Assessment of Hepatitis B Virus DNA Stability in Serum by the Chiron Quantiplex Branched-DNA Assay

Abstract

Quantification of hepatitis B virus (HBV) DNA in serum is used to establish eligibility for treatment and to monitor therapeutic response. With the trend toward centralized testing, defining the conditions that preserve sample integrity is of paramount importance. We therefore evaluated the stability of HBV DNA in 26 previously frozen (PF) and 5 fresh, never previously frozen serum specimens. PF specimens, covering a 3-log₁₀ HBV DNA dynamic range, were thawed and stored at −70, 4, 23, 37, and 45°C (±1.5°C) for 0, 24, 72, and 120 h (±2 h) and were refrozen at −70°C prior to testing. Five fresh specimens were split into two groups. Both group FG1 and group FG2 specimens were handled as described above; however, group FG1 specimens were subsequently maintained at 4°C and were never frozen prior to testing. Linear regression analysis of PF specimens demonstrated no significant HBV DNA degradation at ≤4°C over 5 days; however, HBV DNA levels decreased by 1.8, 3.4, and 20% per day at 23, 37, and 45°C, respectively. Three independent statistical methods confirmed that the probability of specimen failure, defined as a loss of 20% or more of HBV DNA and/or coagulation of serum, was lowest at ≤4°C and increased with temperature. Because only 10 to 20% of individual patient specimens demonstrated losses of HBV DNA of ≥20% at 23 or 37°C, sufficient numbers of serum specimens must be evaluated to determine overall statistical trends. In conclusion, HBV DNA integrity in separated serum specimens is preserved for at least 5 days when the specimens are stored at −70 or 4°C.

Quantification of hepatitis B virus (HBV) DNA in serum is used to determine eligibility for antiviral therapy and to monitor treatment response. With the trend toward centralized testing of HBV DNA, it is important to define the shipping and storage conditions that preserve specimen integrity. This information can minimize the risk of HBV DNA degradation, can ensure that specimens are properly handled within laboratories, and may reduce handling costs. We therefore evaluated the stability of HBV DNA in serum specimens stored at different temperatures for various lengths of time.

Assessment of the effect of serum storage conditions on the ability to detect an analyte such as HBV DNA quantitatively requires the following: (i) an assay capable of generating a reproducible relationship between the quantity of the analyte and its output signal; (ii) the measurement of the quantity of HBV DNA in sufficient numbers of specimens to ensure that the study has sufficient statistical power to demonstrate that changes in the quantity of HBV DNA reflect the effect of specimen storage and/or handling and not inter- or intra-assay variability, (iii) a descriptive endpoint that reflects a clinically relevant change in the quantity of the analyte which can be reliably measured, and (iv) appropriate statistical analysis of the data.

HBV DNA levels were measured by the Chiron Quantiplex HBV DNA assay, which is based on branched-DNA (bDNA) technology, because it has the widest dynamic range of the commercially available HBV DNA assays and it can reliably detect small (twofold) changes in HBV DNA concentration (2, 7). We used several statistical methods to examine the stability of HBV DNA in serum and evaluated sufficient numbers of specimens to ensure that the statistical power of our analysis could reliably detect changes in HBV DNA stability. First, linear regression was used to estimate the change in the quantity of HBV DNA stored at the various temperatures over time. Subsequently, three different statistical methods (Kaplan-Meier, two-way probability table, and logistic regression analyses) were used to estimate the probability of specimen failure, defined as a loss of 20% or more of HBV DNA and/or coagulation of serum.

(This study was presented in part at the 97th General Meeting of the American Society for Microbiology, Miami Beach, Fla., 4 to 8 May 1997 [9a].)

MATERIALS AND METHODS

Specimens.

HBV DNA-positive sera were obtained from routine clinical samples or from patients involved in clinical trials. All sera were separated from the clot within 4 h of collection. Both previously frozen (PF; n = 26) and fresh, never previously frozen (n = 5) specimens were evaluated.

