High-Voltage Leak Detection of a Parenteral Proteinaceous Solution Product Packaged in Form-Fill-Seal Plastic Laminate Bags. Part 1. Method Development and Validation

High-Voltage and Sensitivity Set Points Selection:

An optimum test method requires that the high-voltage set point be set high enough to maximize electrical current reaching the ground, but not set so high that current arc from the electrode to ground occurs bypassing the package and generating a false-positive result. Once the optimum high-voltage set point is identified, an optimum range for the sensitivity set point is chosen. Based on the HVLD instrument manufacturer's recommendation, an optimum sensitivity range ensures voltage readings of approximately 3 V for no-defect packages (negative controls), while allowing for some set point adjustment in day-to-day instrument operation.

In the current study, it was quickly determined using negative controls that a high-voltage set point much greater than 13.0 kV for all six test methods risked current arc and should be avoided. Therefore, a series of tests was performed at high-voltage set points 12.0, 12.5, and 13.0 kV, at sensitivity set points adjusted in 5 unit increments to yield readings ranging from about 2 to 4 V. This process was repeated for each of the six test methods.

Test results for Method 1 are illustrated in Figure 9. All other test methods yielded similar results, demonstrating a strong linear correlation between sensitivity set point and voltage reading, at 12.0, 12.5, and 13.0 kV high-voltage set points. The linear regression analyses of voltage as a function of high-voltage and sensitivity set point for all six methods yielded regression coefficients ranging from 0.9695 to 0.9953. From these data, a high-voltage set point of 12.5kV was chosen. This 12.5 kV setting was low enough to avoid current arc yet was within a demonstrated optimal performance range, thus allowing for slight variation in actual high-voltage output.

• Download figure
• Open in new tab
• Download powerpoint

Figure 9

HVLD Test Method 1—Voltage as a function of sensitivity and high voltage set points.

The sensitivity set point window for each test method was identified by first determining the sensitivity (x-axis) along the 12.5 kV linear regression corresponding to a 3 V reading (y-axis). Bracketing this sensitivity by ±5 units provided the sensitivity set point lower and upper range limits, listed in Table I.

Ref Min and Ref Max Limits Selection:

Method development requires the establishment of appropriate pass/fail acceptance criteria for the HVLD voltage readings. This is done by repeatedly testing negative controls, employing optimized high-voltage and sensitivity parameters, to generate baseline voltage readings. These data are then statistically evaluated to predict both a lower limit (Ref Min), warning of system operation drift of test package misalignment; and an upper limit (Ref Max), the pass/fail voltage limit indicative of package leakage.

In the current study, population sets of negative controls were tested on each of three days (minimum) at the target high voltage set point (12.5 kV), and at the upper and lower sensitivity set points outlined in Table I, for each of the six test methods. Test packages included a subset supplied prefilled and sealed by Novo Nordisk A/S and a subset filled and sealed at WAL to account for product–package preparation variations. For a given method, the same package subsets were tested on each replicate test day. The total numbers of packages in each subset tested on each day by each method are outlined in Table II. With this negative control baseline information, the operational verification lower voltage limit (Ref Min) and the pass/fail voltage upper limit (Ref Max) for each test method were established.

A high-level examination of the voltage readings obtained provides an indication of test method repeatability and precision for no-leak bags. As shown in Table III, day-to-day mean voltage readings within a test package subset (WAL versus Novo Nordisk—Columns A and B, respectively) agreed to within 3 standard deviation units of the mean values. Most agreed within 1 standard deviation unit. Novo Nordisk–made bags' voltage readings obtained over all test days matched the mean WAL-made bags' voltage readings within 2 standard deviation units or less (Column C). These results infer that there is little noticeable day-to-day difference in HVLD test results for any test method, and there is little, if any, difference in HVLD test results for WAL-made versus Novo Nordisk–made bags.

Table IV includes an analysis of the voltage results from all bags (WAL-made plus Novo Nordisk–made) for each test method, pooling all data collected at both the lower and upper sensitivity range settings. As shown, the coefficient of variation for all negative controls tested according to Method 1 was 0.05. Similarly, the coefficients of variation for Test Methods 2 through 6 ranged from 0.05 to 0.06. Thus all HVLD test methods demonstrate similar degrees of precision when testing negative controls at 12.5 kV within the selected sensitivity ranges.

Table IV pooled subset data were also used to calculate Ref Min and Ref Max limits for each test method. In other words, for each method the mean voltage ±3 standard deviation units range was calculated; the lower limit of this range was deemed to be the Ref Min, and the upper range limit was Ref Max. For example, for Method 1, the mean ±3 standard deviation units range is 2.47 to 3.43 V. The lower limit (2.47 V) is the Method 1 Ref Min—the operational verification limit. HVLD test results equal to or less than this value warn of possible system operation drift or test package misalignment. The upper limit (3.43 V) is the Method 1 Ref Max—the pass/fail voltage limit. HVLD test results equal to or greater than this value are indicative of package leakage. The test parameters and reference limits for all six test methods are summarized in Table V.

Protocol:

All six test methods outlined in Table V were validated for robustness and limit of detection confirmation. For each test method, the protocol required a 50:50 random mix of positive and negative controls be tested on each of three replicate test days. The positive controls included holes approximately 3, 5, and 10 μm in nominal diameter—the number of units tested per day being equally divided among hole size group and hole site location. (Note: The actual defects in each hole size group ranged from 2.46 to 3.40 μm, 3.94 to 7.01 μm, and 8.31 to 11.17 μm for the entire positive control population. For hole site locations, refer to Figure 4.)

The number of test units for each Test Method 1 and 2 (employing bags with no metal clip) included 54 negative controls, plus 54 positive controls to account for all three hole size groups and six bag hole site locations. The number of test units for each of the Test Methods 3 through 6 (employing metal clipped bags) included 18 negative controls, plus 18 positive controls to account for the three hole size groups and the two bag hole site locations investigated. (Note: Metal-clipped positive controls had holes only at the septum end right and left bag corners, to check the impact of metal clips near these hole sites.)

Positive controls required for each of the three replicate test days were filled and sealed the day of testing to eliminate possible storage time effects and to minimize liquid product leakage loss. Each positive control was tested only once to eliminate possible hole size change from repeated HVLD exposure. [Note: Experiments performed prior to method validation found laser-drilled hole size widens as a function of HVLD exposure count. For example, holes averaging approximately 6.5 μm (±1.4) in nominal diameter widened to about 11.0 μm (±0.8) after 6 test exposures (n = 10).]

For each test method, all positive controls were tested at the lower sensitivity range set point (to maximize false-negative result risk), and all negative controls were tested using the upper sensitivity range set point (to maximize false-positive test result risk). Thus the robustness of each method at operational setting extremes could be confirmed.

The protocol acceptance criteria for each test method were as follows.

1. All negative controls must pass all tests.

2. The test method's limit of sensitivity will be based on the smallest hole size consistently detected as leaking (fails). All positive controls with holes equal to or larger than the demonstrated limit of sensitivity must fail all tests.