FACE DROPS

Abstract

Background and Objectives

Facial palsy is a common manifestation of Lyme disease, accounting for up to 5% of acute facial palsies in endemic regions. Lyme disease–associated facial palsy (LDFP) warrants prompt antibiotic therapy while corticosteroid therapy is indicated for Bell palsy. The role of adjuvant corticosteroids in the treatment of acute LDFP is unclear. Current limitations of diagnostic laboratory tests for Lyme disease render acute differentiation of LDFP and BP challenging in many cases.

Methods

We reviewed records from 285 patients with LDFP (N = 76) and BP (N = 209) referred to a specialized facial nerve center from 2005 to 2021 to determine clinical characteristics at time of presentation to medical care. We developed and internally validated a clinical risk assessment tool (“FACE DROPS”) based on pertinent differences between signs and symptoms of LDFP and BP at presentation.

Results

The risk assessment tool distinguishes LDFP from BP using 7 clinical criteria: fever (+8 to FACE DROPS score), aches (arthralgia/myalgia; +6), cephalalgia (headache; +3), exhaustion (unusual fatigue; +4), dermatomal or radicular pattern (transverse myelitis or radiculitis; +4), otalgia or postauricular pain (−1), and stiff neck (nuchal rigidity; +3). FACE DROPS scores ≤4 predicted BP with ≥93.5% accuracy while scores of ≥7 predicted LDFP with ≥96.0% accuracy.

Discussion

A novel risk assessment tool to distinguish LDFP from BP was developed. This tool may help guide the prescribing of antimicrobials for Lyme disease in the setting of acute facial palsy pending confirmatory laboratory evidence in the absence of an erythema migrans skin lesion.

Introduction

Acute facial palsy occurs in up to 5% of patients with acute Lyme disease (LD) and is the most common manifestation of neuroborreliosis within the United States.^1-3 Bell palsy is the most common form of acute facial paralysis worldwide. Although their presentations may be similar, Lyme disease–associated facial palsy (LDFP) is distinct from Bell palsy (BP) as it may be accompanied by systemic symptoms; erythema migrans (EM); and/or additional neurologic manifestations such as meningitis, transverse myelitis, or radiculitis (Garin-Bujadoux-Bannwarth syndrome). After onset of acute flaccid facial paralysis in BP and LDFP, patients experience a gradual return of facial tone and movements within several weeks to months. Although some patients recover completely normal facial function within several weeks, evidence indicates that approximately 25% of patients with BP and 50% of patients with LDFP will have permanent and disfiguring facial sequelae in the form of postparalytic facial palsy (PPFP).^1,3-6 Patients with PPFP have varying degrees of undesirable coactivation of facial muscles (i.e., synkinesis), impaired expression, gustatory epiphora or rhinorrhea, and hypertonicity yielding facial discomfort^6-9 (Figure).

Figure. Acute Lyme Disease–Associated Facial Palsy With Normal and Aberrant Recovery.

(A) Complete right-sided flaccid facial paralysis is demonstrated in a patient 10 days after onset of LDFP (left to right: rest, brow raise, eye closure, full-effort smile, and lower lip depression). The palsy developed 2 days after initiation of antibiotic monotherapy for confirmed Lyme disease. (B) Four months later, the patient demonstrates complete recovery. (C) The same set of expressions is shown by a patient who developed LDFP 4 years before, treated with antibiotics, corticosteroids, and antivirals. The patient is noted to have narrowing of the right palpebral fissure at rest, right-sided hypomobility of the brow with midfacial and ocular synkinesis with volitional brow elevation, severe restriction of the right oral commissure and right ocular synkinesis occurring with volitional smile, and severely impaired right lower-lip movement with concurrent mentalis and platysma synkinesis with volitional lip depression. LD = Lyme disease; LDFP = Lyme disease–associated facial palsy. [Reprinted with permission from Jowett N, Gaudin RA, Banks CA, Hadlock TA. Steroid use in Lyme disease-associated facial palsy is associated with worse long-term outcomes. Laryngoscope. 2017 Jun;127(6):1451–1458. PMID: 27598389.]