PF specimens were thawed, divided into 50- to 100-μl aliquots, and incubated at −70, 4, 23, 37, and 45°C (±1.5°C) for 0, 24, 72, and 120 h (±2 h). Following incubation, the specimens were centrifuged at 14,000 rpm for 1 min in an Eppendorf Microfuge (catalog no. 5415C) to collect the condensate and were frozen at −70°C until assayed. PF specimens were independently tested by two laboratories. The Toronto Hospital (n = 20) and Covance Laboratories (n = 6; specimens from Covance Laboratories were not incubated at 45°C). All specimens and controls underwent the same number of freeze-thaw cycles prior to being assayed.

Fresh, never previously frozen specimens were evaluated only at The Toronto Hospital. These specimens were split into two groups. The first group (group FG1) of fresh, never previously frozen specimens was divided into 50- to 100-μl aliquots for incubation at 4, 23, 37, and 45°C (±1.5°C) for 0, 24, 72, and 120 h (±2 h). Following incubation, the aliquots were centrifuged as described above and were stored at 4°C until assayed. Matching aliquots of the fresh, never previously frozen specimens served as frozen controls for the first group (group FG2). Aliquots of FG2 specimens were incubated at the same temperatures and times as the FG1 specimens but were then frozen at −70°C prior to undergoing testing for HBV DNA. FG2 samples underwent only one freeze-thaw cycle. Both FG1 and FG2 specimens were assayed within 10 days of collection.

HBV DNA quantification.

The Quantiplex HBV DNA assay (Chiron Corporation), based on bDNA technology, was used according to the manufacturer’s instructions. The Chiron bDNA assay has been shown to be sensitive, specific, and linear over a nearly 4-log₁₀ quantification range (2, 7). The Chiron HBV bDNA assay demonstrates inter- and intrarun coefficients of variation of 10 to 15% and has been shown to reproducibly detect twofold changes in HBV DNA levels (7). All specimens were tested in duplicate, time zero controls were tested in quadruplicate, and the quantity of HBV DNA in each specimen was determined from a standard curve of HBV DNA run in parallel for each assay. To further enhance reproducibility, all of the specimens derived from an individual patient were assayed in the same assay run. Results were expressed as megaequivalents of HBV DNA per milliliter, with 1 Meq defined as the amount of HBV DNA which generates a level of light emission equivalent to that of 10⁶ copies of the HBV DNA standard (7).

Statistical methods.

The raw data were evaluated by examining the daily mean HBV DNA level at each temperature. Changes in the HBV DNA level at each temperature over time were evaluated by linear regression (18). The data from each incubation temperature were grouped into one category and analyzed as a specific or a fixed effect; i.e., any inferences from the data were considered specific to that particular temperature. In contrast, site- and patient-related differences in HBV DNA levels were analyzed as random effects, i.e., as inferences which are nonspecific to the particular patient or site in the study but which are applicable to any selection from the distribution of all possible sites and patients.

The probability of specimen failure was estimated by three methods: Kaplan-Meier, two-way probability table, and logistic regression analyses for PF specimens and by logistic regression analysis for FG1 and FG2 specimens. For PF, FG1, and FG2 samples, specimen failure was arbitrarily defined as the loss of 20% or more of the amount of HBV DNA quantified from the corresponding control specimen (either immediately frozen or stored at 4°C after collection) and/or the coagulation of serum.

To facilitate comparison of the estimates obtained by Kaplan-Meier analysis with those obtained by linear regression analysis, the data were categorized by temperature, and a simple moving time average over three adjacent values was obtained to minimize the impact of an isolated false specimen failure on the results obtained by Kaplan-Meier analysis. In time series signal processing, a moving average serves as a filter for high-frequency patterns, which are assumed to be noise. The moving average serially processes data for each time point, weighting the data for the current time point more heavily than those for adjacent time points; the sum of the weights typically are normalized to 1.

