“…Five studies did not include some kind of adjustment for age either because they used a stepwise model building approach [36, 40, 62, 92] or restricted the age group studied [72]. Age was usually specified as a categorical variable (n=67), the choices of which varied considerably, but was also used as a linear term (n=9 [12, 29, 47, 59, 80, 81, 108, 87, 88, 95]), cubic spline (n=2 [60, 97]), a matching variable (n=3 [14, 64]) and in one instance as a quadratic term [26]. …”