Temporal artery biopsy (TAB) can confirm the diagnosis of giant cell arteritis (GCA), but it is not without risk.1 Therefore, markers of inflammation such as the erythrocyte sedimentation rate (ESR), C-reactive protein level (CRP), and platelet counts are often used to select patients for TAB.2–4 Our aim was to investigate which demographic and laboratory data predict biopsy-proven GCA in patients undergoing biopsy to generate hypotheses about how physicians integrate this information into TAB decisions.
Methods
We retrospectively reviewed patient medical records of all TABs performed at 8 institutions from January 1, 2007, through April 30, 2012, abstracting age at biopsy, sex, preoperative ESR (adjusted for age and sex), CRP level (elevated or not), platelet count (continuous), corticosteroid use, and whether the biopsy was GCA positive or negative. Patients were excluded if they were younger than 50 years or if they had been using corticosteroids for more than 14 days in the 30-day period prior to TAB. Institutional review board approval was obtained by each investigator at the respective institutions (University of Michigan, Louisiana State University Health Sciences Center, Marshfield Clinic, Hadassah Medical Center, Bethesda Neurology, Franciscan St Anthony Memorial Hospital, Goldschleger Eye Institute, and University of Kentucky) and informed consent was waived because the study was retrospective.
Multiple logistic regression estimated the association between all measured variables and biopsy results after adjusting for clustering within institution (Stata version 13 statistical software; StataCorp LP). Multiple imputation using chained equations investigated complete case bias due to missing data (Stata version 13 statistical software).5 For all analyses, P < .05 was considered statistically significant. All patients included were selected for biopsy based on clinical and laboratory data (eFigure in the Supplement). Consequently, variables that continue to predict a positive biopsy have retained or “leftover” predictive value, suggesting that the predictive information contained within these variables is being underused by clinicians in selecting patients for TAB. See eAppendix 1 in the Supplement for additional details.
Results
We identified 544 patients who had biopsy; 140 met exclusion criteria. Of the remaining 404 patients, 271 (67.1%) were women and 90 (22.3%) had biopsy-proven GCA. Additional patient characteristics are summarized in Table 1. Biopsy results by institution are shown in eTable 1 in the Supplement. The ESR, CRP level, and platelet count were missing in 15 patients (3.7%), 98 (24.3%), and 101 (25.0%), respectively.
Table 1.
Patient Characteristics
| Variable | TAB Result | OR (95% CI) | P Value | |
|---|---|---|---|---|
| Positive | Negative | |||
| Age, mean (SD), y | 77.2 (7.9) | 72.8 (10.1) | 1.05 (1.02–1.08) | <.001 |
| Sex, No. (%) | ||||
| Female | 58 (21.4) | 213 (78.6) | 0.86 (0.53–1.41) | .55 |
| Male | 32 (24.1) | 101 (75.9) | 1 [Reference] | |
| ESR | ||||
| Median (IQR), mm/h | 69.0 (51.0–89.0) | 53.5 (29.5–78.5) | 1.01 (1.01–1.02) | .001 |
| Elevated for age and sex, No. (%) | 66 (25.3) | 195 (74.7) | 1.94 (1.11–3.40) | .02 |
| Platelet count, median (IQR), ×103/μL | 391.0 (274.0–491.5) | 275.0 (210.0–366.0) | 1.01 (1.00–1.01) | <.001 |
| CRP level elevated, No. (%) | 68 (30.8) | 153 (69.2) | 2.70 (1.38–5.31) | .004 |
Abbreviations: CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; IQR, interquartile range; OR, odds ratio; TAB, temporal artery biopsy.
SI conversion factor: To convert platelet count to ×109 per liter, multiply by 1.0.
After controlling for sex, CRP level, adjusted ESR, and clustering within institutions, age at biopsy and platelet count continued to predict the biopsy result (Table 2). Results were unchanged after multiple imputation of missing values for platelet count, adjusted ESR, and elevated CRP level (eTable 2 and eTable 3 in the Supplement).
Table 2.
Multiple Logistic Regression Model Predicting Temporal Artery Biopsy Result in 239 Patientsa
| Covariate | OR (95% CI) | P Value |
|---|---|---|
| Age, per 1-y increase | 1.06 (1.03–1.09) | <.001 |
| Female | 1.01 (0.46–2.21) | .98 |
| Adjusted ESR | 1.02 (0.59–1.79) | .94 |
| Platelet count, per 1 ×103/μL increase | 1.01 (1.00–1.01) | <.001 |
| CRP level | 1.64 (0.97–2.79) | .07 |
Abbreviations: CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; OR, odds ratio.
Area under the receiver operating characteristic curve = 0.731.
Hosmer-Lemeshow test was not significant (P = .76), suggesting that the model fit the data.
