Does This Patient With Diabetes Have Osteomyelitis of the Lower Extremity?

A 52-year-old woman is referred from the emergency department with a diabetic foot ulcer. She has type 1 diabetes mellitus that was first diagnosed at age 12 years. Her condition is complicated by nephropathy, retinopathy, and peripheral vascular disease. She has recently noticed erythema, swelling, and pain over the left foot. On physical examination, she has a pulse of 90/min, blood pressure of 136/84 mm Hg, and temperature of 36.1°C. Pedal pulses are diminished. There is a 2.2 × 1.5-cm ulcer in the toe webbing that probes to underlying bone. Investigations reveal a white blood cell count of 9500/μL and an erythrocyte sedimentation rate (ESR) of 75 mm/h. Wound swab Gram stain reveals gram-positive cocci and gram-negative bacilli. Radiographs of the foot identify soft tissue swelling and cortical erosion in the area of the ulcer. Should magnetic resonance imaging (MRI) of the foot be ordered?

A 64-year-old man is referred with a nonhealing ulcer. He has a small wound overlying the fourth metatarsal head of his right foot that was first noted 3 weeks ago. He was prescribed a 10-day course of antibiotics and referred to a podiatrist for further foot care. This man has a history of type 2 diabetes complicated by neuropathy, nephropathy, and retinopathy. He has no prior history of neuropathic diabetic foot ulcer. On examination, the wound is round, measuring 1 cm in diameter, with associated erythema and swelling. It does not probe to bone. Laboratory investigations reveal a white blood cell count of 14 500/μL, an ESR of 25 mm/h, and a hemoglobin A_1c of 8.2%. Foot radiographs reveal no abnormalities. Gram stain and culture of the wound swab identify gram-positive cocci and gram-negative bacilli. How likely is osteomyelitis?

Foot-related complications account for up to 20% of all diabetes-related admissions in the North American diabetic population.¹ Medicare data from 1995 to 1996 revealed that US $1.45 billion was spent directly on diabetic foot ulcer care in the United States alone.² Quiz Ref IDDiabetic foot problems are the most common cause of nontraumatic amputations,³ with infection responsible for a large proportion of these cases.⁴ When inadequately treated, osteomyelitis increases the risk of amputation.⁵ The perioperative 30-day mortality of patients having lower extremity amputation is reported to be 7.4%.⁶

The diagnosis of lower limb osteomyelitis in patients with diabetes remains a challenge. The classic signs and symptoms of infection may be absent or masked by the coexistence of vascular disease and neuropathy. The gold standard for the diagnosis of osteomyelitis is a bone biopsy and culture. This invasive procedure is not always practical⁷ and may be contraindicated in patients with diabetes and severe peripheral vascular disease. It is therefore important to determine the features of the history, physical examination, and preliminary investigations that aid in the diagnosis of osteomyelitis to minimize costly and invasive investigations and initiate appropriate and timely therapy. Herein, we summarize the test characteristics of the history, physical examination, routinely available laboratory measurements and radiographs, and MRI for evaluating lower extremity osteomyelitis in patients with diabetes.

The evaluation for osteomyelitis and other infections begins with an assessment of a patient's diabetes and the risk factors that predispose a patient to skin breakdown. The history taking should include duration of diabetes, glycemic control, microvascular or macrovascular disease, and presence of peripheral neuropathy and peripheral vascular disease. In addition, the clinician should inquire about recent trauma and history of ulcers.

On physical examination, the clinician should assess for local and systemic features of infection such as fever, chills, hypotension, and presence and appearance of wound or ulcer (erythema, swelling, purulence, size, and depth).^{7 - 10} The clinician should note the presence of foot deformities, tenderness, neuropathy, and venous or arterial insufficiency.^{11 - 12} Because of neuropathy, patients may not perceive foreign bodies within the ulcer.

