Screening for West Nile Virus in Organ Transplantation: A Medical Decision Analysis

Authors


*Corresponding author: Bryce A. Kiberd, bkiberd@dal.ca

Abstract

The Organ Procurement and Transplant Network (OPTN) has recently announced that screening for West Nile Virus (WNV) in deceased organ donors is not recommended at this time. The purpose of this report was to examine the impact of this recommendation by using medical decision analysis.

Without screening the rate of disease transmission was assumed to be the same as in donated blood with a case fatality rate of 25%. With screening we assumed the baseline screening test specificity and sensitivity to be 99.5% and 95%, respectively. The analysis was confined to heart, liver and kidney recipients. Survival probabilities and transplant rates were taken from UNOS.

Annual screening could result in the loss of potentially 452.4 life years (113.8 for heart, 272.6 for liver and 66.0 for kidney). Most positive test results would be false-positive. Screening would be preferable for kidney donors in areas of high disease prevalence and high test specificity. However, for heart and liver most scenarios were associated with a net loss of life with screening, except if patients were stable on the wait list with particularly high case fatality rates from WNV.

Current recommendations by OPTN that screening is not mandatory seem appropriate until further data are available.

Introduction

West Nile Virus (WNV) is a mosquito-borne, single-stranded RNA flavivirus, which was first identified in 1937 but introduced to North America in 1999 (1). Since then human infection has been reported in most states and in many provinces of Canada. The transmission season is long, with cases occurring into December in some parts of the United States. The virus has a longer history in Europe, Africa, Asia and the Middle East. Although the mosquito is the vector of disease transmission, human disease has also been acquired through blood transfusion, breast-feeding, transplacental transmission, occupational exposure in laboratory workers and stem cell and solid organ transplantation (2).

Most infections are mild, however, people older than 50 years with chronic medical conditions and those receiving immunosuppression are at highest risk of severe disease, which may include encephalomyelitis and death. Approximately 200 Americans were believed to have died from WNV in 2002 (3). Case fatality rates are low among the general population but ranges between 4 and 29% for hospitalized patients. In the solid organ transplant recipient the case fatality rate is quite high. The limited number of documented cases and the lack of routine post transplant screening in all patients preclude an accurate estimation of this rate. Currently, there is no specific drug treatment or vaccine against the infection.

With respect to the transplant recipient, the most likely mode of transmission is also the mosquito. Preventive strategies, such as avoiding outside activities at dawn and dusk, using N,N-diethyl-m-toluamide (DEET)-containing repellents and eliminating breeding sites, will help reduce mosquito bite transmission (1,3). Routine blood-donor screening should reduce this mode of transmission. However there is as yet no endorsed strategy to test deceased organ donors in North America. The difficulty is detecting the infected potential donor. By the time serologic tests are positive the diseased individuals may no longer be infective, and the tests for detecting virus are time consuming, cross react with other viruses and may not detect low-level viremia (1,3).

The Organ Procurement and Transplantation Network (OPTN) and the Health Resources and Services Administration (HRSA) recently released their recommendations on the role of deceased donor screening in January 2004 (4). They recommended that donors with encephalitis, meningitis or flaccid paralysis in geographic areas with human WNV infection be deferred. They acknowledged that some centers were testing deceased donors and added that organs with known reactive tests before transplantation be used only if the recipients were informed of the risk and had an emergent life-threatening illness requiring transplantation and no suitable alternative therapy. Because of testing logistics, the lack of a FDA-approved test and known limitations of testing with false-positive and -negatives, this group did not maintain that screening should be required at this time. Once a test is licensed, the organization recommends the appropriateness of screening be reassessed. Given the seriousness of this infection, delaying screening at this time may have a large negative impact on transplant outcomes. Quantifying the impact of screening would provide useful information to the transplant community.

The purpose of this report was to examine the impact of testing for WNV in deceased organ donors using a medical decision analysis model. This study examines the impact of lives lost or gained by introducing routine screening and examines the impact of test characteristics (sensitivity and specificity), specific solid-organ recipients (heart, liver and kidney), and the prevalence of the WNV-infected donor on these outcomes.

Methods

A medical decision analysis was created comparing two strategies, ‘screen’ and ‘no screen’. In the ‘no screen’ strategy, the model assumed that deceased donors had the same likelihood of being infective for WNV as blood donors. As the prevalence varies by region and season, a range of prevalence rates were examined. Patients receiving an organ from an infected donor had a baseline case fatality rate of 0.25 (0.10–0.75; 5). As few cases have been reported, a wide range of case fatality rates was examined. The outcome of interest was life years.

