Secondary Logo

Journal Logo


Intraoperative Frozen Section Analysis in Revision Total Joint Arthroplasty

Banit, Daxes M. MD; Kaufer, Herbert MD; Hartford, James M. MD

Author Information
Clinical Orthopaedics and Related Research: August 2002 - Volume 401 - Issue - p 230-238
  • Free


The infection rate for revision arthroplasty is reportedly four times higher than that of primary arthroplasty. 12,13 With the cost of revision arthroplasties being $10,000 to $50,000 more than primary arthroplasty 4,10,18,19,24,31 and the increased number of revision arthroplasties being done, it is important to have a reliable test which detects occult infection at the time of surgery to distinguish between septic and aseptic loosening. The treatment for aseptic loosening includes component removal, debridement of the intramembranous tissue, and proceeding with the revision. The treatment for septic loosening involves operative debridement, antimicrobial therapy, and delayed reimplantation. Many surgeons think two-stage reimplantation yields better results than one-stage reimplantation. 5,9,17,21,27,29,33,39 Alternative treatment options include debridement with retention of components and antimicrobial suppression, or debridement with resection arthroplasty.

To identify infections before revision surgery, the surgeon relies on several factors including patient history (history of pain, drainage, prolonged healing), physical examinations (drainage, sinus tracts, induration), serum laboratory test (erythrocyte sedimentation rate, C-reactive protein, cell count), radiologic studies (plain radiographs, nuclear medicine scans), and joint aspirations. The reliability of each factor has been questioned. 3,6,7,23,25,29,42

Currently, intraoperative frozen section analysis is used to detect the presence of an infection at the time of surgery. Although several authors have described the efficacy of intraoperative frozen section analysis, this test has been questioned for its reliability. 2,7,11,14,15,26,33 In the detection of infected arthroplasties no gold standard test is available, and operative cultures provide the best reference for comparision. 15,26 The current study determined the sensitivity, specificity, and positive and negative predictive values of frozen section histologic examination for patients having revision joint replacement surgery.


From August 1996 to March 2000, the results of frozen section analysis and culture results from a consecutive series of 121 patients who had revision total joint arthroplasties, independent of whether infection was suspected, were collected from one surgeon’s (JMH) experience at two institutions. Institutional review board approval and patient consent were obtained.

No patients received preoperative or intraoperative antibiotics before culture specimens were obtained. Intraoperatively, at least two specimens were sent for frozen and permanent section examinations. These specimens were obtained immediately on entering the joint space. The sampled areas were taken from any area suspicious for infection. Three culture swabs and tissue samples were obtained from the same areas sampled for the frozen sections. Specimens were obtained by sharp dissection. All cultures were sent for bacteriologic, mycologic, and mycobacteriologic examinations. All samples were handled and sealed in the sterile field to control intraoperative contaminations.

Laboratory policy is to do an immediate Gram’s stain. All culture plating was done in a laminar flow hood to provide quality control to reduce the incidence of contamination. Each aerobic and anaerobic swab and tissue was plated on chocolate agar, anaerobic blood agar, and Field’s enrichment media. Specimens for acid-fast bacilli were plated on Middlebrook agar and Bactec blood. Specimens for fungal elements were plated on brain infusion agar and inhibitory mold agar. Final cultures were reported negative if no growth was seen at 7 days for routine bacteriologic examinations, 4 weeks for mycologic examinations, and 8 weeks for mycobacteriologic examinations per institutional laboratory reporting protocol. Culture reports were reported positive if at least one colony growth occurred. The laboratory protocols at both institutions were the same.

The entire tissue sample submitted was embedded for frozen section analysis. For tissue samples too large for embedding, the pathologist determined the most suspicious area and prepared it. The sample was frozen in dimethyl butane at 30°C and stained with hematoxylin and eosin. The entire slide was examined for polymorphonuclear leukocytes microscopically. The specimens were examined under low power to determine the five most cellular areas. These areas were reassessed under high power (×40) and the neutrophil count of the most cellular region was reported. A frozen section was considered positive for infection if there were greater than 10 polymorphonuclear leukocytes per high power field identified in any of the sampled areas.

