Of all the complications of disseminated coccidioidomycosis, the most lethal is meningitis; it is estimated that there are 200–500 new cases of this complication per year [1
]. Treatment with oral antifungals requires lifetime administration to suppress recurrences [2
], and in many cases intrathecal (and neurotoxic) therapy is required to stop progression [3
]. Because of the many complications, which include hydrocephalus, vasculitis, cerebral or spinal cord infarction, arachnoiditis, cranial nerve palsy, syringomyelia, transverse myelitis, cord compression, paralyses, parenchymal abscesses, and seizures, we deem it essential to begin treatment as early as possible, before the pathologic processes have advanced or become irreversible. Early diagnosis is thus very desirable.
Culture of cerebrospinal fluid (CSF) for Coccidioides
, even in active and untreated cases, is positive in a minority of specimens, presumably relating to the focalization of disease to the meninges themselves and the absence of fungal multiplication in CSF. The classical method for diagnosis is the detection of anti-coccidioidal antibodies in the CSF [4
]. The antibodies that react in coccidioidal assays are IgG and IgM antibodies; IgM is usually detected early in the course and IgG persists during disease activity and beyond. IgG antibodies appear directed against the chitinase enzyme of this fungus (this antigen is often referred to as IDCF, detected in complement fixation (CF) or immunodiffusion (ID)), and IgM antibodies to a polysaccharide-containing fungal antigen, incorporating the coccidioidal beta-glucosidase (this antigen is often referred to as IDTP, detected in tube precipitation assays or ID). IDCF [5
] and IDTP [6
] have been defined at a molecular level.
In the present study, we compared lateral flow assay—a new convenient and rapid method used for CSF antibody detection—to older methods and other assays that have been applied to detection of fungal products in the CSF. All comparisons, including “classical tests” (CF and ID), were made with true disease status (defined in Section 2.1
3.1. Assays in Diagnosing Coccidioidal Meningitis, LFAs
At a 1:21 dilution the one-strip LFA had a sensitivity of 95% and a specificity of 100%, a positive predictive value (PPV) of 100% and a negative predictive value (NPV) of 86% (Table 1
). Six (of 60) indeterminate results, as defined, were excluded; 5 of these were due to some observers recording the test as negative and others as a weak positive. The 1:441 dilution was less sensitive (69%) but as specific (100%).
The 2-strip test, IgG assay had a sensitivity of 92%, specificity 100%, PPV 100%, NPV 91% (Table 2
), with 1:441 again less sensitive (70%). The 2-strip IgM test was 53% sensitive, 100% specific, and the 1:441 assay again less sensitive, 20%. The lesser results with IgM are consistent with the usual dominance of IgG antibodies found in CSF.
When considering testing with both strips at 1:21, the combined sensitivity resulting from both tests was 92%, specificity 100% (PPV 100%, NPV 91%; Table 3
), and at 1:441, less sensitive, 70%. Ten (of 77) 1:21 results were declared indeterminate, as defined here; 7 of these were due to the “negative vs. weak positive” discrepancies- thus again, discrepancies were very rarely between clear positive and negative readings.
In summary then, with the 1-strip, or testing IgG and IgM in the 2-strip test, the results were very similar (slightly improved NPV in the latter), and highly efficacious in use for diagnosing coccidioidal meningitis.
3.2. Classical Methods
The classical test, assaying CF testing alone, was also efficacious, less sensitive (71%), but as specific (100%). The PPV was 100% and the NPV 69% (Table 4
a). When ID testing could also be performed, and ID results available, the combined CF+ID results then increased the sensitivity to 78% (specificity 100%), with PPV 100% and NPV 74% (Table 4
b). More information about CF and ID results is given below, in the section “Correlation of tests”.
With the application of the IMMY EIA to specimens, studying IgG only, the sensitivity was 90%, specificity 93% (PPV 94%, NPV 90%). With the IgM assay only, the sensitivity dropped to 49%, though specificity remained at 93%. When the results were combined (i.e., considering a test of either IgG or IgM positive on any specimen as positive), the sensitivity was 94%, and specificity 90% (PPV 92%, NPV 92%) (Table 5
In none of these assays was there an indeterminate result. Thus in this IMMY EIA assay, IgM testing added little to the efficacy of IgG testing alone of CSF. In considering IgG or IgM alone, or the use of both tests, the 1:441 dilution test was less sensitive, 15–75% among those 3 comparisons, but specificity remained high (98–100%).