In the ‘screen’ strategy we presumed a viral nucleic amplification detection test would be employed using either reverse transcriptase polymerase chain reaction (RT-PCR) or nucleic acid sequence-based amplification (NASBA). We did not examine the serologic test, as evidence to date has shown that the test is not reliable in detecting those that are capable of transmitting infection (although an excellent test to diagnose recent infection). In the cases of transmission through blood transfusion, the serologic tests were negative whereas the nucleic acid tests were positive (6). In cases that are serologic IgM positive, <10% are positive for virus (7). The viremic phase may last only 2–15 days, whereas the duration of the IgM positivity phase may last more than 90 days, with only several days of overlap (3). We assumed that the nonreactive donors on peripheral blood samples did not have infected organs. To assume otherwise would reduce the benefit of screening.

We also assumed that the screening test would be available even for the more time-sensitive transplants such as heart and liver. We assumed that organs screening positive (including false-positives) were discarded. We estimated the test characteristics from the published and network literature (Table 1). Organs that were false-negative were assumed to be transplanted and would transmit disease as in the ‘no screen’ strategy. The impact of a discarded organ was captured by assuming that the benefit of this transplant would be lost and that a wait-listed patient would remain on the list.

Table 1.  West Nile virus prevalence and test characteristics
  1. *Specificity and sensitivity quoted refers to that of detecting an asymptomatic deceased donor who is viremic and capable of transmitting disease: not diagnosing an individual who has had an infection and may be noninfective.

  2. RT-PCR = reverse transcriptase polymerase chain reaction, NASBA = nucleic acid sequence-based amplification.

Prevalence0.00024 (0.0001–0.02)(6,9,13,14)
RT-PCR/NASBA*
 Specificity0.995 (0.985–1.0)(9)
 Sensitivity0.95 (0.90–0.99) 

Baseline patient mortality probabilities while on the wait list and after transplantation for patients of various organs were abstracted from the United Network of Organ Sharing (UNOS) and the literature (Table 2). We assumed a 25-year horizon with an annual 5% discount rate of life years. Kidney organ recipients with allografts that failed were returned to permanent dialysis with an assumed higher mortality rate than dialysis patients active on the wait list. To examine the impact on transplantation in the US, we used the numbers transplanted in the year 2002 from UNOS. We assumed that the number of combined organ transplants (i.e. heart kidney or kidney liver) were negligible. On the other hand we did not include lung, small intestine or pancreas alone transplantation. The software used was Data 4.0 (TreeAge Software, Inc., Williamstown, MA). Figure 1 shows the decision tree for a heart transplant. We examined one-, two-, and three-way sensitivity analyses for test sensitivities and specificities, wait list and post transplant mortality, and prevalence of the infected donor.

Table 2.  Organ-specific baseline variables
  1. 1Organs transplanted in 2002 UNOS registry

  2. 2Wait list mortality calculated by proportion transplanted by status and available published wait-list mortality rates by status. Ranges derived by lowest to highest mortality by status. Three-month mortality from the model for end stage liver disease (MELD) scores were annualized.

  3. 3Mean and 95% confidence intervals from Kaplan-Meier survival rates 1996–2001, analyzed November 28, 2003: http://www.optn.org/latestData/rptStrat.asp

  4. 4Range of wait list and permanent dialysis mortality based on published mortality by age of 0–19 and >60-year-old groups.

  5. 5From years 1 and 3, survival percentages of annual logarithmic decline were calculated and used to project subsequent survival.

Heart transplantation
 Number12155(15)
 Annual mortality on wait list20.293 (0.16–0.43)(16)
Patient survival
 1 year30.853 (0.844–0.862)(17)
 3 years3, 50.780 (0.771–0.789)(17)
Liver transplantation
 Number14969(15)
 Annual mortality on wait list20.30 (0.06–0.80)(18,19)
MELD score mean 20.7 (range 10–30)
Patient survival
 1 year30.871 (0.865–0.876)(17)
 3 years3, 50.797 (0.790–0.803)(17)
Kidney transplantation
 Number18539(15)
 Annual mortality on wait list2, 40.063 (0.02–0.10)(20)
Patient survival
 1 year30.955 (0.953–0.957)(17)
 3 years3, 50.909 (0.906–0.911)(17)
Graft survival
 1 year30.909 (0.906–0.912)(17)
 3 years3, 50.815 (0.811–0.828)(17)
Annual permanent dialysis mortality40.161 (0.036–0.232)(20)
Figure 1.

