A Phantom Trend Inside the Control Limits
Valley Imaging operates a single Hologic Horizon A. At installation the spine phantom was scanned 30 times, giving a mean L1-L4 BMD of 1.042 g/cm2 with RMS-SD 0.006 g/cm2 (CV 0.6%). Policy is a daily phantom scan plotted on a Shewhart chart with control limits drawn at plus or minus 1.5% of the baseline mean. Over the last 10 weeks the values have read 1.038, 1.036, 1.034, 1.033, 1.031, 1.030, 1.029, 1.028, 1.027, and 1.026 g/cm2. Every point sits comfortably inside the plus or minus 1.5% limits. The technologist who runs QC leaves on maternity leave next week, and the lead technologist says, 'It's all within limits, so keep scanning patients and don't create a fuss.' A 2-year follow-up spine DXA is booked for this afternoon.
1. Using the numbers provided, decide whether these QC data represent a real problem. Which decision rules apply, and what is your conclusion?
Read the worked answer
Do not rely on the limit lines alone; apply pattern rules as well.
- Direction/tendency rule: ten consecutive points fall below the baseline mean and each is lower than the one before it. Most QC programs reject a run of six or more consecutive points on one side of the mean, or a monotonic trend of six or more points. This chart violates both.
- Magnitude in SD units: with RMS-SD 0.006 g/cm2, the drop from 1.042 to 1.026 g/cm2 is 0.016 g/cm2, about 2.7 baseline SDs. That is far outside what random precision alone would produce.
- The 1.5% limit is itself too loose: 1.5% of 1.042 is about 0.016 g/cm2, roughly 2.5 CVs. A limit that generous will hide real drift, which is exactly what has happened here.
Conclusion: this is statistically significant drift, not random variation. The chart has failed even though no single point crossed a line, and clinical scanning should not continue unexamined.
2. List, in priority order, the causes you would rule out before calling service, and how you would rule each one out.
Read the worked answer
Work from cheapest and most likely to most expensive:
- Phantom handling - confirm it is the same phantom, undamaged, correctly oriented and seated, and that nothing (pad, strap, patient belonging) is on the table above it. Scan it twice back to back; reproducibility within the tube tells you whether you have an operator problem or a machine problem.
- Analysis rather than acquisition - record the current software version and re-analyze the stored baseline phantom scan with today's software. If the historical curve shifts when reanalyzed, the trend is in the software/ROI, not the hardware. Verify the auto-analysis box placement on the phantom.
- Environment and setup - room temperature logs, table cleanliness, table height/position, whether a service or software update occurred at the start of the 10-week window.
- Hardware - a smooth monotonic decline with no operator explanation points to tube output, detector, or calibration drift. This is the engineer's call, and it should be made promptly.
Document each step with dates and values; the QC record is your defense if a clinical comparison is later questioned.
3. What do you do about this afternoon's follow-up scan, and what do you do about patients scanned during the drift?
Read the worked answer
- This afternoon: a phantom shift of about 1.5% is the same order of magnitude as a clinically meaningful spine change (LSC is commonly 2-3% at the spine). A true 2-3% treatment response could be erased, doubled, or invented by this drift. The safest course is to postpone the follow-up until the cause is identified and corrected. If the physician insists the scan cannot wait, acquire and archive it, but do not report a change from the prior study until the drift is resolved and the data are re-examined.
- Patients already scanned in the window: build a list of every exam in the last 10 weeks. If the cause proves to be analysis-side, reanalyze the stored scans and issue corrected reports. If the cause is hardware, flag any serial comparison made in that window and repeat the study after repair if a treatment decision depends on it.
- Communicate: inform the lead technologist and medical director in writing that the chart failed on a trend rule, not on a limit line, and record the corrective action. 'Within limits' is not synonymous with 'in control.'