One-way table analysis is typically used to show the effects of different levels of a single factor. Similarly, two-way table analysis show the effect of combinations of levels for two factors and are frequently analyzed by using a chi-square statistic. Two-way probabilities were created by dividing the number of measured specimen failures by the total number of measurements in each temperature category.

Logistic regression analysis was performed in two stages (10). Each temperature level was considered to be a specific category or fixed effect, and both site- and patient-related variations in assay levels were considered to be random or nonspecific effects. For the fresh specimens, the probability of specimen failure was evaluated only by logistic regression because the sample size was too small for Kaplan-Meier and two-way table analyses. In all cases, a sufficient number of specimens was assessed to achieve at least a 95% confidence interval (CI) or its equivalent.

All the data were natural log transformed prior to analysis. Such transformation was necessary to ensure that the error was constant, because regression analysis assumes that error properties are the same over the dynamic range, whereas intrinsic assay error tends to be proportional to the viral load. Another advantage of natural log transformation is that the standard error in natural log units is directly related to the coefficient of variance of the raw units when the coefficient of variance is less than 50%.

RESULTS

PF specimens.

Twenty-six PF specimens were assessed. The quantity of HBV DNA in these specimens ranged from 1.57 to 5,029 Meq/ml, with a mean of 641 Meq/ml. The scatter in quantitation for each day is shown for each temperature category in Fig. 1A to D by plotting the quantity of HBV DNA (in megaequivalents per milliliter) on the y axis and time (in days) on the x axis. In order to illustrate the overall trends in the daily means, the y axis has been cropped such that four data points above 3,000 Meq are not shown. Day 0 has the highest number of datum points because the quantity of HBV DNA in the control samples was measured in quadruplicate and specimens that coagulated (which occurred only in some specimens incubated at 45°C) were not assessed. For Fig. 1D (45°C), only 20 of 26 PF specimens were tested (specimens tested at Covance Laboratories were not tested at 45°C). The specimens for this temperature category demonstrated a lower baseline mean HBV DNA level.

FIG. 1.

Scattergrams showing the range of quantitation of HBV DNA in PF specimens at each time point. The four incubation temperatures tested (±1.5°C) are shown in different plots: (A) 4°C; (B) 23°C; (C) 37°C; (D) 45°C. Each dot represents the observed quantity of HBV DNA (in megaequivalents per milliliter). Hollow dots indicate samples tested at the Toronto Hospital and solid dots indicate samples tested at Covance Laboratories. The horizontal line which bisects the diamonds represents the average of the data points at the respective time points, and the vertical apex of each diamond represents the 95% CI for the average value on each day (18). The horizontal line representing the average of the data points at time zero has been extended to day 5 in order to demonstrate the baseline average (panel B does not demonstrate the two high-value datum points present on panels A and C for days 1 and 3 because values for patient 23 were >3,000 Meq/ml at the 23°C incubation temperature).

While the scattergrams indicate trends in the data, they also show the inherent problem in drawing conclusions from the raw data; i.e., at certain temperatures, both site and patient variability obscure the overall statistical trend. Although the scattergram for the samples incubated at 4°C illustrates that the average trend is nearly zero, sample variability is evident as fluctuations in the daily mean. The scattergram for samples incubated at 23°C illustrates that the DNA loss is more consistent over time. The scattergrams of the samples incubated at 37 and 45°C show a comparatively greater loss of DNA, although mean sample variability is also greater.