Medical decision tree for heart transplantation. Sens = sensitivity, spec = specificity, CHF = congestive heart failure health state, CHFamr = CHF annual mortality rate, Htransplant = first year after heart transplantation health state, Htransplant2 = after the first year with a heart transplantation health state, Txmr1 = mortality within the first year after heart transplantation, Txmr = mortality after the first year after heart transplantation, CFR = case fatality rate of West Nile Virus (WNV) disease.

Results

The results show that testing for WNV in 2002 could have resulted in the loss of approximately 66.4 (113.8 undiscounted) life years for heart transplant recipients, 159 (272.6 undiscounted) life years for liver transplant recipients and 38.8 (66.0 undiscounted) life years for kidney recipients. As shown in Table 3 most of the positive tests would be false-positives. Figure 2 shows the range of life years gained/lost for the three organs over a range of test specificities.

Table 3.  Baseline outcomes of the West Nile virus reverse transcriptase polymerase chain reaction test
 Net loss life years*
discounted/
undiscounted**

Total test positive
number/year

True positive
number/year
  1. *Assumed 2155 heart, 4969 liver and 8539 kidney transplants in 2002.

  2. **Future life years discounted at 5%.

Heart66.5 (113.8)11.30.5
Liver 159 (272.6)29.01.1
Kidney 38.8 (66.0)44.71.9
Figure 2.

Net life gained/lost with West Nile Virus (WNV) screening strategy over range of test specificities.

Under one- and two-way sensitivity analyses there was a consistent loss of life years with the ‘screen’ strategy compared with the no screen strategy (data not shown). For screening to result in a net gain in life years several conditions were required (Figure 3). With a case fatality rate of 75% (in addition to the baseline life lost within the first year), a lower wait list mortality rate (16% annual mortality rate) and a high disease prevalence (one of every 200 individual with active infection), the test specificity must exceed 99.7% to have a net benefit. The interface between ‘screen’ vs. ‘no screen’ is the point where screening results in zero net life years gain (loss). From this figure low prevalence rates demand very high test specificities. However for kidney transplantation (Figure 4), using the baseline case fatality rate, the test specificities can be lower and still result in net benefit. For example WNV donor screening with a test specificity of 99.4% would have resulted in better outcomes compared with the no screen strategy if the prevalence had exceeded 0.005 (1/200). Test sensitivities had a lesser impact on differences in strategy outcomes (data not shown).

Figure 3.

Sensitivity analysis for heart transplantation assuming 75% case fatality rate and mortality on the wait list of 16% annually over a range of test specificities and West Nile Virus (WNV)-infected donor prevalence rates.

Figure 4.

Sensitivity analysis for kidney transplantation over a range of test specificities and West Nile Virus (WNV)-infected donor prevalence rates.

Discussion

Although screening to eliminate WNV-infected donors seems intuitive, it is possible that screening from a societal prospective may do more harm than good. The results suggest that transplanting heart or liver wait-listed patients with an untested organ provides a greater societal benefit than screening and discarding reactive organ donors. Our findings lend support to the recommendation by OPTN not to consider mandatory screening at this time (4). However the organization did state that donors with encephalitis, meningitis and flaccid paralysis of undetermined etiology in geographic regions with known WNV human infection be deferred. Our medical decision analysis did not analyze this component of the recommendation.

Their recommendations went further and considered offering organs testing positive to informed recipients with life-threatening illness requiring urgent transplantation (4). This analysis highlights that many of these reactive organs may be false-positive, the case fatality rate for WNV is not 100% and that the mortality waiting for selected patients is quite high. Given the uncertainty, full informed consent would be essential if this action is to be taken.

Creating a safe organ procurement system is a high public priority, no less than for blood services. There are essentially no wait lists for most blood products and discarding blood products at high risk for WNV infection is an acceptable trade-off. However, even a small organ discard rate has important consequences for patients on the wait list. Despite this, our results also show that not all recipients should willingly accept an untested organ. Patients with relatively good survival on the list and patients receiving organs from endemic areas would be best served to have their donor tested with a very specific test. Under these circumstances once a reliable specific test is licensed, this analysis would support testing.

The perspective taken in our analysis is mostly societal, as ‘not screening’ does place a rare patient at risk. The societal perspective maximizes good at the predictable expense of an unlucky individual. More patients may benefit from a ‘no screen’ strategy. Nonetheless, if the screened positive organ for a particular patient was discarded and if a guarantee given that the next available suitable organ would be earmarked to this particular patient and if this patient could be assured that he/she would survive this wait, then screening should be instituted. In reality there are no such guarantees for the sick heart and liver wait-listed patient. Nonetheless decisions should be made on an individual basis and full informed consent of any decision including discussing the risks of not screening is recommended.