Appropriate antibiotic therapy is indicated in all cases of LDFP but not in BP. Corticosteroid therapy speeds recovery and decreases the likelihood of PPFP after onset of acute Bell palsy.^10,11 There exists no prospective evidence to support adjuvant corticosteroid use in LDFP.^12-14 Some studies even suggest that adjuvant corticosteroid therapy in acute LDFP may worsen long-term facial outcomes.¹⁵ In Lyme disease–endemic areas, a management dilemma often exists at the point of care for patients presenting with acute facial palsy because LDFP can occur without the characteristic EM skin lesion¹ and before the development of a measurable antibody response to Borrelia burgdorferi¹⁶; such cases may be erroneously diagnosed as Bell palsy and treated with corticosteroids instead of antibiotic therapy.

Distinguishing between LDFP and BP in the acute setting is crucial to guide appropriate therapy. Our goal herein was to determine whether relevant differences in acute symptomatology exist between patients presenting with acute BP and LDFP and to exploit such differences to develop a clinical scoring tool to aid in their differentiation before the results of serologic testing are available. We reviewed patients with BP and LDFP referred to a facial nerve clinic at a major medical center in the Boston area over the span of 16 years to develop and validate the tool.

Methods

Enrollment and Grouping

Patients were enrolled from the Massachusetts Eye and Ear Infirmary (MEEI) Facial Paralysis Patient Data Repository (FPPDR) between October 18, 2005, and December 13, 2021. Patients were all initially seen at various primary care, acute care, and emergency care facilities throughout the United States before being referred to our tertiary care center. Medical records at time of their initial presentation for facial paralysis were reviewed. Inclusion criteria included rapid onset of unilateral flaccid facial paralysis, with full evolution of weakness over 72 hours and onset of recovery of facial tone and movement occurring within 4 months, no middle ear disease, no evidence of a tumor or infectious process affecting the facial nerve, and no evidence of an active underlying autoimmune or inflammatory disease associated with facial palsy. Patients were excluded if they presented with synchronous or metachronous bilateral facial palsy, a vesicular (herpetic) rash, a history of head and neck cancer, or neurologic disease affecting motor nerves or were subsequently diagnosed with facial paralysis secondary to a disease other than BP or LDFP.

Cases were classified as LDFP if they met the Centers for Disease Control and Prevention (CDC) 2017 criteria for confirmed LD, namely acute-onset facial palsy (FP) with concurrent EM in a LD-endemic region or acute onset of FP in addition to laboratory evidence of B. burgdorferi infection.¹⁷ Laboratory evidence of infection includes positive two-tier LD antibody testing on a blood sample or positive CSF testing for the production of intrathecal antibody to B. burgdorferi, or positive PCR testing for detection of B. burgdorferi in blood or CSF. Cases were also included in the LDFP group if the patient had been diagnosed with LDFP by an infectious diseases specialist in the setting of ambiguous laboratory results. Cases were otherwise classified as BP.

Standard Protocol Approvals, Registrations, and Patient Consents

FPPDR enrollment for involved patients occurred at the time of presentation for facial palsy. In cases where medical records were missing during data collection for this study, patients were contacted and informed consent obtained for transfer of the records. All data collection and patient contact were approved by MGB and MEEI institutional review boards.

Data Collection

Clinical data were gathered for each case using secure REDCap electronic data capture tools hosted at Mass General Brigham (MGB) and used as input variables for statistical models aimed at predicting Lyme disease.^18,19 Data were obtained through retrospective chart review of the MGB and MEEI electronic medical records (EMRs) from 2005 to 2021 and kept on a secure server run through MGB. Two data collectors first gathered data in duplicate on 20 cases, and data were reviewed to ensure agreement. The remaining cases were then divided equally between the 2 collators to complete data collection. Next, all cases were reviewed individually by a third author to confirm correct group assignments. Data were collected in February 2022 and analyzed between November 2022 and May 2024. In the course of data collection, 59 patients were randomly selected to receive phone calls to confirm the symptoms listed in the EMR. Data collectors were blinded to diagnoses coded in the repository, although diagnoses were often stated or implied in the medical documentation being reviewed.