After correlating the results of the frozen section with the final culture results, the sensitivity, specificity, and positive and negative predictive values were calculated. All parameters were calculated using a polymorphonuclear leukocyte count greater than 10 per high power field to indicate infection. To reduce the threshold, a value greater than five polymorphonuclear leukocytes per high power field was recalculated. To further define the use of frozen sections in revision joint arthroplasties, the sensitivity, specificity, and positive and negative predictive values were calculated separately for total hip arthroplasty and total knee arthroplasty.

Statistics were done comparing sensitivity, specificity, and positive and negative predictive values using a z test with tails and Yates continuity correction (SigmaStat for Windows, SPSS Inc, Chicago, IL).


During the study from August 1996 to March 2000, 121 consecutive patients were evaluated. The followup of each patient was a minimum of 2 months to allow for final culture reporting. There were 92 men and 29 women. The average age of the patients at the time of the revision procedure was 64 years (range, 23–93 years). The surgical revisions included three shoulder hemiarthroplasties (three patients), 55 total knee arthroplasties (55 patients), and 63 total hip arthroplasties (63 patients).

All positive cultures were considered pathologic (Table 1). All patients with a positive culture result received an infectious disease consultation for appropriate treatment recommendations.

Summary of Positive Cultures

A comparison of the Gram’s stain with culture results revealed a 43% correlation (nine of 21 patients). The shoulder of one patient had a positive Gram’s stain and culture result. In the revision total hip arthroplasty group, four of 11 (36%) Gram’s stains correlated with a positive final culture result. In the revision total knee arthroplasty group, four of nine (44%) Gram’s stains correlated with a positive culture result. There were no positive Gram’s stains reported in any patients with negative final culture.

The results of the frozen section analysis were reported as less than five, five to nine, and at least 10 polymorphonuclear leukocytes per high power field. With the criterion greater than 10 polymorphonuclear leukocytes per high power field, the sensitivity of the frozen section was 67%. The specificity of the frozen section to correctly identify the absence of infection was 93%. The percent of patients with a positive frozen section having a true infection, or the positive predictive value, was 67%. The number of patients with a negative frozen section who were not infected, or negative predictive value, was 93% (Table 2).

Comparison of Frozen Section to Culture

If the diagnosis for the presence of an infection was changed to greater than five polymorphonuclear leukocytes per high power field, the sensitivity of the frozen section was 67%. The specificity of the frozen section would have changed to 84%. The corresponding positive predictive value of a frozen section would be decreased to 47% with a 92% negative predictive value. Statistical analysis comparing the sensitivity, specificity, and positive and negative predictive values using the criterion of greater than 10 versus greater than five polymorphonuclear leukocytes per high power field was nonsignificant (p > 0.05).

Using the criterion of at least 10 polymorphonuclear leukocytes per high power field to define a positive frozen section, a subgroup analysis was done separately on the patients who had a total hip arthroplasty and the patients who had a total knee arthroplasty (Tables 3, 4). The criterion of greater than 10 polymorphonuclear leukocytes per high power field was selected because of its better positive predictive value. Eleven of 63 total hip arthroplasties had positive results of culture specimens, but six of these 11 were false negative and were treated with primary exchange arthroplasty. Fifty-two total hip arthroplasties had negative results of culture specimens, but four were positive by frozen section analysis. Analysis of frozen sections in total hip arthroplasties revealed a 45% sensitivity, 92% specificity, 55% positive predictive value, and 88% negative predictive value. Nine of 55 total knee arthroplasties were infected; all nine were diagnosed correctly by frozen section. Forty-six total knee arthroplasties had negative culture specimens, but two were positive on frozen section analysis. Statistics revealed 100% sensitivity, 96% specificity, 82% positive predictive value, and 100% negative predictive value (Table 5). A z test comparison between the total hip arthroplasties versus the total knee arthroplasties revealed significant differences (p < 0.05) between sensitivity, and positive and negative predictive values. Specificity did not achieve statistical significance (Table 5).

Comparison of Frozen Section to Culture to Total Hip Arthroplasty
Comparison of Frozen Section to Culture in Total Knee Arthroplasty
Summary of Statistical Analysis

Of the 121 revision arthroplasties done, 21 replacements (21 patients) were infected as determined by culture results: one of three shoulders (one patient), 11 of 63 hips (63 patients), and nine of 55 knees (55 patients). All 21 joints were treated for infection. The one shoulder prosthesis was removed based on clinical impression even though the frozen sections were negative. This patient did not have reimplantation because of worsening of his medical problems and persistent infection. The other two patients had rerevision surgeries and remain symptom free.

