“Straight Talk About the Diagnosis of Lyme Disease”
Phillip J. Baker, Ph.D.
American Lyme Disease Foundation
It is often stated in public discussions that Lyme disease is vastly under diagnosed and that the number of cases reported annually to the CDC represents only about 1/10th of the true number of cases. This should not be surprising since under reporting seems to be an almost unavoidable feature of the reporting process per se.
Although the CDC acknowledges that Lyme disease may be under-reported by a factor of 3-12 times (1,2,3,4) , the degree of under reporting is not unlike that observed for other reportable diseases such as gonorrhea (2 times), Chlamydia infections (4-6 times), E. coli O157:H7 infections (13-27 times), hepatitis B virus infections (20 times) and Salmonella infections (35-38 times). Typically, under-reporting is measured by comparing the number of patients who are diagnosed and reported vs the number who are diagnosed, but not reported. Since in both instances, most –if not all—of the patients are or have been receiving treatment for Lyme disease, it is not true to say that those whose disease is not reported are denied medical treatment. The situation, therefore, may not be as dire as it may appear at first glance.
It is important to understand that cases are not counted for surveillance solely on the results of serological tests; patients also must have other consistent clinical features of the disease to be included in the count. For example, the State of Connecticut receives about 15,000 positive laboratory tests for Lyme disease each year; however, it reports <4,000 cases, due to the lack of additional clinical data required for surveillance reporting. This disparity could be remedied if physicians were required to complete the health department forms to include all of the requisite missing data; that is easier said than done. However, if the data submitted are not accurate and typical of that associated with Lyme disease, over diagnosis may result as indicated by the following observations that emerged when data submitted to the Connecticut State Department of Health were subjected to greater scrutiny. Of 678 patients identified from a random sample of all case reports submitted from 1984-1991, only 64% were found to have met the surveillance case definition for Lyme disease; the remainder (36%) were found to have arthralgia, rather than the arthritis characteristic of Lyme disease (5). In another examination of surveillance reports submitted from 2000 to 2003, a random sample of 709 patients showed that a similar number (66%) were classified as definite cases of Lyme disease; however, 15% were classified as possible cases of Lyme disease, and 19% were classified as unlikely cases of Lyme disease (6). Although one might argue that Lyme disease may be either under diagnosed or over diagnosed, the truth most likely lies somewhere in between.
The development of improved diagnostic tests certainly is a desirable goal, and the C6-ELISA represents a major advance in that regard (7). The specificity of this recently developed test is improved due to the use of a defined ligand (C6) that greatly reduces the degree of variation encountered when the same specimens are tested by different laboratories. Although studies reveal that the results obtained using the C6-ELISA compare favorably with those obtained using the conventional two-tiered testing procedure (8, 9), it is premature to recommend the C6-ELISA be used in place of two-tiered testing until more data are acquired.
The key issue is not so much whether improved tests ought to be developed, but rather: (a) how newly developed tests might best be evaluated and compared to existing procedures, and (b) how to establish superiority so that recommendations then can be made with confidence concerning their routine use. Since comparative studies are time consuming and not always conducted using the same panel of reference specimens under the same set of experimental conditions, the results obtained are often difficult –if not impossible—to interpret. The establishment of a serum reference repository with a computerized data base would greatly accelerate this decision making process. It would enable one to compare the results of newly developed and existing diagnostic tests under identical conditions using the same panel of well-characterized reference specimens. At a relatively modest cost (<$1M), it can be designed to yield the precise type of information needed by the FDA and CDC to make sound recommendations on the best diagnostic tests to be used routinely, as well as to provide pertinent information on a test’s strengths and limitations.
Another important consideration is the use of what most physicians would consider to be an excellent diagnostic test, under circumstances in which the probability of having Lyme disease is low. In 2007, 27,444 cases of Lyme disease were reported to the CDC; this represents an average of 9.1 cases per 100,000 persons, or an incidence of 0.009% for the entire population. If one assumes that Lyme disease is under reported by a factor of 10, the incidence then could be as high as 0.09%. In the ten States where Lyme disease is most prevalent, the average number of reported cases – for the same year—was 34.7 cases per 100,000, an incidence of 0.035%. Once again, if one assumes the true number of cases to be 10 times more than the numbers reported, the incidence could be as high as 0.4% in endemic areas. Data provided by the New York State Department of Health (for 2004-2006) indicate the highest reported incidence of Lyme disease to be in the Hudson Valley Region (1.3%), which includes Duchess (3.8%), Putnam (1.8%), and Westchester (0.5%) counties (http://www.health.state.ny.us/statistics/chac/general/lyme.htm). These counties rank among the highest in the reported incidence of Lyme disease for the U.S.
Seltzer and Shapiro (10) showed that, under circumstances where the theoretical incidence of Lyme disease is 1%, and using what one would consider to be an excellent diagnostic test with 95% sensitivity and 90% specificity, the predictive value of a positive test (the percentage of persons with disease among all those having a positive test) is only 8.7%. They also showed that, of all tests that were positive for Lyme disease under the same conditions, 91.3% were found to be falsely positive. Once again, this applies to a situation where the theoretical incidence of Lyme disease is 1% — an incidence far greater than even the maximum projected for the U.S. as a whole (0.09%) and for major endemic areas (0.4%). If one assumes an incidence of disease of 10% and 50%, the projected positive predictive values would rise to 51.4% and 90.5%, respectively. Based on these observations, Seltzer and Shapiro emphasized “the need for judicious use of diagnostic tests to ensure that the predictive value of a positive test is high.” If a test is used in groups of patients with a high probability of having Lyme disease (e.g., in patients with objective physical signs commonly associated with Lyme disease), its predictive value will be high. However, if it is used in groups of patients with nonspecific symptoms or in situations where there is a low probability of having Lyme disease (e.g., in patients from areas where Lyme disease is relatively rare or in areas where the Ixodes ticks that transmit Lyme disease are not found), its predictive value is low. Porwancher attests to the fact that the long-used — and often much maligned– conventional two-tiered test for Lyme disease, when applied under conditions where the pre-test risk of Lyme disease is high, indeed provides excellent and reliable results (11). By contrast, use of the two-tired test for Lyme disease by physicians for any type of joint pain/arthritis, where the pre-test risk of Lyme disease is low, provides poor and unreliable results.
- Pub. Health Manag. Prac. 2: 61-65, 1996.
- J. Infect. Dis. 173: 1260-1262, 1996.
- Amer. J. Epidemiol. 148: 1018-1026, 1998.
- Amer. J. Epidemiol. 155: 1120-1127, 2002.
- JAMA 283: 609-616, 2000.
- Emerging Infect. Dis. 14: 210-216, 2008.
- J. Clin. Microbiol. 37: 3990-3996, 1999.
- Clin. Diag. Lab. Immunol. 11: 924-929, 2004.
- Clin. Infect. Dis. 47: 188-195, 2008.
- Ped.Infect.Dis. J. 15: 762-763, 1996.
- See complete article on diagnosis by Porwancher posted on www.aldf.com .