Figure 2 illustrates the HBV DNA degradation trends at different incubation temperatures, as determined by linear regression analysis. HBV DNA levels decreased by approximately 1.8%/day at 23°C, 3.4%/day at 37°C, and 20%/day at 45°C. No significant change in HBV DNA levels was observed in PF serum specimens which were incubated at 4°C. HBV DNA levels decreased by approximately 0.006 ln Meq per day, which was not significantly different from the decrease in HBV DNA levels in the immediately frozen specimens. As the incubation temperature increased, the rate of HBV DNA loss demonstrated parallel increases. For example, in samples incubated at 23 and 37°C, daily decreases in HBV DNA levels were 0.015 and 0.010 ln Meq, respectively. Although the rate of HBV DNA loss from specimens incubated at 23 and 37°C was significantly greater than that from control samples, only 10 to 20% of individual patient specimens showed a ≥20% HBV DNA loss at these temperatures. Of note, the rates of HBV DNA loss at 23 and 37°C did not differ significantly. The highest rate of HBV DNA loss was observed in samples incubated at 45°C, which showed a decrease in HBV DNA levels of 0.246 ln Meq per day, and this was the only temperature at which specimen coagulation occurred.

FIG. 2.

HBV DNA degradation trends by linear regression analysis. Each line represents the linear regression at the indicated incubation temperature (±1.5°C). Percentages to the right of each line represent the amount of HBV DNA after 5 days compared to the initial amount. At 45°C approximately 20% of the HBV DNA is lost within 24 h.

The probability of specimen failure over time at each incubation temperature, as estimated by the Kaplan-Meier, two-way probability table, and logistic regression analyses, is indicated in Table 1. All three statistical methods unanimously demonstrated that incubation at the lowest temperature, 4°C, was associated with the lowest probability of specimen failure and that the probability of specimen failure increased with increasing temperature. The viral load at time zero was not correlated with the probability of specimen failure on the basis of an analysis performed after arbitrarily dividing specimens into groups with viral loads of 0 to 150, 150 to 800, and >800 Meq/ml.

TABLE 1.

Estimated probability of specimen failure

Specimen and analysis method	Estimated probability (%) of failure
	−70°C	4°C	23°C	37°C	45°C
PF specimensa
Kaplan-Meier	NAb	5	31	38	77
Two-way tables	NA	1.3	13	16	39
Logistic regression	NA	2	9	22	35
Fresh specimens
Logistic regression (never frozen)c	NA	0	20	34	50
Logistic regression (subsequently frozen)d	14	23	49	66	79

^aPF specimens were thawed, stored at the indicated temperature ± 1.5°C for 0, 24, 72, and 120 h (±2 h), and then refrozen at −70°C prior to testing for HBV DNA. All PF specimens underwent the same number of freeze-thaw cycles. 

^bNA, not applicable. 

^cFresh specimens stored at the indicated temperature ± 1.5°C for 24, 72, and 120 h (±2 h) and then stored at 4°C for up to 10 days until they were tested (group FG1). These fresh specimens were never frozen. 

^dFresh specimens stored at the indicated temperature ± 1.5°C for 24, 72, and 120 h (±2 h) and then frozen at −70°C for up to 10 days until they were tested (group FG2). These fresh specimens underwent one freeze-thaw cycle. 

As expected, each statistical method produced different estimates of specimen failure, with Kaplan-Meier analysis yielding a higher estimated risk of specimen failure (11) and two-way table analysis yielding a lower estimated risk of specimen failure. More realistic estimates of the probability of specimen failure were provided by logistic regression (10, 11).

Fresh specimens.

Five fresh, never previously frozen specimens containing between 0.70 and 73.58 Meq of HBV DNA per ml, with a mean of 42.79 Meq/ml, were also assessed. Specimens from FG1 were divided into aliquots and incubated at 4, 23, 37, or 45°C (±1.5°C) for 0, 24, 72, and 120 h (±2 h). Following the allotted incubation times, each specimen was either stored at 4°C for up to 10 days (FG1) or frozen at −70°C (FG2) prior to testing.

The changes in HBV DNA levels that occurred at each of the incubation temperatures over time were evaluated by linear regression. There was no significant change in the quantity of HBV DNA in specimens incubated and subsequently stored at 4°C (FG1) or in specimens that were immediately refrozen (FG2) after the incubation period.