It is sometimes difficult to distinguish between lower extremity ulcers due to diabetes and those caused primarily by venous or arterial insufficiency. Quiz Ref IDVenous ulcers are typically found above the medial or lateral malleoli and frequently have irregular borders. Arterial ulcers often affect the toes or the shins, with the borders of the ulcer being pale and appearing as if they have been punched out. These ulcers may lack granulation tissue and are typically painful in the absence of coexisting neuropathy. Diabetic ulcers usually occur at areas of increased pressure, such as the sole of the foot, or areas where shoes have rubbed against the skin.¹³ Although any ulcer is a risk factor for osteomyelitis, the traumatized skin in a patient with vascular insufficiency is also prone to this disease process.¹⁴

Ulcer Area. One method for quantifying ulcer size is to multiply the longest and widest diameters of the lesion.⁹ This may not be a completely accurate estimation of ulcer area, as some ulcers may be round or irregularly shaped.

Probe-to-Bone Test. Quiz Ref IDThe probe-to-bone test of a foot ulcer is performed at the bedside with a sterile, blunt, stainless steel probe. The examiner gently probes the wound for the presence of a rock-hard, gritty structure at the wound base in the absence of any intervening soft tissue. The presence of such a finding indicates a positive probe-to-bone result, whereas the inability to probe the base of a wound to periosteum or bone is a negative result.¹⁰

Wagner Grade. Wagner⁸ developed a scale to grade foot ulcers based on observations. Foot ulcers are graded from 0 to 5 based on depth of lesion and presence or absence of features of infection and/or gangrene (Box 1). A limitation of the Wagner grading scale is that all deep tissue infections (including abscess, tendinitis, and osteomyelitis) are accounted for in a single grade.

We searched the MEDLINE electronic database for English-language articles between 1966 and March 2007 using the following search terms: osteomyelitis, diabet$, signs and symptoms, physical examination, diagnosis, diagnostic tests, and “sensitivity and specificity.” We identified additional references by modifying a previously published search strategy.¹⁵ This strategy combined 9 exploded Medical Subject Headings (physical examination, medical history taking, professional competence, “sensitivity and specificity,” reproducibility of results, observer variation, “diagnostic tests, routine,” decision support techniques, and Bayes theorem). We then took the intersection of this set with osteomyelitis and diabetes mellitus (exploded). We identified additional articles through a hand search of references from retrieved articles, previous reviews, and polling experts. The titles and abstracts (when available) of the articles retrieved were evaluated to determine their eligibility for our review. Publications in abstract and letter form were included to minimize publication bias.

Articles were included for review if they fulfilled all of the following criteria: (1) they were original studies describing historical features, physical examination, laboratory investigations, or plain radiograph in the diagnosis of lower extremity osteomyelitis in patients with diabetes mellitus, (2) data could be extracted to construct 2 × 2 tables or the article reported operating characteristics of the diagnostic measure, and (3) the diagnostic test was compared with a reference standard. Studies in pediatric populations or mixed populations of patients with and without diabetes were excluded. Data from a single high-quality meta-analysis, not conforming to the prespecified search criteria and data extraction, were used to summarize the test characteristics of MRI in patients with diabetes who were thought to have osteomyelitis.

Bone biopsy is the gold standard for the diagnosis of osteomyelitis. Ideally, bone specimens should have both microbiological and histological analysis; however, we required the gold standard for osteomyelitis to be culture or histological results because patients had often received antibiotic therapy before bone biopsy or surgical intervention, thus lowering the yield of bone culture.

One author (S.B.) identified potential articles by screening the retrieved articles and by searching through the bibliographies of these articles. Two authors independently reviewed articles for quality and extracted the operating characteristics of the diagnostic tests. Each article was rated using a topic-specific quality rating scale that used published principles¹⁶ as well as a modified quality checklist specific to the Rational Clinical Examination series¹⁷ (Box 2).

^aWhen a finding occurs in 0 of 30 patients, the upper confidence limit for 0% is 10%. A study with 0 of 20 patients having the finding would have an upper confidence limit of 14%.