Tool Development

A total of 285 patients met inclusion criteria and were used to develop the clinical scoring tool. A total of 12 patients were excluded based on the abovementioned criteria. Several modeling techniques were applied, including logistic regression with stepwise selection based on AIC criteria, Lasso regression, decision trees, random forest classification, and gradient boosting. The evaluation of model performance included metrics such as overall accuracy (number of correctly predicted patients divided by the total number of patients), sensitivity (true-positive rate), and specificity (true-negative rate). To ensure robust validation, a 3-fold cross-validation strategy was used for all models. This was driven by the limited occurrence of radiculo-spinal manifestations (radiculitis, transverse myelitis), which were present in only 3 patients. The final prediction model was selected based on the criteria of achieving the highest accuracy and specificity (eTable 1). In this case, the Lasso regression model demonstrated the best performance and was chosen as the final prediction model, given that it encourages sparsity in the model by setting some coefficients to zero, making it particularly useful for feature selection and model simplification when dealing with high-dimensional data sets.

R version 4.2.2 was used for execution of Lasso regression. The data set was randomly divided into 3 approximately equal-sized folds, with each fold containing patients with all types of symptoms. In each iteration, 2 folds were used for training the Lasso regression model and the remaining fold was used for validation. The regularization parameter λ was tuned during the cross-validation process to find the optimal value that balances model complexity and performance. The performance metrics were calculated for the validation fold, and the final evaluation of the Lasso regression model was obtained by averaging the performance metrics from the 3 iterations. This resulted in the retention of 7 predictors. Lasso regression coefficients were obtained for these 7 predictors. The coefficients were floating-point numbers with low values used in probability calculations ranging from −0.56 to 4.13. The coefficients were then leveraged as estimates of the weights attributed to each predictor variable in the model, by multiplying them by 2, rounding to the nearest integer to achieve a user-friendly “risk score” for each predictor. These 7 total predictors composed the FACE DROPS (fever, arthralgias or myalgias, cephalalgia, exhaustion or unusual fatigue, dermatomal or radicular pattern, nuchal rigidity, otalgia or postauricular pain, and stiff neck) clinical scoring tool. Of the 7 factors in the model, 6 were predictive of LDFP and were assigned positive values while one factor (otalgia/postauricular pain) was more predictive of Bell palsy and was assigned a negative value.

Fifty-nine of the 285 patients (14 with LDFP, 45 with BP) having the most recent dates of diagnosis were then contacted to confirm the presence or absence of the 7 predictor symptoms to improve the rigor of the scoring tool. A 3-fold cross-validation Lasso regression model was then used to determine final weights of the 7 predictors, which were then reduced to user-friendly “risk score” for each predictor as given above.

Data Availability

Data not provided in the article because of space limitations may be shared (anonymized) at the request of any qualified investigator for purposes of replicating procedures and results, subject to institutional board approvals.

Results

Descriptive Statistics

The study cohort used in calculation of the FACE DROPS score included 285 patients. The BP group included 209 patients with 125 women (60%) and 84 men (40%) while the LDFP group included 76 patients with 35 women (46%) and 41 men (54%). Age at presentation in the BP group ranged from 9 to 85 years (mean 46.2 years, standard deviation (SD) 16.1 years) while age in the LDFP group ranged from 10 to 85 years (mean 38.0 years, SD 20.4 years). Descriptive statistics are presented in Table 1. Presentation for LDFP generally clustered in summer months while presentation for BP was spread approximately evenly throughout the year (Table 2).

Table 1.

Descriptive Characteristics of Patients in the BP and LDFP Groups

Characteristic	Bell palsy (N = 209)	LDFP (N = 76)
Female, n (%)	125 (59.8)	35 (46.1)
Male, n (%)	84 (40.2)	41 (53.9)
Age at presentation (y), n (%)
Younger than 18	6 (2.9)	16 (21.0)
18–30	33 (15.8)	15 (19.7)
31–40	43 (20.4)	13 (17.1)
41–50	43 (20.4)	11 (14.3)
51–60	40 (19.1)	9 (11.8)
61–70	31 (14.8)	7 (9.2)
71–80	10 (4.8)	4 (5.2)
81–90	3 (1.4)	1 (1.3)
Race/ethnicity (self-identified), n (%)
Black	14 (6.7)	2 (2.6)
Arab	0	1 (1.3)
Asian/Pacific Islander	4 (1.9)	2 (2.6)
Caucasian	156 (74.6)	65 (85.5)
Latino/Hispanic	19 (9.0)	2 (2.6)
Other/no response	16 (7.7)	4 (5.2)
Signs/symptoms, n (%)
Radiculitis/transverse myelitis	0	3 (4.0)
Rash	4 (1.9)	29 (38.2)
Fever	1 (0.5)	36 (47.4)
Nuchal rigidity	1 (0.5)	21 (27.6)
Fatigue	9 (4.3)	31 (40.8)
Headache	37 (17.7)	55 (72.4)
Retroauricular pain or otalgia	114 (54.5)	32 (42.1)
Arthralgia	1 (0.5)	25 (32.9)
Dysgeusia	41 (19.6)	12 (15.7)

Table 2.