With the Meridian EIA, tested 1:21, studying IgG only, the sensitivity (71%) was less than the IMMY EIA, though specificity was 100%. With IgM assay, the sensitivity was much less than with Meridian IgG, or IgM in IMMY EIA, only 7% (specificity, 100%). When comparing the utility of both the IgG and IgM results, as discussed above re the IMMY test, the sensitivity (71%) was inferior to the IMMY EIA, though specificity was 100% (PPV 100%, NPV 68%). One of 68 of all these Meridian assays was indeterminate (as defined by the kit).
3.4. Summary: Test Correlations
For summary and ease of comparison, the results of all tests, used in their optimal fashion, is repeated in Table 6
3.5. Percent Agreement between Tests
The percent agreement between all tests, on positives or negatives, was very good, with the least agreement between BDG antigen results and the antibody assays. These comparisons will be affected by the differences in sensitivity between tests, already described.
For example, considering all tests used in their optimal (CF + ID; for other tests, combining the results of IgG and IgM testing), the agreements among positive assays for each of the 5 antibody tests with each other ranged from 87–100%, with all but one being ≥90% (Table 7
). Agreements on positives of the 5 antibody tests with the BDG assay ranged from 68–90%.
3.6. Correlations of Tests
The correlations between tests are seen in the comparison of the assay results (first Results section), and also the above section on agreements between tests. Two tests, the CF and the BDG test, give quantitative results, beyond only positive or negative. In a previous study [7
], there was no correlation found between the BDG CSF antigen titer test and the CSF CF antibody titer. In examination of the present results, there was also no impressive correlation with CSF BDG antigen titer and 1-strip LFA, 2-strip LFA, IMMY EIA, or Meridian EIA antibody tests outcome either; the r2
correlation coefficients ranged from 0.12 to 0.21.
On the other hand, with both the IMMY EIA and the Meridian EIA, the only negative CSF EIA with each test that had CF titration, had CF titers of only 1:2. The r2, with CSF CF for the IMMY IgG EIA was 0.38, and for the Meridian EIA 0.25. Overall, the CSF CF positive titers in the present study were: undiluted only, 4; 1:2, 8; 1:4, 2; 1:8, 6; 1:16, 2.
Two patients had serum drawn the same day as a CSF sample was available. Both samples from both were positive in IMMY EIA, but there was insufficient testing by other assays to draw conclusions about concordance between the 2 samples.
Of the CSFs from cases in the present study, we obtained the ID results for 2 that were CF negative and 2 that were CF positive (titers 1:2), and in these 4 the ID results were positive. In another case, the quantity was insufficient for a CF test undiluted, a 1:2 dilution was CF negative, and the ID was positive.
The CSF CF negative cases were of particular interest, as these might reflect CSFs with the very lowest antibody concentrations. The 2 CF negative, ID positive specimens above were positive in both LFAs, IMMY EIA, and BDG, and one of two Meridian EIA positive.
In addition to those, there were 7 others CF negative. All but one appeared to be on therapy, and thus their disease possibly stabilized or in remission. One of these was a ventricular CSF, this entity to be discussed later. Of the other 6, 2 were ID negative as well and 4 had had no ID done (apparently owing to inadequate sample). Of these 6 CF negatives, 5 were positive in the 1-strip LFA and the IMMY EIA, 3 of 4 were positive in the 2-strip LFA (2 indeterminate), 5 of 5 tested were BDG positive, and only 1 of 6 positive in the Meridian EIA. This suggests a weakness in Meridian EIA when antibody titers are low. In one of the 6 (this one positive in the 2 LFAs and IMMY EIA, negative in Meridian EIA; not BDG tested), we were able to find a record that a second CSF (unavailable to us), tested 1 mo. later, had turned CF positive, indicating those LFA and IMMY EIA positive assays on our earlier specimen had detected a case in the early stage.
Of cases that were studied with PCR, and negative, 2 of 3 could be studied presently. One was a ventricular specimen, these discussed below. The other was negative in CF, IMMY and Meridian EIA testing, indeterminate in the 2-strip LFA, and only positive in the 1-strip and BDG assays.