Likelihood ratios (LRs) predicting the presence of osteomyelitis, given a positive test result (sensitivity/1 − specificity) and a negative test result (1 − sensitivity/specificity) were calculated for each outcome of interest using published raw data.¹⁸ Where 2 or more studies examined the same clinical variable, we calculated summary LRs and 95% confidence intervals (CIs) using the DerSimonian and Laird random-effects approach.¹⁹ Estimated variances of LRs were computed using the usual methods for ratios of proportions,²⁰ with their reciprocals used as study weights. In studies with a zero cell count, the value 0.5 was added to each cell count to permit use of this variance estimation. All analyses were performed using R software, version 2.01.²¹

The electronic literature search identified 279 studies, of which 21 met our inclusion criteria and form the basis of our review (Table 1 and Figure).^{9 - 10 ,22 - 40} The included studies accounted for a total of 1027 patients. Three studies reported sensitivity or specificity but not both.^{23 ,32 ,36} Eight studies prospectively evaluated patients.^{9 - 10 ,22 - 24 ,30 - 32} One meta-analysis of the test performance of MRI for the diagnosis of osteomyelitis of the foot and ankle included 11 studies in patients with diabetes and accounted for a total of 275 patients.⁴¹

The prevalence of osteomyelitis in the selected studies ranged from 12% to 100%. A retrospective cohort study of 8905 patients with diabetes found that 15% of those with a foot ulcer developed osteomyelitis at or after diagnosis.⁴² The prior probability that a leg ulcer in a patient with diabetes will end up being osteomyelitis is more likely to reflect this value than the values found in the highly selected group of patients in the included studies.

There were no studies identified that addressed the precision of symptoms, signs, or investigations in the diagnosis of lower extremity osteomyelitis.

No studies were identified that addressed the utility of any component of the history in the diagnosis of osteomyelitis.

Seven studies assessed physical examination findings in the diagnosis of lower extremity osteomyelitis.^{9 - 10 ,24 ,26 ,30 ,39 - 40} The 7 studies included a total of 509 patients. No studies examined the test characteristics of a combination of findings. Temperature was examined in a single study, which received a level V quality rating (nonindependent comparisons of test results with a standard of uncertain validity) and demonstrated a poor sensitivity of 19%; patients without osteomyelitis were not included, so specificity and LRs could not be calculated.³⁹

“Bone exposure” as a single finding suggests osteomyelitis (LR, 9.2; 95% CI, 0.57-146; Table 2). While the absence of bone exposure, defined as visualization of bone either directly or after probing, has a much narrower CI and lowers the likelihood of osteomyelitis, the LR of 0.70 (95% CI, 0.53-0.92) is not low enough to rule out osteomyelitis unless the pretest probability is already very low.

Good-quality evidence⁹ suggests that an ulcer area larger than 2 cm² (calculated as described above) makes osteomyelitis more likely (LR, 7.2; 95% CI, 1.1-49), while an ulcer area smaller than 2 cm² decreases the likelihood of osteomyelitis by about half (LR, 0.48; 95% CI, 0.31-0.76). The presence or absence of ulcer inflammation (erythema, swelling, purulence) does not modify the probability of disease (positive LR, 1.5; 95% CI, 0.51-4.7; negative LR, 0.84; 95% CI, 0.56-1.3).⁹

The probe-to-bone test has been evaluated in 3 studies.^{10 ,26 ,30} Shone et al²⁶ and Lavery et al³⁰ studied the probe-to-bone test in an outpatient setting, while Grayson et al¹⁰ prospectively evaluated the probe-to-bone test in 75 patients with suspected severe limb-threatening infection. A positive probe-to-bone test result increases the likelihood of osteomyelitis (summary LR, 6.4; 95% CI, 3.6-11). A negative probe-to-bone test result has a summary LR of 0.39 (95% CI, 0.20-0.76).

Three studies described the diagnostic accuracy of clinical gestalt.^{9 ,24 ,40} One study described the diagnostic accuracy of “clinical judgment”⁹ and 2 studies used the Wagner grading scale.^{24 ,40} Because the Wagner grade is a subjective assessment, we elected to summarize the data together with clinical judgment. The clinical impression of osteomyelitis, without formal rules or weighting of the findings, increases the likelihood of osteomyelitis about 5-fold (summary LR, 5.5; 95% CI, 1.8-17). When a clinician judges that osteomyelitis is absent, the likelihood decreases (summary LR, 0.54; 95% CI, 0.30-0.97) (Table 2). These data suggest that clinicians might be more proficient at detecting the presence of osteomyelitis than detecting its absence.

