Our company developed the most widely used forehead SpO2 incorporated into the ARES (sometimes referred to as Watermark) home sleep testing device.  Pressure applied to the forehead/sensor is critical to 'limiting' the baseline shift.  Too much and too little pressure can impact SpO2 accuracy.  The algorithm used to adjust the intensity of red and infrared emitted light can make a difference as well.  Despite the baseline shifts, the forehead location has advantages including faster response to desaturations due to sleep disordered breathing (less circulation delay).  We conducted breath hold studies with the ARES, Nonin forehead and finger, and Nellcor and found the Nellcor to be really good.

Hi Daniel. Thanks for your answer. What Nellcor probe are you referring to? I presume it is not the max fast. And how many percentage points of spo2 baseline change can be expected in an ideal setup?

In our study we used the Nellcor 595 with its reflectance foehead sensor.  I have attached a unpublished report that explains what we did.  In Figure 1 you scan see there will not be a baseline shift under ideal conditions (i.e., subject was slightly inclined).  With a loose strap and when the patient shifts to a lateral position (when baseline shift is most impacted) I have seen >5% error.  The oximeter continues to respond to desaturation events so the changes are accurately detected but measures like the % of time below 90% can be impacted.  

There has been research related to improved reflectance accuracy using multiple LEDs.