3.8. Ventricular Specimens
Ventricular CSF specimens represent an especially difficult instance in CSF testing, because, unless there is ventriculitis, the ventricles have disproportionately less inflammation than concurrent lumbar specimens, and because of the dilution in ventricular volume, detecting antibody and antigen would also be more difficult. Three specimens were included in the present study, but not all the specimens could be tested by all assays. Two were CF negative, but one of these was ID positive (also PCR negative). Two of 3 were Meridian EIA positive, 2 of 2 were BDG positive, and all 3 were positive by 1-strip LFA, 2-strip LFA and IMMY EIA.
3.9. Serial Specimens
Multiple lumbar CSF specimens were available from 7 patients. Of patients with samples obtained 1 year or less apart, their assays (CF, 1- and 2-strip LFAs, IMMY and Meridian EIAs, and BDG) were generally very consistent over time. However, samples were obtained from 2 patients separated by 7 years on treatment, and in both patients their results generally trended to turn from positive to negative in all assays. In another patient we were able to determine that his CSF inflammation (as determined by CSF leukocyte count) was declining over 4 serial samples, and all his assays in the present study also turned negative. The final patient worthy of individual mention is a patient with AIDS, as it is noteworthy that all the CSF antibody assays were positive, indicating this co-morbidity did not interfere with CSF antibody production. Over a 4-year period, the antibody tests remained positive on treatment.
3.10. Rabbit CSF
Twelve rabbit study CSFs were available for testing, 4 from animals prior to infection, and 8 (half untreated controls and half treated with intravenous amphotericin B) on day 14 or 15 post-infection. With the 1-strip LFA, all 8 infected rabbit CSFs gave a strong positive test for coccidioidal antibody, with a positive control test line as well, the latter indicating (as it would in human CSF testing) a valid assay. The CSFs from the 4 uninfected rabbits also yielded a positive control test line, but no positive test line.
The study here of 49 patient specimens, and 40 control specimens, is, to our knowledge, the largest comparative study of coccidioidal CSF diagnosis to date. Most assays studied had excellent sensitivity for the diagnosis of coccidioidal meningitis, and were efficacious in separating cases from controls (Table 6
). Assays that do not require sending specimens to special reference laboratories offer logistical advantages. With the availability of kits, local hospital laboratories in endemic areas can be the source of testing. LFA assays do not even require any laboratory, are simple to use, and give rapid results; potentially even at the bedside.
Previous studies with the classical CF antibody test on CSF have reported false negative results that range from 17–41% [14
]. A recent study [17
] found CSF CF antibody testing to be 70% sensitive and 100% specific, virtually identical to what we found (Table 4
a) with our specimens and assays. This same study reported pooled EIA results on CSF from two manufacturers (not stated), and found a sensitivity for IgM of 8%, and for IgG, 85%.
A recent addition to the CSF diagnostic armamentarium is the detection of coccidioidal antigen [17
]. In CSF, a sensitivity of 93%, specificity of 100%, PPV of 100%, and NPV of 97% were the reported indices. These results are virtually the same as what we report with the 1-strip and 2-strip LFAs, and IMMY EIA assays (Table 6
), although the NPV is higher. That antigen test is more rapidly performed than the classical antibody tests, but requires sending to one reference laboratory, with the ensuing turnaround time delay. The sensitivity is virtually identical to what was reported re CSF with the BDG antigen assay [7
], although the other indices were higher for the coccidioidal antigen test. The BDG assay is not specific for coccidioidal antigen, but as the only other common mycosis involving the CNS is cryptococcosis (for which highly useful and widely available diagnostic methods, proven over decades [18
], and newer convenient diagnostic tools [19
] exist, and is a condition where BDG assay is less sensitive anyway), the lack of BDG specificity is less of a problem than one might expect. A possible convenience for BDG testing is that many hospitals have BDG kits in-house, because they are used for serum diagnosis of opportunistic mycoses in compromised hosts. There was no correlation found between coccidioidal CSF antigen tests with CSF coccidioidal antibody titers [17
], as was also found in the present study between BDG antigen results and the results of the antibody tests (Table 7
), and previously for CSF BDG titers and CSF CF titers [7
]. Our interpretation of such findings would be that antigen-antibody binding in CSF can remove antigen and/or antibody from the CSF, and then the undetectable analytes ablate the possibility of the concordance such as seen among various antibody test results (Table 7
The few, infrequently encountered and potentially diagnostically difficult ventricular specimens did not present problems for several of the assays in the present study.