Five of 11 hips (11 patients) had positive specimens by frozen section analysis. All five prostheses were removed during the initial revision. Two of these patients had a two-stage reimplantation after an appropriate course of antibiotics and are doing well more than 1 year after revision surgery (Fig 1). Three of five of the hip prostheses in three patients with positive culture specimens were removed during the initial procedure and have not been reimplanted secondary to the patients’ deteriorating medical conditions. Six patients (11 hips) with negative frozen sections and positive culture specimens and had primary reimplantation. All six patients were treated with an appropriate course of antibiotics after positive results were obtained. Four of six patients are symptom free 6 to 12 months after reimplantation, but the other two patients continue to have groin pain 1 year after surgery. Of the remaining 52 patients, 51 are doing well with no evidence of infection at followup of at least 6 months. One patient who had a total hip arthroplasty had a false negative culture with a delayed growth of coccidiomycosis. This patient had excision arthroplasty and is being treated with antifungal medication.

Fig 1A–B.
Fig 1A–B.:
(A) Anteroposterior and (B) lateral radiographs are shown of a patient with false negative results of culture of a total hip arthroplasty prosthesis. The patient had primary revision surgery with final cultures positive for Staphylococcus epidermidis. The patient postoperatively has had persistent groin pain despite 12 weeks of antibiotics.

All nine knee revision prostheses, which were infected as determined by culture specimens, had positive frozen section analysis (Fig 2). The nine patients who had a total knee arthroplasty with culture specimens that were positive had excision arthroplasty and an adequate course of antibiotics. Six of nine patients had a staged replacement and currently are symptom free. One patient chose to have a fusion and two patients no longer were candidates for implantation because of deteriorating medical conditions. Of the remaining 46 patients, two had false positive frozen section analysis and had a two-stage reimplantation. These two patients are doing well at followup. Forty-four of 46 patients had primary reimplantation. Two of these 44 patients had persistent knee pain, but followup evaluation for infection and aspiration have been negative.

Fig 2.
Fig 2.:
An anteroposterior radiograph shows the knee of a patient with a positive frozen section and cultures for Methicillin-resistant Staphylococcus aureus treated with staged reimplantation. More than 2 years after surgery the patient shows no signs of infection.


A laboratory test which differentiates between septic and aseptic loosening intraoperatively would allow for an improved success rate in revision surgery. Determination of a prosthetic infection may be difficult. An indolent infection may produce pain and no other symptoms indicative of an infection, such as erythema, swelling, or drainage. Routine laboratory studies including cell count, sedimentation rate, and C-reactive protein may be more valuable, but these tests are not unique to total joint arthroplasties, and other conditions can produce positive culture results. 1,7,16,34,38,40 Radiographs may be inconclusive for detection of infection. Nuclear medicine studies also have been used to detect periprosthetic infections. Technetium and indium-111 labeled scans have shown variable rates of sensitivity and specificity. 28,35,36,38

Bone marrow biopsy has been proposed as a diagnostic tool for osteomyelitis, but there are no data for revision arthroplasty surgery. Preoperative joint aspiration allows for direct sampling of the periprosthetic joint fluid with a reported 95% chance of detecting an infection, but there is a high false positive rate from skin flora contamination. 22,25,32 The sensitivity of prosthetic joint fluid aspiration has been reported to range from 60% to 75%. 3,22,25,32 Given the high rate of false positive and false negative results with aspirations, a more reliable test is needed.

Review of the data in the current study revealed the poor ability of the intraoperative Gram’s stain to predict the culture result. Only four of 11 (36%) patients who had revision total hip arthroplasty and four of nine patients who had (44%) revision total knee arthroplasty had a positive correlation between Gram’s stain and culture results.

Currently, intraoperative frozen section analysis has been proposed as a reliable test for detection of infection of a total joint replacement. 8,15,29,38 Unlike a joint aspiration, the site of tissue and joint fluid sampling is better controlled under direct observation. The technique of using frozen section analysis intraoperatively first was reported by Charosky et al. 8 Mirra et al 29 initially reported a series of 15 frozen sections from revision total joint arthroplasties. A positive frozen section was defined as greater than 10 polymorphonuclear leukocytes per high power field. In the series of Mirra et al, all of the frozen sections correlated with the final intraoperative cultures. In a later study by Mirra et al, 30 only 22 of 26 frozen section analyses correlated with final intraoperative cultures.