Discussion
Our data suggest that physicians integrate ESR and patient sex into TAB decisions but may not completely account for age and platelet count. Unlike ESR and CRP level, the diagnostic utility of platelet count has only recently been recognized, potentially explaining why physicians use it with less certainty. Predictions in eTable 4 and eAppendix 2 in the Supplement demonstrate how the addition of platelet count and age, when fully used to generate a posttest probability, could influence a clinician’s decision to perform a biopsy.
Important limitations must be acknowledged. This study was retrospective and therefore susceptible to selection bias and confounding by unmeasured or poorly measured laboratory variables. Both the complete case and imputed model demonstrated the same findings; however, missing data, including approximately 25% of platelet data, limit our conclusions. The largest centers had similar positive biopsy rates, but overall this varied greatly between sites. Although likely partly due to the small samples at some sites, this also suggests that variation in criteria used by even specialist physicians may exist when making biopsy decisions. Lastly, some institutions contributed few cases and there was no central biopsy reading center to confirm the tissue diagnosis or adequacy.
Recognizing the multiple limitations of retrospective design and missing data, these results suggest that clinicians are fully integrating the predictive diagnostic information in ESR and patient sex, but not platelet count or age. A clinical risk prediction tool may help clinicians overcome the hurdle of applying diagnostic information to individual patients at risk for GCA.
Supplementary Material
Acknowledgments
Funding/Support: Dr De Lott is supported by grant 5T32NS7222-32 from the National Institutes of Health. Dr Burke is supported by grants K08 NS082597 and R01 MD008879 from the National Institutes of Health.
Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Footnotes
Supplemental content at jamaophthalmology.com
Author Contributions: Dr De Lott had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: De Lott, Burke.
Acquisition, analysis, or interpretation of data: De Lott.
Drafting of the manuscript: De Lott.
Critical revision of the manuscript for important intellectual content: De Lott, Burke.
Statistical analysis: De Lott.
Administrative, technical, or material support: De Lott, Burke.
Study supervision: Burke.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Burke reported providing expert testimony for Sullivan, Ward, Asher and Patton, PC; receiving fees for lecturing from the University of Rochester; receiving fees for articles written for the American Academy of Neurology; and receiving fees for adjudicating clinical events for AstraZeneca. No other disclosures were reported.
Group Members: The Michigan Neuro-Ophthalmology Research Consortium includes Marie D. Acierno, MD, Louisiana State University Health Sciences Center, New Orleans; Adeela Alizai, MD, Franciscan St Anthony Memorial Hospital, Michigan City, Indiana; Dennis Anderson, MD, Marshfield Clinic, Marshfield, Wisconsin; Wayne Cornblath, MD, University of Michigan, Ann Arbor; Lindsey B. De Lott, MD, University of Michigan, Ann Arbor; Shlomo Dotan, MD, Hadassah Medical Center, Jerusalem, Israel; David Katz, MD, Bethesda Neurology, Bethesda, Maryland; Edward Margolin, MD, University of Toronto, Toronto, Ontario, Canada; Iris Mizrachi, MD, Goldschleger Eye Institute, Tel Hashomer, Israel; Padmaja Sudhakar, MD, University of Kentucky, Lexington; and Jonathan Trobe, MD, University of Michigan, Ann Arbor.
Contributor Information
Lindsey B. De Lott, Department of Ophthalmology and Visual Sciences, University of Michigan Kellogg Eye Center, Ann Arbor.
James F. Burke, Department of Neurology, University of Michigan, Ann Arbor.
References
- 1.Murchison AP, Bilyk JR. Brow ptosis after temporal artery biopsy: incidence and associations. Ophthalmology. 2012;119(12):2637–2642. doi: 10.1016/j.ophtha.2012.07.020. [DOI] [PubMed] [Google Scholar]
- 2.Kermani TA, Schmidt J, Crowson CS, et al. Utility of erythrocyte sedimentation rate and C-reactive protein for the diagnosis of giant cell arteritis. Semin Arthritis Rheum. 2012;41(6):866–871. doi: 10.1016/j.semarthrit.2011.10.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Lincoff NS, Erlich PD, Brass LS. Thrombocytosis in temporal arteritis rising platelet counts: a red flag for giant cell arteritis. J Neuroophthalmol. 2000;20(2):67–72. [PubMed] [Google Scholar]
- 4.Walvick MD, Walvick MP. Giant cell arteritis: laboratory predictors of a positive temporal artery biopsy. Ophthalmology. 2011;118(6):1201–1204. doi: 10.1016/j.ophtha.2010.10.002. [DOI] [PubMed] [Google Scholar]
- 5.Rubin DB. Multiple imputation after 18+ years. J Am Stat Assoc. 1996;91(434):473–489. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.