INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

There are 89.9 million cases of genital Chlamydia trachomatis infection every year worldwide (13), 4.5 million of which occur in the United States (4). Although many C. trachomatis infections are asymptomatic (16), the sequelae from infection, including pelvic inflammatory disease (PID), and infertility, represent a large burden for populations worldwide. Furthermore, inflammatory sexually transmitted diseases, such as those caused by C. trachomatis, increase the risk of both human immunodeficiency virus (HIV) transmission and infection (7, 11). Together, the high percentage of asymptomatic infections, the sequelae of infections, and the increased association with HIV transmission underscore the importance of screening as a necessary intervention to reduce the burden of diseases caused by C. trachomatis.

Detection of genital C. trachomatis infection by ligase chain reaction (LCR) with first-void urine is a noninvasive, highly sensitive, and highly specific procedure (2, 8). Although the cost of LCR is higher than that of other tests such as direct fluorescent antibody, antigen detection by enzyme immunoassay, and nucleic acid probe tests, LCR is more sensitive and more specific (15, 17). Culture has been considered to be the "gold standard" in the past but costs more and is less sensitive than either LCR or PCR (3, 5, 10, 12, 14).

Pooling serum samples for HIV testing was found to be accurate and has been used to reduce the cost of enzyme-linked immunosorbent assays for detection of antibody to HIV (1, 6). Pooling for HIV testing has been used to develop both population estimates and, in a multiple-step procedure, to determine which individual sample is positive. Pooling has also been applied to the PCR detection of C. trachomatis in endocervical and urethral scrapes (9), but in that study the sample size was small. The investigators acknowledged the need for subsequent studies to rule out the possibility of reduced sensitivity by diluting out individual specimens in the pool.

The screening of women at risk for C. trachomatis has been recommended by the Institute of Medicine as a cost-effective program which would prevent the high cost of untreated infections which lead to PID (4). As a screening and treatment intervention reduces the prevalence of C. trachomatis infection over time, the cost per specimen tested with the pooling protocol algorithm would be further decreased. The reduction in price occurs for two reasons: (i) as prevalence decreases, pooling a greater number of samples increases cost savings and (ii) the samples from fewer pools would test presumptive positive such that fewer samples would be retested individually. Therefore, the cost for finding one case does not increase dramatically as prevalence decreases, as is the case when samples are tested individually.

In this study we examined the accuracy and cost-saving ability of pooling urine specimens for the detection of genital C. trachomatis infections by LCR. A cost analysis of the pooling protocol was conducted to determine the number of specimens it would be necessary to pool in order to provide the highest cost savings, taking into account the prevalence of infection in the population screened.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Sample size and parameters. As part of an ongoing study to determine chlamydia prevalence in asymptomatic U.S. Army females with a mean age (± standard deviation [SD]) of 22 (±4 years), urine samples were tested by LCR to ascertain genital C. trachomatis infection. A sample of 568 processed urine specimens was pooled by 4 into 142 pools, and 520 specimens were pooled by 10 into 52 pools. Pools were formed by order of consecutive laboratory accession number. All 1,088 pooled urine samples were also tested individually. For all discrepant individual and pool results, both the individual samples and the pools were retested to confirm results.

Urine specimen, collection, preparation and assay setup. Specimen, collection, preparation, and assay setups were performed according to the manufacturer's instructions for the urine-based chlamydia LCR assay (Abbott Laboratories, Abbott Park, Ill.).

DNA amplification and detection. Unit dose tubes containing DNA preparations were amplified under the following conditions: 40 cycles of denaturation at 93°C for 1 s, annealing at 59°C for 1 s, extension at 62°C for 1 min, 10 s, and soaking at 25°C in an LCR thermocycler (Abbott Laboratories). Amplified DNA was detected in an LCR-automated machine which performed a particle-based enzyme immunoassay with a fluorescent signal. For individually tested samples, a sample-to-cutoff ratio (S/CO) of 1.0 was considered positive, and borderline negative samples (0.80 to 0.99 S/CO) were retested, according to manufacturer's instructions.