The current study was done prospectively to assess the ability of frozen section analysis to accurately determine the presence of infection in revision total joint replacement as determined by intraoperative cultures. Previous studies 2,11,14,15,26 assessing the effectiveness of frozen section analysis in determining infection are unclear as to whether final sections or intraoperative cultures were used as the gold standard for comparison. There is variability in the literature regarding the ability of frozen section analysis to detect a prosthetic infection. In a series of 33 revision arthroplasties, Feldman et al 15 reported 100% sensitivity and 96% specificity for detection of infection using more than five polymorphonuclear leukocytes per high power field as their cutoff. These results are in contrast to the findings of Fehring and McAlister 14 who reported a sensitivity of 18% and specificity of 89% in 107 revision joint replacements. Some inconsistencies in the literature arise from the retrospective nature of data collection. Another problem is in the definition of the study population. 2,15 Some studies examined patients who had the second stage of a reimplantation procedure. 11,20,37,41 Other groups studied patients who had initial revision arthroplasty secondary to loosening. 14,26 Another study combined both patient populations, those having the initial procedure or the reimplantation procedure. 33 A concern in the second stage of two-stage reimplantation procedures is the increased false positive rate in frozen sections compared with operative cultures from the residual inflammation. The studies of Feldman et al, 15 Lonner et al, 26 and Della Valle et al 11 showed a minimal risk of this occurrence.

The current study was done in a prospective manner examining all patients having initial revision of their arthroplasty. All patients, once entered in the study, were not recounted a second time when a reimplantation procedure was done. As stated previously, all cultures were considered positive. The results of the current study do not compare as favorably with those obtained by Lonner et al, 26 but the sensitivity and specificity are improved compared with those reported by Fehring and McAlister. 14 In the current series, meticulous surgical technique was used including aseptic skin preparation, sampling the most questionable areas, and using sharp dissection without electrocautery. There is no good explanation for the discrepancy between the results of Fehring and McAlister 14 and the results of the current study. In comparison with other studies, the current study only examined patients having initial revision surgery.

A reason for the decreased sensitivity may have been inconsistent tissue sampling. All samples were obtained at initial arthrotomy. Additional samples if indicated were analyzed later. No limit was placed on the number of areas sampled. An attempt was made to submit all suspicious areas of tissue for frozen section analysis, but focal areas of inflammation could have been missed.

A second source of error is handling of the specimens by the pathologist. If a tissue sample was large, a representative sample of tissue was prepared for frozen section analysis at the discretion of the pathologist. The prepared tissue was examined to identify the five most cellular areas before reporting the polymorphonuclear leukocytes per high power field count for the most cellular region. The possibility exists that a nonrepresentative sample may have been prepared for analysis.

A third source of error may be attributable to fastidious organisms that require special media or prolonged incubation times. The policy of observing bacteriologic cultures for 7 days, mycologic cultures for 4 weeks, and mycobacteriologic cultures for 8 weeks controls for the incubation periods of fastidious organisms. Another issue is contamination of cultures. The most common positive cultures in arthroplasties are Staphylococcus aureus and Staphylococcus epidermidis, which are common contaminants. Contamination from skin flora would increase the false negative rate. This increases the risk of labeling a culture with a contaminant positive. During the data collection period in the current study, all cultures were reported with at least moderate growth.

When statistical analysis was done on the use of frozen section analysis in total hip arthroplasty compared with total knee arthroplasty with a positive result defined as greater than 10 polymorphonuclear leukocytes per high power field, the sensitivity was 45% for total hip arthroplasty versus 100% for total knee arthroplasty. The respective specificity, positive predictive value, and negative predictive value of total hip arthroplasty were 92%, 55%, and 88%. The corresponding values for total knee arthroplasty were 96%, 82%, and 100%. All parameters except specificity were significantly different (p < 0.05). At the authors’ institutions, the ability to detect infections by frozen section analysis was better in total knee arthroplasty than in total hip arthroplasty.

