For each right censored event, the device has performed for a period of time, after which its performance is unknown. Thus, only the time the device has undergone active surveillance is incorporated into the analyses. The following example is intended to provide an overview of the analysis process.

In Figure 1, the first patient’s device (serial number BH004134) operated for 32 months. At that time a product performance related event occurred. The fourth patient’s device (serial number BH008926) did not have an event but is censored because it was still in service and without product performance related events at the time of the analysis. This patient’s device had 66 months of implant experience. In this example, Figure 1 shows that 7 of the 16 devices had product performance events (gray bars), and 9 devices (blue bars) are censored.

Figure 1. Implant times for an individual device in 16 patients. Gray bars indicate devices removed from service due to a product performance event. Blue bars indicate right censored devices.

The first step in the life table method is to divide the time since implant into intervals of a specific length and determine how many devices entered each interval, how many were censored in each interval, and how many devices had events in each interval. This example will use 12-month intervals and determine a 60-month, or 5-year cumulative device survival estimate. For the first two 12-month intervals, all 16 devices survived and none were removed. In the 24-36 month interval, device BH004134 was removed due to an event. Therefore the table entries show that 16 entered the interval, none were censored, and 1 was removed due to a product performance event. For the 36-48 month interval, only 15 devices entered the interval and 1 was removed for a product performance event (device BH002883). For the 48-60 month interval, 14 devices entered the interval and 1 was removed for a product performance event (device BH006049). The device survival estimate for the first interval would be 16/16 = 100%. Likewise, the second interval would have a device survival estimate of 16/16 = 100%. The third interval would have a device survival estimate of 15/16 = 94%. The fourth interval from 36-48 months would have a device survival estimate of 14/15 = 93%. The fifth interval from 48-60 months would have a device survival estimate of 13/14 = 93%. In order to determine the overall risk from the first 5 intervals (also known as the device survival at 60 months), the interval specific estimates must be multiplied. The result of this multiplication is 100% * 100% * 94% * 93% * 93% = 81% cumulative device survival at 5 years.

Effective sample size or devices at risk for each interval is defined as the number of devices with full opportunity to experience a product performance event in the interval. Since censored devices are not fully followed throughout the interval, an adjustment must be made from the total number of devices that enter the interval. This is computed by subtracting one half the number censored in the interval from the number that entered the interval. This adjustment more accurately reflects the number of devices that could have experienced a product performance event than simply using the number that entered the interval. Using the number that enter an interval would over-estimate the sample size because the censored devices do not complete the interval. Completely ignoring the censored devices in the interval would under-estimate the sample size because censored devices would not be credited with their full service time. Using one half the number of censored devices effectively splits the difference. Expanding the example above to determine a 72 month, or 6-year device survival estimate, involves a censored device and adjusting the effective sample size. For the 60-72 month interval, 13 devices entered the interval and 1 was right censored (device BH008926) and 1 was removed for a product performance event (device BH 007561). The sixth interval from 60-72 months would have a device survival estimate of [13 – (0.5 * 1 censored event) – 2 total events] / [13 – (0.5 * 1 censored device)], or 10.5/12.5 = 84%. The 6-year cumulative device survival would be the 5-year cumulative device survival multiplied by the sixth interval device survival estimate, or 81% * 84% = 68%.

Active surveillance normally begins at the time of implant and continues until a product performance or censoring event occurs. In some cases in the Implantable Systems Performance Registry (ISPR) and more predominately in older device models, active surveillance of a device started well after the device was implanted. Because the device was not actively followed for some time after implant and before enrollment, this time should not be included in the analysis. For the ISPR, a method to incorporate data from these previously implanted devices was required that would appropriately adjust the follow up time. This method is called left censoring². Left censoring provides a statistical technique that uses data from existing devices while appropriately adjusting the device survival curves for the time the device was not actively followed in the registry.

Figure 2. Implant times for devices in 4 patients that were implanted prior to the device being enrolled in the ISPR. Green bars represent the time from implant to enrollment in the ISPR, or the time interval that is left censored. Blue bars represent the time since enrollment into the ISPR, or the time interval when active surveillance occurred. Gray bars indicate devices removed from service due to a product performance event.

For example, the first patient’s device (serial number BH004134) was implanted for 12 months prior to being enrolled in the ISPR. That period of time is left censored and is not included in the device survival analysis from 0 to 12 months. The period of active surveillance began at 12 months and the device enters the device survival curve at the 12-month time point. Thus, in some cases sample sizes may get larger from one time interval to the next interval.

Device survival for this report is evaluated over discrete 3-month intervals. For each interval, the effective number of devices that successfully functioned throughout the interval is divided by the number of devices that were at risk during the interval.  Cumulative device survival probability at any time point is obtained by multiplying the device survival probabilities of all intervals occurring prior to the time point of interest. A cumulative device survival curve is generated by plotting the cumulative device survival probability of all discrete intervals for which an adequate amount of data is present. The device survival curves shown are only presented where at least 20 total devices were still being followed in any given interval, except where otherwise noted.  Device survival estimates are presented at the device level, not at the system level which involves the combination of 2 or more devices.

Since device survival curves are derived from a sample of the total implanted population, they are only estimates of device survival.  The larger the effective sample size, the more confident the estimate. A confidence interval can be calculated to assess the confidence in an estimate. Confidence intervals for 1-year device survival estimates are shown at the end of each section. This can be roughly interpreted as meaning that the true survival of the device will fall somewhere in the interval, with 95% probability.

When confidence intervals for device models overlap, survival estimates for product performance related events are not statistically significantly different between models.  When confidence intervals do not overlap, survival estimates for product performance related events are statistically significantly different between models.