 1.100∗∗∗
(-2.625 to -1.989) (1.441 to 2.348) (-3.372 to -2.824) (0.818 to 1.382)
Observations 10,535 10,535 10,535 10,535
R2 0.015 0.349 0.036 0.376

The first two columns measure use the existence of each category of side effect as independent variables, while the second two columns use the percent of possible in each category of side effects. The second and fourth columns include the starting symptom level. In all four regressions, the outcome is the change in symptom severity. Robust standard errors are clustered at the user level. Coefficients are reported with 95% Confidence Intervals below. ∗∗∗p < 0.01, ∗∗p < 0.05, p < 0.10.