Any explanation for this difference is speculative. One reason may be that the hip with its soft tissue coverage is better at suppressing infections. This possibly may lead to more indolent infections that were found with cultures, but not with frozen section analysis. The knee with less soft tissue coverage also may not be able to suppress an infection.

According to the current study, the possible presence of an infection is suggested by at least 10 polymorphonuclear leukocytes per high power field. With a sensitivity of 67% and specificity of 93%, the frozen section cannot be used as the sole determinant of infection. At the authors’ institutions, the ability to detect infection by frozen section analysis is better in patients having total knee arthroplasty with a 100% sensitivity. The data obtained intraoperatively need to be correlated with the history, physical examination, radiologic studies, and joint aspiration when determining the need for a staged reimplantation.


1. Aalto K, Osterman K, Peltola H, Rasanen J: Changes in erythrocyte sedimentation rate and C-reactive protein after total hip arthroplasty. Clin Orthop 184:118–120, 1984.
2. Athansou NA, Pandey R, DeSteiger R, Crook D, McLardy Smith P: Diagnosis of infection by frozen section during revision arthroplasty. J Bone Joint Surg 77B:28–33, 1995.
3. Barrack RL, Harris WH: The value of aspiration of the hip before total hip arthroplasty. J Bone Joint Surg 75A:66–76, 1993.
4. Barrack RL, Swahney J, Hsu J, Cofield RH: Cost analysis of revision total hip arthroplasty. Clin Orthop 369:175–178, 1999.
5. Buchholz HW, Elson RA, Engelbrecht E, et al: Management of deep infection of total hip replacement. J Bone Joint Surg 63B:342–353, 1981.
6. Canner GC, Steinberg ME, Heppenstall RB, Balderston R: The infected hip after total hip arthroplasty. J Bone Joint Surg 66A:1393–1399, 1984.
7. Carlson AS: Erythrocyte sedimentation rate in infected and non infected total hip arthroplasties. Acta Orthop Scand 49:287–290, 1978.
8. Charosky CB, Bullough PG, Wilson Jr PD: Total hip replacement failures: A histological evaluation. J Bone Joint Surg 55A:49–58, 1973.
9. Cherney DL, Amstutz HC: Total hip replacement in the previously septic hip. J Bone Joint Surg 65A:1256–1265, 1983.
10. Coyte PC, Hawker G, Croxford R, Wright JG: Rates of revision knee replacement in Ontario, Canada. J Bone Joint Surg 81A:773–782, 1999.
11. Della Valle CJ, Bogner BA, Desai P, et al: Analysis of frozen sections of intraoperative specimens obtained at the time of reoperation after hip or knee resection arthroplasty for the treatment of infection. J Bone Joint Surg 81A: 684–689, 1999.
12. Dupont JA: Significance of operative cultures in total hip arthroplasty. Clin Orthop 211:122–126, 1986.
13. Dupont JA, Charnley J: Low friction arthroplasty of the hip for failures of previous operations. J Bone Joint Surg 54B:77–87, 1972.
14. Fehring TK, McAlister Jr JA: Frozen histologic section as a guide to sepsis in revision joint arthroplasty. Clin Orthop 304:229–237, 1994.
15. Feldman DS, Lonner JH, Desai P, Zuckerman JD: The role of intraoperative frozen sections in revision total joint arthroplasty. J Bone Joint Surg 77A:1807–1813, 1995.
16. Forester IW, Crawford R: Sedimentation rate in infected and uninfected total hip arthroplasty. Clin Orthop 168:48–52, 1982.
17. Garvin KL, Evans BG, Salvati EA, Brause BD: Palacos gentamicin for treatment of deep periprosthetic hip infection. Clin Orthop 298:97–105, 1994.
18. Heck DA, Melfi CA, Mamlin LA, et al: Revision rates after knee replacement in the United States. Med Care 36:661–669, 1998.
19. Herbert CK, Williams RE, Levy RS, Barrack RL: Cost of treating an infected total knee replacement. Clin Orthop 331:140–145, 1996.