In contrast, there were substantial decreases in the quantity of HBV DNA in all specimens incubated at the higher temperatures, and this was exacerbated by subsequent storage at −70°C. The greatest DNA losses were observed in specimens incubated at the higher temperatures (23, 37, and 45°C) for 120 h prior to being frozen at −70°C. HBV DNA levels decreased by 0.062, 0.070, and 0.124 ln Meq per day at the respective temperatures. There were also HBV DNA losses, albeit to a lesser extent, for the aliquots incubated at 23, 37, and 45°C followed by storage at 4°C, with HBV DNA levels decreasing by 0.033, 0.064, and 0.099 ln Meq per day, respectively.

The probability of specimen failure for stored fresh specimens was evaluated by logistic regression (Table 1). The probability of specimen failure did not reach significance for specimens maintained at 4°C for 120 h. However, the probability of specimen failure increased with increasing temperature, such that at the highest temperature, 45°C, there was a 50% probability of specimen failure for samples that had never been frozen. Of interest, the probability of specimen failure was higher for all fresh samples that were frozen at −70°C (FG2) than for specimens that had never been frozen (FG1).

DISCUSSION

Improvements in molecular biology-based technologies have dramatically enhanced the accuracy and reproducibility of quantitative viral nucleic acid detection in serum and plasma (5, 7, 20). Accurate viral load determinations are important because the viral load provides an indirect measure of the amount of viral replication in vivo and is used to help predict an individual’s response to therapy and/or clinical outcome (12, 15). In order to properly validate the relationship between HBV viral load and clinical disease and to monitor the effects of antiviral therapy, it is crucial to maintain specimen integrity during shipping and handling. To date, we are unaware of an adequate stability study that has evaluated the stability of HBV DNA in serum under various temperature and storage conditions.

Most published studies of nucleic acid stability in serum or plasma have focused on hepatitis C virus and human immunodeficiency virus. In many cases those studies were limited because they involved the use of nonstandardized assays, tested a limited number of samples, used different and sometimes poorly defined endpoints, and did not involve rigorous statistical analysis (1, 3, 16). While PCR-based assays are generally more sensitive than hybridization assays for nucleic acid quantification, in-house PCR assays tend to be very poorly standardized (9, 17, 21).

When investigators have used standardized, commercially available hybridization or PCR-based assays such as the Chiron bDNA or the Amplicor HCV Monitor assay (4, 6, 8, 13, 14), several factors have been shown to affect either specimen integrity or the ability of a given assay to accurately quantify the analyte. These factors include (i) the time between specimen collection and the separation of the plasma or serum, with short intervals of 4 to 6 h maximizing the amount of nucleic acid that can be detected in a given sample (4, 6, 13); (ii) the nature of the sample collection tube and/or the anticoagulant used (8, 14, 19); and (iii) the temperature and duration of storage. For example, once a specimen has been separated, storage at 4°C for a few days does not generally lead to a loss of the viral nucleic acid (3, 6, 8, 13).

Although this study did not assess the impact of the time between separation of the serum from the clot, because all specimens were separated within 4 h, we have defined the critical elements required for accurate assessment of analyte stability in clinical specimens. These elements include (i) the use of a standardized and reproducible assay, (ii) assessment of sufficient numbers of specimens to ensure statistical power, (iii) the use of a clinically relevant descriptive endpoint, and (iv) the use of appropriate statistical analysis.

Our data demonstrated that in individual specimens there was a marked variability in HBV DNA stability and that this variability was a function of the storage temperature and the viral load. We found that examination of the raw data alone could obscure the overall degradation trends, in part, because only 10 to 20% of individual patient specimens lose ≥20% of their HBV DNA when they are stored at temperatures above 4°C. Thus, it is critical that sufficient numbers of specimens be evaluated. We also found that logarithmic transformation of the raw data, which has the effect of linearizing the decay rate, enhanced the ability to assess the overall statistical trend.