Four studies evaluated the utility of laboratory investigations in the diagnosis of osteomyelitis^{9 ,36 ,38 - 39} (Table 2). The 4 studies included a total of 108 patients, a rather small number. Three studies of varying quality evaluated the utility of ESR.^{9 ,36 ,39} The cutoff used to define an elevated ESR varied among the studies. An ESR of more than 70 mm/h increases the probability of the diagnosis of osteomyelitis with a summary positive LR of 11 (95% CI, 1.6-79.0), while an ESR of less than 70 mm/h has a summary LR of 0.34 (95% CI, 0.06-1.90)^{9 ,36} (Table 2) with a 95% CI that crosses 1.0.

The value of an elevated white blood cell count was examined in a single study and demonstrated poor sensitivity (range, 14%-54%) regardless of the cutoff studied.³⁹ Patients without osteomyelitis were not included in this study, so specificity and LRs could not be calculated.³⁹

Swab culture³⁸ was examined in a single study and had no diagnostic utility, with positive and negative LRs of 1.0 (95% CI for positive LR, 0.65-1.5; 95% CI for negative LR, 0.08-13) (Table 2), suggesting that in patients with suspected osteomyelitis, a positive swab was equally common in patients with and without biopsy-proven osteomyelitis.

Sixteen studies that included 567 patients assessed the accuracy of plain radiographs in the diagnosis of lower extremity osteomyelitis.^{9 ,22 - 25 ,27 - 29 ,31 - 35 ,37 - 38 ,40} The studies included a broad range of patients and plain radiographs were often studied concurrently with another radiologic technique (magnification radiography, ultrasound, bone scan, or MRI). Only 6 of the 16 studies were prospective.^{9 ,22 - 24 ,31 - 32} Some studies used 3 views^{9 ,28 ,35} and others used 2 views,^{24 ,31} but for the most part it was unclear how many views were taken.^{22 - 23 ,25 ,27 ,29 ,32 - 35 ,37 - 38 ,40}

Quiz Ref IDThe characteristic signs of osteomyelitis on plain radiograph include focal loss of trabecular pattern, periosteal reaction, and frank bone destruction, often accompanied by soft tissue swelling.⁴³ The studies that clearly stated a definition for a positive radiographic result incorporated 1 or more of these features and were used to calculate summary LRs (Table 3).

Radiographic results alone appear to be marginally useful if positive, with a summary LR of 2.3 (95% CI, 1.6-3.3), and less useful when negative for osteomyelitis (summary LR, 0.63; 95% CI, 0.51-0.78) (Table 3). We found no studies that address the utility of serial radiographs in the diagnosis of osteomyelitis.

Soft Tissue Culture. We identified 3 studies^{44 - 46} that compared superficial swab culture with bone culture. Swab culture identified the identical pathogens as bone culture in only 19% to 36% of isolates. Superficial swab cultures do not reliably predict bone microorganisms.

Other Radiological Modes Including MRI. A comprehensive review of all radiological modes in the evaluation and diagnosis of lower extremity osteomyelitis is beyond the scope of this review. The utility of nuclear scans and MRI in the diagnosis of osteomyelitis have been recently reviewed elsewhere.^{41 ,47} A recent meta-analysis reported that nuclear imaging (technetium, indium, and white blood cell scans) lacks specificity (62%-88.5%) in the diagnosis of osteomyelitis.⁴⁷ Kapoor et al⁴¹ recently summarized the test characteristics of MRI in foot osteomyelitis. Eleven of the 16 MRI studies included exclusively patients with diabetes. Magnetic resonance imaging was shown to have a sensitivity of 90% (range, 77%-100%) and a specificity of 83% (range, 40%-100%) in all patients and a summary positive LR of 3.8 (95% CI, 2.5-5.8) and a summary negative LR of 0.14 (95% CI, 0.08-0.26) in patients with diabetes.⁴¹ Magnetic resonance imaging was also shown to be more accurate than technetium Tc 99m bone scan, plain radiography, and white blood cell scan. The overall accuracy (ie, the weighted average of the sensitivity and specificity) of MRI is 89% (95% CI, 83.0-94.5).