The situation is different for the kidney failure patient, especially in a wait list-dominated allocation strategy. Forgoing a kidney today for a patient at the top of the wait list is not likely to be of great harm. From this patient's perspective screening should be performed. Our results show that this would also be the case, from a societal perspective, for potential kidney recipients, if the organ is from an endemic area and there is a good test available.

The results also demonstrate that patients on heart and liver transplant lists who are more stable may also benefit from waiting for specific tests. However, this assumed a very high case fatality rate. As with all risks, patients should be appraised of risk preferably at the time of listing rather than at the time of organ availability.

Cadaver donor screening for other infectious diseases is standard practice (8). The other viruses screened in donors can all cause serious morbidity and mortality. Ideally these screening strategies should have high negative and positive predictive values. High negative prediction reduces the risk of disease transmission and high positive prediction reduces organ wastage. Transmission rates of disease from organs with negative viral screens and organ discard rates because of positive viral screens are not widely available.

Differences in positive prediction largely reflect the prevalence in the population and test specificity. As WNV exposure will vary with region and season, knowledge of regional epidemiology is critical if screening is implemented. With better technology the specificity of the test may improve. It is possible that the specificity of nucleic acid amplification testing (even when performed on an emergent basis and often after hours) will be higher than those used in the development of this model. At present no approved test exists for this application. Testing performed by proficient laboratories operated by blood donation programs such as Canadian Blood Services and the American Red Cross is likely to be better than testing performed on cadaver organs by in-house laboratories gearing up on a stat, 24-h, 7-day per week basis. For the deceased organ donor there will be no time for confirmatory testing.

Limitations to the study should be discussed. Accurate assessments of case fatality rates and disease prevalence at this time are impossible. Using blood transmission rates assumes that deceased organ donors are at the same risk and this may not be true. As patients presenting for blood donation are likely to be self selected and feeling well, the WNV infective rate may well be higher in deceased donors. The clinical specificity and sensitivity of the test is essentially unknown. We used a baseline screening test sensitivity and specificity of 95% and 99.5%, respectively, as this was a goal judged to be an acceptable goal during an FDA-sponsored symposium on WNV (9). We examined in a one-way sensitivity analysis higher values for the above inputs to make the analysis more robust and in favor of screening. Despite this, screening from a societal perspective for heart and liver transplantation is marginal unless WNV prevalence, test specificity rates and case fatality rates are all much higher.

We also did not include the potential of short or long-term disability (reduced quality of life) in an infected recipient nor perform a cost-effective analysis. Unfortunately there were no studies measuring short- or long-term quality of life preferences in patients disabled from the WNV infection. However for the small group of individuals who develop WNV disease, morbidity can be prolonged. Including disability in the rare recipient will only make a trivial change compared with the lost benefits from discarded false-reactive organs. As significant changes in the case fatality rate up to 75% had little effect on results, even large reductions in permanent quality of life of the WNV-infected survivors would not be expected to change the conclusions. Similarly adding increased hospital or long-term costs in these patients will be small compared with testing all donors and the missed savings of forgone kidney transplantation. A full cost-effective analysis should be performed if screening is to be recommended.

Issues with WNV in the transplant and the general population are evolving. Further improvement in testing is critical and should be supported. Although most centers do not have the capacity to perform sensitive and specific tests within a short time period, suitable tests may be performed with current instruments within 1–3 h. Given the logistics and costs a clear recommendation will be required before widespread testing is implemented. As more patients are infected, knowing who has been exposed may prove to be a partial solution. If prior infection results in life-long immunity, a pool of recipients may be suitable to accept these organs. As WNV-naive potential recipients will have a lifelong risk of WNV, it makes sense that this population is ideal as candidates for a vaccine, ideally pretransplant (10). If testing is delayed or impossible before engraftment, the use of high WNV titer immune globulin may prove beneficial if the test is later found to be positive (4,11). Other options for treatment may be available in the future (12).

In summary, this paper has shown how use of a medical decision analysis helps decide whether or not to implement deceased donor WNV screening by integrating differences in the type of organ transplanted, WNV disease prevalence, test characteristics and survival on the wait list. The current position of the OPTN and HRSA not to recommend deceased WNV donor screening seems justified for the time being. However this will likely change in the future. Given the variations in infection prevalence, the recommendations may have to be flexible.

Ancillary