Although the percentage of HBV DNA lost could be projected from the estimated daily rates of change at each temperature, a more meaningful estimation of the potential clinical implications was provided by the probability of specimen failure, which we arbitrarily defined as a 20% or greater loss of HBV DNA and/or coagulation of the specimen. Although it could be debated whether a 20% loss of HBV DNA is clinically relevant, we did test sufficient numbers of specimens to document a 20% loss of HBV DNA, and it is common laboratory practice to discard coagulated specimens (which occurred for only some of the specimens stored at 45°C). We then used three different statistical methods to assess the probability of specimen failure (i.e., Kaplan-Meier, two-way probability table, and logistic regression analyses), and all three statistical methods yielded concordant results, such that the lowest probability of specimen failure was observed at the lowest temperature and the probability of specimen failure increased with increasing temperature. Kaplan-Meier analysis produced the highest estimates for the probability of specimen failure. Results of the two-way table analysis, which indicated a lower probability of specimen failure than Kaplan-Meier analysis, provided a lower limit of HBV DNA loss. Logistic regression analysis provided intermediate and probably the most realistic estimates of specimen failure, primarily because it is a model-based analysis that uses intrinsic trends in the data (11).

A substantial number of patient specimens had to be tested in order to achieve adequate statistical power at the 95% CI assuming that measurements are taken at three to four time points for each temperature. Of the three methods, linear regression analysis required the least number of samples to achieve statistical power. In order to approximate the number of uncorrelated samples required to achieve a 95% CI, the following formula was used: N ≃ 8 (ɛ/Δ)², where N equals the number of patients, ɛ is the inherent error for each quantitation, and Δ is the smallest change that is detected with a 95% CI. Typically, a relative error of 80% would require the inclusion of five patients.

The power of two-way probability tables is based on the equation n ≃ 4p(1 − p)/(Δp)², where n is the total number of time points multiplied by the total number of patients, p is the estimated probability of specimen failure, and Δp represents the acceptable uncertainty of p. When the probability of specimen failure is 70%, which occurred at 45°C, the number of patients required to achieve adequate statistical power at the 95% CI is 24. The requirements of logistic regression fall between those of linear regression and two-way tables. For instance, if a 10% error rate in the probability of specimen failure is considered acceptable, a cohort of 8 to 10 patients would satisfy the requirements for adequate statistical power (95% CI) for all three statistical methods. The numbers of specimens in the cohort of frozen specimens far exceeded these criteria because a total of 26 specimens were used.

We selected PF specimens with quantities of HBV DNA over the entire range of detectability by the bDNA assay to assess HBV DNA stability but found no correlation between viral load at time zero and specimen failure. It is possible that specimens containing less than 0.7 Meq of HBV DNA per ml, the lower limit of quantitation of the Chiron bDNA assay, may not have the same stability profile. Such specimens will need to be evaluated when newer, more sensitive standardized assays become available.

We also evaluated five fresh serum specimens that had never been frozen prior to testing. From this limited data set, it appears that freezing somewhat alters the stability of HBV DNA in specimens once they have been maintained at ≥4°C for any length of time compared to the stability of HBV DNA in a matched set of unfrozen specimens. Further study of fresh specimens is warranted to confirm this observation. It is possible that freezing may cause immune complex and/or viral precipitation, which may affect the amount of HBV DNA that is subsequently detected in the clinical specimen after it has been frozen. This may be especially important when using as controls specimens which have never been frozen (6).

In summary, we have evaluated the stability of HBV DNA in PF and in fresh serum specimens that were incubated at various temperatures for discrete lengths of time and which were then stored at 4°C or frozen at −70°C prior to testing by the Chiron bDNA assay. The data support the fact that HBV DNA in separated serum stored at −70 or 4°C is stable for at least 5 days.