Clinical diagnosis of cerebral amyloid angiopathy: Validation of the Boston Criteria

Methods.

We identified all available brain tissue specimens from our prospective cohort of patients aged ≥55 years with primary lobar ICH admitted or referred to the Massachusetts General Hospital (MGH) or Spaulding Rehabilitation Hospital (SRH) between July 1994 and March 2000. To avoid a systematic bias, we excluded patients who had a pathologic diagnosis of CAA at the time of referral to MGH or SRH. Of 233 eligible patients, 39 (17%) were determined to have available pathologic samples containing cerebral vessels.

Each of these primary lobar ICH patients was diagnosed on clinical and radiographic grounds with probable or possible CAA according to the Boston criteria (see the Appendix). The diagnostic radiographic studies consisted of gradient-echo MRI scans in 15 patients, MRI without gradient echo in seven, and CT scans alone in 17. All images were obtained prior to or within 1 week of the acquisition of the neuropathologic sample, and all were evaluated for this study by a stroke neurologist (J.R.) without knowledge of the pathologic diagnosis. Four patients taking warfarin with an international normalized ratio (INR) ≤3.0 were diagnosed with probable CAA, whereas an additional four patients were instead diagnosed with possible CAA because of an INR >3.0.

Neuropathologic specimens were obtained from patients by postmortem brain examination (n = 14), hematoma evacuation (n = 21), or cortical biopsy (n = 4). All samples were stained by conventional methods including Congo red and examined in random order by a neuropathologist (D.K.) blinded to clinical diagnosis. CAA was diagnosed as the cause of ICH in postmortem brains by the complete replacement of blood vessel walls with amyloid, cracking of the vessel wall (creating a “vessel-in-vessel” appearance), and at least one focus of paravascular blood leakage, without other definite source of hemorrhage. CAA was diagnosed in hematoma or biopsy samples by complete replacement of a vessel wall with amyloid, an approach previously shown to correlate with CAA-related ICH with high specificity.4

Results.

Of 39 primary lobar ICH patients with neuropathologic samples containing vessels, 13 met criteria for probable CAA (see the Appendix). The multiple hemorrhagic lesions in these patients were demonstrated by CT or MRI scans performed at the time of each hemorrhage (n = 5) or by gradient-echo MRI sequences (n = 8) sensitive to old as well as recent hemorrhagic lesions.5,6⇓

All 13 patients (100%) diagnosed clinically with probable CAA demonstrated CAA pathologically ( table 1). Of the remaining 26 patients with possible CAA, 16 (62%) were pathologically diagnosed with CAA. No alternative pathologic cause of hemorrhage was found in the 10 samples without CAA (including four full postmortem brain examinations). None of the pathologic specimens showed severe hypertensive vascular changes, vascular malformations, or saccular aneurysms.

The overall prevalence of CAA was 29/39 (74%). The clinical characteristics of those patients with and without CAA are shown in table 2. Each of the 10 lobar ICH patients without CAA had one or more clinical risk factors for ICH: hypertension, use of warfarin, or use of an antiplatelet agent. Though the prevalence of each of these factors was higher in the group without CAA, none of the differences was significant.

Discussion.

Although neuropathologic examination remains the definitive diagnostic approach to CAA, our data suggest that a reliable diagnosis can be reached from clinical and radiographic information alone. Though the criteria are unlikely to remain perfectly accurate in future studies, the data suggest that their specificity will be at least comparable to clinical criteria used for other CNS diseases.

A second notable finding was the overall high prevalence of CAA (29/39, 74%) as the cause of lobar ICH. The prevalence was substantially greater than that seen (9/24, 38%) in a pathologic series of nontraumatic lobar ICH in Japanese patients between 1979 and 1990,7 perhaps reflecting a lower rate of hypertensive ICH in the western population or secular improvements in blood pressure control. The high rate of CAA in our study highlights its important role as a cause of primary lobar ICH in the elderly.

The alternative approach of validating the Boston criteria by comparing APOE genotypes among patients with clinically and pathologically diagnosed CAA also supported their accuracy.3 The idea underlying this approach was that if the diagnosis of probable CAA was accurate, then a group of patients with this clinical diagnosis should show similar elevations in frequency of the APOE ε2 and ε4 alleles1,8,9⇓⇓ as a group with pathologically proved CAA. Among 50 patients with a diagnosis of probable CAA, the allele frequencies for APOE ε2 (0.18) and ε4 (0.20) corresponded closely to the frequencies in 68 published cases of pathologically diagnosed CAA (0.17 and 0.26, respectively) and were significantly greater than those in populations of elderly controls or patients with hypertensive hemorrhage.3

Although APOE genotype is a risk factor for the presence and progression of CAA,1,8-10⇓⇓⇓ it appears insufficiently specific or sensitive on its own to diagnose this disorder in individual patients. In the current study, for example, eight of 24 (33%) patients with pathologically diagnosed CAA with available DNA samples carried neither of the APOE alleles (ε2 or ε4) associated with the disease. Larger validation studies will determine whether the Boston criteria should be modified to incorporate information on APOE genotype.

The sensitivity of the “probable CAA” diagnosis in this series was underestimated by the fact that most patients (23 of 39) did not have gradient-echo MRI scanning, generally because of the severity of the hemorrhagic strokes. Gradient-echo MRI is a sensitive, noninvasive technique for identifying small chronic hemorrhagic lesions in patients with lobar ICH,5,6⇓ and is thus expected to offer the greatest chance for detecting the additional lobar hemorrhages required for the diagnosis of probable CAA. Analysis of those patients who did undergo gradient-echo MRI supports its sensitivity for detection of CAA. Of the 11 patients with pathologically diagnosed CAA who had a gradient echo study, eight (73%) demonstrated the pattern of multiple hemorrhagic lesions diagnostic of probable CAA. CT or conventional MRI alone, conversely, reached the diagnosis of probable CAA in only five of the remaining 18 patients (28%) with pathologically diagnosed CAA.

Diagnosis of CAA may contribute to decision making in certain clinical situations, particularly when treatments such as anticoagulation or thrombolysis are under consideration. The proposed diagnostic criteria for CAA may also provide a foundation for identifying potential research subjects for the spectrum of antiamyloid treatment strategies likely to enter clinical trials in future years.

Appendix

Boston Criteria for Diagnosis of CAA-Related Hemorrhage *

• 1. Definite CAA

• Full postmortem examination demonstrating:

  • Lobar, cortical, or corticosubcortical hemorrhage

  • Severe CAA with vasculopathy†

  • Absence of other diagnostic lesion

• 2. Probable CAA with supporting pathology

• Clinical data and pathologic tissue (evacuated hematoma or cortical biopsy) demonstrating:

  • Lobar, cortical, or corticosubcortical hemorrhage

  • Some degree of CAA in specimen

  • Absence of other diagnostic lesion

• 3. Probable CAA

• Clinical data and MRI or CT demonstrating:

  • Multiple hemorrhages restricted to lobar, cortical, or corticosubcortical regions (cerebellar hemorrhage allowed)

  • Age ≥55 years

  • Absence of other cause of hemorrhage‡

• 4. Possible CAA

     Clinical data and MRI or CT demonstrating:

  • Single lobar, cortical, or corticosubcortical hemorrhage

  • Age ≥55 years

  • Absence of other cause of hemorrhage‡