20. Hughes PW, Salvati EA, Wilson PD, Blumenfeld EL: Treatment of subacute sepsis of the hip by antibiotics and joint replacement: Criteria for diagnosis with evaluation of twenty-six cases. Clin Orthop 141:143–157, 1979.
21. Hunter GA: Results of insertion of a total hip prosthesis after sepsis. J Bone Joint Surg 61B:422–423, 1979.
22. Insall JN, Thompson FM, Brause DB: Two stage reimplantation for the salvage of infected total knee arthroplasty. J Bone Joint Surg 65A:1087–1098, 1983.
23. Johnson JA, Christie MJ, Sandler MP: Detection of occult infection following total joint arthroplasty using sequential technetium bone scintigraphy and indium-111 WBC imaging. J Nucl Med 29:1347–1353, 1988.
24. Lavernia CJ, Drakeford MK, Tsao AK, et al: Revision and primary hip and knee arthroplasty: A cost analysis. Clin Orthop 311:136–141, 1995.
25. Levitsky KA, Hozack WJ, Balderston RA, et al: Evaluation of a painful prosthetic joint: Relative value of bone scan, sedimentation rate, and joint aspiration. J Arthroplasty 6:237–244, 1991.
26. Lonner JH, Desai P, DiCesare PE, Steiner G, Zuckerman JD: The reliability of analysis of intraoperative frozen sections for identifying active infection during revision hip and knee arthroplasty. J Bone Joint Surg 78A:1553–1558, 1996.
27. McDonald DJ, Fitzgerald Jr RH, Ilstrup DM: Two stage reconstruction of a total hip arthroplasty because of infection. J Bone Joint Surg 71A:828–834, 1989.
28. Merkel KD, Brown ML, Dewangee MK, Fitzgerald Jr RH: Comparison of indium-labeled leukocyte imaging with sequential technetium-gallium scanning in the diagnosis of low grade musculoskeletal sepsis: A prospective study. J Bone Joint Surg 67A:465–476, 1985.
29. Mirra JM, Amstutz HC, Matos M, Gold R: The pathology of joint tissue and its clinical relevance in prosthesis failure. Clin Orthop 117:221–240, 1976.
30. Mirra JM, Marder RA, Amstutz HC: Pathology of failed total joint arthroplasty. Clin Orthop 170:175–183, 1982.
31. Mushinski M: Average charges for hip replacement surgeries: United States, 1997. Stat Bull Metrop Insur Co 80:32–40, 1999.
32. O’Neill DA, Harris WH: Failed total hip replacement: Assessment by plain radiographs, arthrograms, and aspiration of the hip joint. J Bone Joint Surg 66A:540–546, 1984.
33. Pace TB, Jeray KJ, Latham JT: Synovial tissue examination by frozen section as an indicator of infection in hip and knee arthroplasty in community hospitals. J Arthroplasty 12:64–69, 1997.
34. Peltola H, Vahvanen V, Aalto K: Fever, c-reactive protein, and erythrocyte sedimentation rate in monitoring recovery from septic arthritis: A preliminary study. J Pediatr Orthop 4:170–174, 1984.
35. Pring DJ, Henderson RG, Keshavarzian A, et al: Indium-granulocyte scanning in the painful prosthetic joint. Am J Roentgenol 147:167–172, 1986.
36. Rand JA, Brown ML: The value of indium-111 leukocyte scanning in the evaluation of painful or infected total knee arthroplasties. Clin Orthop 259:179–182, 1990.
37. Rand JA, Bryan RS: Reimplantation for the salvage of an infected total knee arthroplasty. J Bone Joint Surg 65A:1081–1086, 1983.
38. Rand JA, Fitzgerald Jr RH: Diagnosis and management of the infected total knee arthroplasty. Orthop Clin North Am 20:201–210, 1989.
39. Sanzen L, Carlsson AS, Josefsson G, Lindberg LT: Revision operation on infected total hip arthroplasty: Two to nine years follow up study. Clin Orthop 229:165–172, 1990.
40. Shih LY, Wu JJ, Yang DJ: Erythrocyte sedimentation rate and C-reactive protein values in patients with total hip arthroplasty. Clin Orthop 225:238–246, 1987.
41. Walker RH, Schurman DJ: Management of infected total knee arthroplasties. Clin Orthop 186:81–89, 1984.
42. Wukich DK, Abre SH, Callaghan JJ, et al: Diagnosis of infection by preoperative scintigraphy with indium labeled white blood cell. J Bone Joint Surg 69A: 1353–1360, 1987.
© 2002 Lippincott Williams & Wilkins, Inc.