Pilot study. Because the volume for each individual urine specimen is decreased in the pooled assay, a pilot study was conducted to determine an appropriate S/CO for the pooled assays. The desired S/CO would detect all positive pools while not detecting most, if not all, negative pools. The pilot study consisted of 148 processed urine samples from the ongoing study of female U.S. Army recruits. The technician, blinded to the individual test results, pooled and tested these 148 samples by four. By lowering the S/CO from 1.0 to 0.2, all of the positive pools were detected (100%) (25 of 25) and only 2.7% (1 of 37) of the negative pools tested presumptive positive. Since all pools which test positive are retested, specificity with the pooling algorithm is 100%, i.e., no different than with testing processed specimens individually.

Cost analysis. A model was developed to determine the pool size that yielded the highest cost savings. The binomial distribution was used to estimate the number of pools that are likely to be positive given a selected pool size and population disease prevalence. Next, the optimal pooling number for a range of disease prevalences was calculated. For a dichotomous outcome (i.e., positive or negative test result for a genital C. trachomatis infection), independence was assumed (i.e., the order of the samples received was random with regard to the distribution of the positive or negative samples in the population). The expected percentage of positive pooled assays was determined using the following equation: s = [(l  r/n)^c] × 100%, where s is the expected number of positive pools, r is the number of positive samples tested, n is the total number of samples tested, r/n is the prevalence of disease, and c is the number of specimens pooled. This equation accounted for the probability that from 1 to c samples in the pool were positive.

Estimation of population prevalence with pooled data. Pooling can also be used to reduce the cost of estimating population prevalence. Based on calculations from a previous study, the estimated population prevalence and 95% confidence interval (CI) were back calculated from the pooled data (6). Separate estimates were made for samples pooled by 4 and by 10. Calculations were based on the following equations: (i) Estimated prevalence: p = 1  [1  (s/n)]^1/c (ii) (SD): SD = {[(s/n) × (1  s/n)^(2/c)  1]/(n × c²)}^0.5 (iii) 95% CI: p ± 1.96 (SD) where s is the total number of presumptive-positive pools, n is the total number of pools, and c is the number of specimens in each pool.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Sensitivity and specificity of the pooled assays. A comparison of the distribution of S/COs for the individual and pooled samples indicated that lowering the S/CO from 1.0 to 0.20 for determining positive pools resulted in high sensitivity with a low proportion of specimens from negative pools that need to be retested individually (Fig. 1). There were two weakly positive individual specimens (i.e., an S/CO of 1 but <2.0) in the pilot study (pooled by 4), six weak positives in the study pooled by 4, and four low positives in the study pooled by 10. These weak positives were the only positive specimens in the pool. These pools all tested between 0.2 and 1.0 and sometimes higher.

View larger version (24K): [in this window] [in a new window]   FIG. 1.   Detection of genital C. trachomatis infection by LCR. Graphs A and C show the distribution of individual urine samples from two groups of 568 and 520 women taken from the study population. Graph B shows the distribution of the samples in A pooled by 4 (n = 142), and graph D shows the distribution of the samples in C pooled by 10 (n = 52). The S/CO for individual samples, which was 1.0, was lowered to 0.2 for pooled samples, as indicated.

Cost analysis. For a population with 8% genital C. trachomatis prevalence, which is close to the 8.5% prevalence found in our study population of female U.S. Army recruits, pooling by four provided the highest cost savings. The reduction of total assay costs per specimen, which included technician time, decreased from $12.76 to $7.78, i.e., by 39%. The model demonstrated that with a 2% prevalence, pooling eight samples would reduce the cost per sample by 59%. A population prevalence graph was constructed from the model to determine the number of pooled samples that would achieve the highest cost savings (Fig. 2).

View larger version (19K): [in this window] [in a new window]   FIG. 2.   Cost-saving ability of pooling processed urine specimens before the performance of the LCR test for the detection of genital C. trachomatis infections. The graph shows the cost per sample when the pooling algorithm was used, depending on the number of specimens pooled and taking into account various prevalences of infection in the population screened. A baseline total cost of $12.76 per individual sample which included laboratory consumables, technician time, and LCR unit dose costs was used.

