This article provides information, status, and current activities surrounding a Corrective and Preventative Action (CAPA) set forth by ActiGraph under CAPA number 16CPAN03.
Determining if you are impacted
This issue impacts wGT3X-BT and wActiSleep-BT (serial numbers beginning with MOS) devices shipped between November 7, 2013 and March 24, 2014 with serial numbers listed in the spreadsheet below. These devices are only impacted if they were used to collect data while running firmware version 1.4.0, 1.4.1, 1.5.0, 1.6.0 or 1.7.0. To verify your firmware version, simply connect the monitor to a computer running ActiGraph's ActiLife software and observe the firmware version column. If you are unsure whether or not the data you have collected in the past was collected with firmware version 1.4.0 or later, follow these instructions to check.
In addition, ActiGraph has developed a file scanning tool which can be used to scan *.agd and *.gt3x files quickly to discern whether or not those files are impacted by the issue. The tool can be downloaded here [wGT3X-BT, wActiSleep-BT Firmware Issue - File Scanner (16CPAN03)]. Instructions on how to use the tool can be found here.
During manufacturing, each ActiGraph activity monitor is calibrated to compensate for known variations in the sensitivity of the solid state accelerometers from unit to unit. The activity monitors listed in the referenced spreadsheet were calibrated during manufacturing (firmware version 1.0.0) using a specific accelerometer scale setting that our engineers later deemed to be less than optimal. In order to optimize the process, this calibration scale setting was changed with the release of firmware version 1.1.0 on 3/24/2014. To compensate for the 1.0.0 calibrated devices, our engineers placed a line of code in the firmware which allowed these prior monitors to continue to operate using the previous scale factor. With the release of version 1.4.0 firmware this code was removed inadvertently, causing these activity monitors to revert back to an incorrect calibration scale factor.
How did this happen?
Prior to the release of version 1.4.0 of our wGT3X-BT firmware, some changes that were made in the code-base of another product were inadvertently back-ported into the 1.4.0 code-base. Because of the subtlety of the impact on the data, this issue went unnoticed during our regression testing process. The code-review process failed to identify the issue because of the focus on the actual changes being made/improved in the wGT3X-BT firmware. To prevent this in the future, we will introduce both a full code-base review and unit testing in this specific portion of code into our firmware release process. In addition, we will intentionally scrutinize our output waveform sweep during testing with a tighter tolerance on the allowed variability in the signal response.
What has been done thus far?
The ActiGraph engineering team has spent considerable some time assessing the impact on user data. We have performed several empirical lab tests as well as some real-world activity test so that we would be able to accurately understand the impact on customers who are affected by this issue in advance of notifying customers. We have also spent some time examining the nature of the problem to determine the repairability of affected files.
How am I impacted?
If your data was collected with an affected monitor, both your summarized (epoch or *.agd) data and raw data will be slightly attenuated. According to our research, the average reduction across lifestyle-like activities is around 6% for epoch data, but that will vary depending on the types of activity.
The chart below provides an overview of the results of our epoch-level testing. The distances (57mm, 38mm, and 25mm) represent the radius of rotation of the test setup. The activity monitors were run through a sweep from 30RPM (0.5Hz) to 210RPM (3.5Hz) with a dwell time of 30s for each frequency. The percentage differences shown here are for 10s epoch counts. Test 1, Test 2, Test 3 and Test 4 were performed with one affected monitor compared to one unaffected monitor. Control 1 and control 2 tests results represent a comparison of two unaffected monitors.
What do I need to do?
At this time, our goal is to make you aware of the issue so that you are able to properly document and adjust any findings that you may have made or are currently making. We would also suggest downloading our file scan tool which has been specially designed to scan local directories containing *.agd and *.gt3x files to determine if any of those data are impacted by this issue. Instructions on using the file scan tool can be found here.
What will happen next?
ActiGraph engineers have fixed the firmware problem and have completed testing on a firmware release. However, releasing this firmware for public, automatic updates (the default setting in ActiLife) could possibly cause customers with ongoing studies to have data to containing both attenuated and non-attenuated data. For this reason, our plan is to hold this release for 30 days (until May 16, 2016) to allow affected users the opportunity to plan for the update. For more information, see the following topic.
What if I am currently collecting data in an active study?
If you are currently collecting data in an active study, be aware that the pending firmware release (scheduled for May 16, 2016) will cause your activity monitor to immediately correct the attenuation issue. For that reason, you can elect to continue collecting data with the existing firmware by disabling automatic firmware updates in ActiLife (instructions here). You will then have the ability to update the firmware on these monitors at a later time, thereby ensuring that your data collection remains consistent for your active study.
Can my data be fixed?
We are investigating techniques for repairing raw data files. Unfortunately, summary (epoch) data (*.agd or exports thereof) cannot be repaired directly due to the nature of the issue. If you have files which need to be repaired, click here to contact us directly (if hyperlink does not work, you can email us at email@example.com with subject "16CPAN03 - wGT3X-BT Firmware Issue Support Request").
What are the next steps?
ActiGraph will finalize testing on the firmware release fix which will be queued for release on May 16, 2016. In the meantime, we would encourage you to review your monitor list against the serial numbers listed in this spreadsheet to see if you are impacted.
May 16, 2016 - Firmware version 1.8.0 has been released! Updating your devices to this version of the firmware will correct all of the issues mentioned above moving forward. If you have disabled automatic firmware updates and would like to get the firmware file, please go here to find the download for manual implementation. If you require more information on how to manually update the firmware, please visit this help article.
Additionally, We are currently developing a small application which will use one of two of the following techniques to repair your files:
File scan technique - This technique will scan through your files to look for periods of non-movement from which the tool can estimate the calibration error. The tool will then correct the entire file by scaling each data point.
Hardware calibration technique - This technique will use the same process to correct the files as described in "1" above. However, rather than scanning the file, this technique will utilize measured offsets from the actual device used to collect the original data. This requires that you have the hardware available.
A couple of important notes about these two repair methods:
- Some files will not contain enough data to utilize the first technique.
- Neither technique is better than the other in recovering your data.
- For the second technique, we will provide you with instructions on how to gather the correct calibration data from your device. To do so, your monitor must be running firmware version 1.7.0 or older. If you believe you will use the second technique, we would recommend declining the pending 1.8.0 firmware release by disabling automatic firmware updating in ActiLife if you plan to use these devices in the near future, prior to performing the repair process.
We anticipate these tools to be tested and completed by 6/3/2016.
November 08, 2016 - Our publication "Scaling and Correcting Accelerometer data Affected by 16CPAN03" was rejected by Medicine & Science in Sports & Exercise (MSSE), but not necessarily in a negative context. On the advice of the MSSE Editor-in-Chief, we have elected to simply list the documentation on our website instead of publishing the article in a journal; as it does not technically constitute as a scholarly work or scientific contribution. The official documentation can be found here.
If you have any questions, concerns, or comments, please contact us through our support channels.