There are several important limitations to consider when interpreting the presented studies. We identified only 10 studies that attained a level II or III quality rating, with the remainder considered to be of overall poor quality. The majority of studies evaluated were of retrospective design and had unblinded protocols. Aside from study design, the physical examination findings and maneuvers were often described in a single study^{9 ,38 - 39} and the precision was not addressed.^{9 - 10 ,24 ,26 ,40} The accuracy of combinations of variables is unknown.

Patients were studied most often in tertiary care centers, thus resulting in a selection bias. For example, 1 study that evaluated the probe-to-bone test was in patients with severe limb-threatening infection.¹⁰

The study by Newman et al⁹ that discussed the diagnostic accuracy of clinical impression did not clearly define what this entailed, thus limiting reproducibility.

In the studies that addressed the utility of plain radiographs, details of the definition used to interpret the radiograph were missing in more than half of the studies.^{22 - 25 ,31 - 33 ,37 ,40} Also not defined was the time between clinical presentation and the time the radiograph was performed. Bony changes associated with osteomyelitis may take 7 to 15 days after the onset of the infectious process before they are identified on plain radiographs.²⁹

Patients who lacked features commonly suggestive of osteomyelitis were less likely to have a bone biopsy. This form of bias, referred to as verification bias, occurs when the decision to perform the gold standard test (bone biopsy or culture) is influenced by the results of clinical variables or screening test being studied. This will in turn inappropriately increase the apparent sensitivity of the test and decrease its apparent specificity.

While there is some uncertainty about the pretest probability of osteomyelitis in a patient with diabetes and a lower extremity ulcer, our best estimate is 15%. Using each clinical predictor individually, an ulcer area of larger than 2 cm² with an LR of 7.2 (95% CI, 1.1-49) increases the likelihood of osteomyelitis to 56%, and the positive finding of probing to bone (LR, 6.4; 95% CI, 3.6-11) increases the probability of osteomyelitis to 53%. The elevation in ESR with a positive LR of 11 (95% CI, 1.6-79) increases the probability of osteomyelitis from 15% to 66%. While most clinicians believe that the presence of these 3 variables together would make the diagnosis of osteomyelitis certain, the literature does not speak to the accuracy of combinations of variables (as noted in the aforementioned limitations). At this point, most clinicians would treat as osteomyelitis. Those who are uncomfortable with the uncertainties in these data might choose to order an MRI. A positive MRI result in combination with any one of these clinical variables increases the probability of osteomyelitis to greater than 80% (assuming independence of the LRs).

The physical examination features and laboratory and radiographic results are not helpful in this case presentation to make a diagnosis of osteomyelitis. Quiz Ref IDAs indicated herein, ulcer area larger than 2 cm², a positive probe to bone, an ESR greater than 70 mm/h, or an abnormal radiographic finding are more likely to be associated with osteomyelitis, but none of the testing modes in our review, with the exception of MRI, displayed clinically useful negative LRs for ruling out osteomyelitis. Assuming a prevalence of 15%, a negative MRI finding decreases the probability of osteomyelitis to 2.4%, thus effectively excluding the diagnosis. The elevated white blood cell counts and positive wound swab cultures are not diagnostically helpful.

Osteomyelitis of the foot causes significant morbidity in patients with diabetes, with a significant financial burden to patients and the institutions caring for these patients. Although there is no substitute for a detailed history, its utility in the diagnosis of osteomyelitis in patients with diabetes has not been well studied. The available evidence suggests that an ulcer that measures more than 2 cm² or a positive probe-to-bone finding may be helpful to establish the diagnosis. An ESR greater than 70 mm/h or positive plain radiograph findings appear to be helpful in increasing the likelihood of osteomyelitis. Magnetic resonance imaging results should be interpreted in the context of the pretest probability. Temperature, ulcer inflammation, white blood cell count, and swab culture do not appear to be helpful in establishing the diagnosis or directing therapy in patients with diabetes and a lower extremity ulcer.