Estimation of population prevalence with pooled data. The observed prevalence for the individual samples in the 142 pools of 4 was 8.5% (48 of 568), and that for the 52 pools of 10 was 11.9% (62 of 520) (Table 1). The estimated population prevalence, back calculated from the number of positive pools, for the 142 pools of 4 was 9.1 (95% CI: 6.5 and 11.6), and for the 52 pools of 10 it was 12.9 (95% CI: 8.8 and 17.0). Each 95% CI included the observed prevalence of the subsample, 10.1% (110 of 1,088). Additionally, each 95% CI included values within 8 to 9%, the overall prevalence measured in a much larger sample (>10,000) of this population.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

In this study we evaluated pooling of processed urine specimens for LCR detection of C. trachomatis for both accuracy and cost-saving ability. The high sensitivity and specificity of LCR was not affected by pooling up to 10 samples when the S/CO was adjusted from 1.0 to 0.2. Although a small percentage of negative pools tested presumptive positive, no specificity was lost with the pooling algorithm, since all specimens in pools which test presumptive positive are retested individually with the manufacturer's specified S/CO for the individual test. Since retesting negative pools does increase costs, the specificity of pools must be high.

The cost analysis model showed that depending on the prevalence of C. trachomatis, the number of specimens that should be pooled for optimal cost savings varies. As prevalence decreases, the pooling protocol for screening could save more than 59% of the cost per specimen compared to that for testing individual samples only. Also, early studies have shown that C. trachomatis screening and treatment programs are cost effective; the Centers for Disease Control and Prevention has estimated that for every dollar spent on prevention, $12 is saved in treating sequelae (4). The use of the pooling algorithm for testing samples obtained during screening could further increase savings in health care costs.

Since C. trachomatis prevalence levels have ranged from 4 to 20% in various populations in the United States, pooling three to four samples is likely to provide the highest cost savings. Furthermore, the cost saved does not significantly change the sensitivity or specificity of the assay. In the event that screening is not conducted, pooling can be used to determine population prevalences over time in order to measure the benefits of disease interventions such as mass treatment or behavioral interventions. The population prevalence back calculation, described previously (6), gave an accurate estimate of the observed population prevalence in this study.

Use of the pooling algorithm would benefit investigators and program planners in two ways: (i) money saved from the use of the pooling algorithm could be applied to other areas of disease prevention and/or (ii) the amount of money allocated to screening would allow more specimens to be tested for the same total cost. Pooling samples for the detection of genital C. trachomatis infection in urine samples is cost saving and simple to perform and could be applicable in screening programs in the United States and in population-based research worldwide.

Pooling is a technique which could be immediately used for significant cost savings in high-volume laboratories such as state labs and referral labs. Laboratories which are currently using less sensitive and specific and less costly techniques could introduce both LCR and pooling into their laboratories.

Specific populations or laboratories that might benefit from pooling include any lab in which the combination of turnaround time and volume allows at a minimum a combination of 19 pools and retests per day. With 96 specimens at a population prevalence of about 4%, pooling by six would fill up one full run (38 test unit doses) per day. The run would include, on average, 16 pools of six and 22 retests.

Laboratory managers should consider two points before using pooling. First, processed specimens from presumptive-positive pools need to be amplified and detected individually. This additional step adds a minimum of 3 hours until individual test results for specimens in presumptive-positive pools are known. Second, laboratory managers should estimate the cost savings they expect to gain for their laboratories. This estimate is a combination of both technicians' salaries and their benefits, institutional overhead, and the prevalence of chlamydia in the populations served by the laboratory. Pooling a greater number than is recommended for certain population prevalences can cost more money than testing specimens individually.

A potential limitation of the pooling algorithm is the possibility of technician error while processed samples are pooled in the LCR run. The use of tray maps simplifies this process. Samples should be organized by skipping a space after each pooled group in the specimen rack. Thus, pooling adds no significant complexity to setting up unit doses. Additional technician error can be avoided when samples from presumptive-positive pools (detected in the previous run) are retested individually before the routine testing of the new pooled groups. Therefore, each run has a combination of samples that are retested individually and new pooled samples from the next batch of specimens.

The study laboratory has met Clinical Laboratory Improvement Act requirements for the modification of a clinical laboratory procedure from a Food and Drug Administration-approved diagnostic kit. Investigators consider performance documentation of the required study adequate for including the pooling protocol in testing clinical specimens in the study laboratory. Each laboratory that wishes to introduce pooling must meet the requirements to modify a Food and Drug Administration-approved package insert. These requirements include meeting the regulations as set forth in the Federal Register (3a).

Use of pooling processed urine samples for LCR testing of C. trachomatis will decrease the cost of screening, providing more evidence that screening programs can and should be implemented. Further applications of pooling include pooling urine specimens for the LCR detection of Neisseria gonorrhoeae. The cost savings of pooling urine for both N. gonorrhoeae and C. trachomatis should also be considered.