The previously reported PCR assay study [11
] reported good sensitivity for detection of coccidioidal DNA in tissue, but (although few CSFs were tested), inefficacy in PCR testing CSF. The latter was underscored in the present study, where present assays could detect coccidioidal antibody in previously PCR-negative specimens.
The ability to provide titers, beyond qualitative results, could potentially enhance the utility of several assays studied here, as higher titers might then be studied for correlation with true disease status, and those might provide further assurance in distinguishing cases. In addition, titers offer the possibility of correlating with therapeutic results, and be useful in following the course of patients. Our present results given, comparing 1:21 and 1:441 dilution results, already suggests utility of titration of these assays, and some further pilot studies with IMMY reagents reinforces this conclusion.
With respect to studies of interest using sera
, Meridian EIA has been compared to classical assay methods, e.g., 92% sensitive and 81% specific [20
]. Investigators have previously compared Meridian EIA and IMMY EIA for coccidioidal sera [21
]. The Meridian EIA had more false positive tests for IgM, and was less reproducible across participating laboratories. The 1-strip LFA was, for serum, compared against CF and ID, and 98% agreement on positives reported [22
]. It would be expected that BDG testing in CSF in meningitis would be more sensitive than BDG testing in coccidioidal sera [23
], since the relatively closed CNS compartment does not have the elimination pathways available in the systemic circulation.
There are several shortcomings to the present study. We do not know the onset of disease for the case specimens we tested, thus which specimens (especially ones test negative) were early in the course, which were obtained later (when disease had more time to develop), and which (especially ones test negative) followed remission of disease.
It would also have been of interest to have more serial samples from patients during their course, so as to have more data to attempt to correlate disease activity and test result.
We lacked ID test data on some CF test negative patients, and some more might have been declared “classical test” positive if they were ID test positive. We did not have information on the level of CSF inflammation (CSF leukocytes, protein, glucose) in almost all specimens, to correlate with our test results. To date, testing sera in coccidioidal antibody and antigen assays [16
] has suggested only histoplasmosis presents a significant cross-reaction problem, but meningitis in histoplasmosis is a rare event; more data about performance of the tests studied here in other CNS mycotic infections would still be of interest. Data-gathering addressed to these shortcomings can be derived in future studies, and inform them.
Our conservative method for LFA interpretation. utilizing multiple readers, give the reader an idea about how much variation there could be, among different readers, of a weakly positive test, although it likely increased the number of indeterminate results in this study. Inexpensive densitometry scanners exist for LFAs, and a quantitative cutoff value could be accurately determined with those instruments. Introducing these instruments as part of the LFA testing could have the downside of complicating bedside testing.
With respect to the rabbit studies, the antibody binding protein on the gold particles in LFA isn’t species-specific, so specific antibody from any species might react as would human antibody, explaining the positive test line. For the control LFA line to be positive requires enough cross-reactivity between human and rabbit antibody for there to be a reaction based on binding by the goat anti-human serum. Tests on the rabbit samples were 100% sensitive and specific, suggesting sufficient cross-reactivity with human antibody in LFA so as to be potentially useful in diagnosis of some other species. The animals for which this information could potentially be most useful would be dogs, as dogs are commonly infected in endemic regions [24
], and can develop CNS disease. As this manuscript was being prepared, a paper appeared [25
] on the utility of LFA testing on dog sera in coccidioidal diagnosis.
In conclusion, a number of reliable tools are available to the clinician in the diagnosis of this devastating complication. It is extremely important for clinicians to consider coccidioidal meningitis in the differential diagnosis of CNS infection, in the endemic areas and in patients with a relevant travel history, so that diagnostic tests are not delayed, and treatment instituted before devastating events ensue. The tests available have become increasingly convenient, easy to use and thus to repeat if necessary, including in sequential testing, and give rapid results.