Distribution of BP and LDFP Cases by Month

	January	February	March	April	May	June	July	August	September	October	November	December	Unspecified date
BP	22	19	19	22	15	17	19	14	12	19	15	11	5
Lyme	0	1	2	1	7	20	20	8	7	3	4	1	2

Abbreviations: BP = Bell palsy; LDFP = Lyme disease–associated facial palsy.

FACE DROPS Score

Based on the abovementioned data, the FACE DROPS score was developed constituting the following signs and symptoms as assessed at the time of initial presentation for unilateral facial palsy (Table 3).

Table 3.

FACE DROPS Score and Description of Characteristics

Characteristic	Score	Description
Fever	+8	Temperature ≥100.4˚F
Aches	+6	Arthralgia/myalgia; joint or muscle pains
Cephalalgia	+3	Headache (e.g., frontotemporal, bitemporal, and occipital) other than isolated otalgia or postauricular pain
Exhaustion	+4	Unusual fatigue reported by the patient around time of palsy onset
Dermatomal or radicular signs	+3	Symptoms or signs consistent with myelitis or radiculitis (e.g., paresthesias, allodynia, and hyperalgesia) or diagnostic evidence consistent with myelitis or radiculitis (e.g., electrodiagnostic studies and MRI)
Otalgia or postauricular pain	−1	Pain in the ear or postauricular area ipsilateral to facial droop
Stiff neck	+3	Nuchal rigidity

Patients were categorized into prediction groups based on their FACE DROPS scores (Table 4; eTable 2). The FACE DROPS risk assessment tool accurately predicted BP with scores of 4 or lower (≥93.5% accuracy) and LDFP with scores of 7 or higher (≥96.7% accuracy). Moderate scores of 5–6 were more predictive of LDFP than BP but with less certainty. Of note, nearly one in 5 patients with LDFP had scores in the range of 4 or less.

Table 4.

Prediction Accuracy for FACE DROPS

FACE DROPS score	Prediction accuracy for BP	Prediction accuracy for LDFP
≤4	93.5%
5–6		71.4%
≥7		96.7%

Abbreviations: BP = Bell palsy; LDFP = Lyme disease–associated facial palsy.

For scores from −1 to 4, the tool is predictive of BP while for scores of 5 or more, the tool is predictive of LDFP.

The tool is most strongly predictive of LDFP with scores ≥7.

Fever and arthralgias/myalgias were highly predictive of LDFP, respectively, contributing 8 and 6 points to the FACE DROPS score. A plurality of symptoms contributed 3 points to the FACE DROPS score while one symptom, otalgia or postauricular pain, contributed a negative value to the score, being more commonly reported among patients with Bell palsy.

When using a threshold of FACE DROPS score ≥5 for prediction of LDFP, the clinical scoring tool overall has a sensitivity of 81.6%, specificity of 97.6%, positive predictive value for LDFP of 92.5%, and negative predictive value for LDFP of 93.6%.

Four patients with BP were calculated to have FACE DROPS scores in the range generally predictive of LDFP. Two had scores of 6, a patient in the sixth decade of life and another in the fifth decade of life, both with headache, fatigue, and postauricular pain. One patient in the seventh decade of life with headache and fatigue had a score of 7. One patient in the seventh decade of life with fever had a score of 8.

Discussion

We have developed the FACE DROPS clinical validation tool to aid clinicians in distinguishing between LDFP and BP in the acute care setting.

Reports suggest that most patients who present to medical care with LDFP are administered corticosteroids,^6,14,15 and that initiation of appropriate antibiotic therapy in many cases of LDFP is delayed, owing to challenges in differentiating LDFP from BP in the acute setting. A characteristic EM rash at onset of LDFP is often lacking or overlooked on physical examination, necessitating laboratory confirmation of Borrelia infection to render a diagnosis.^20,21 Serologic testing is typically used to confirm Lyme disease diagnosis in cases where EM rash is absent, but it carries several limitations: there is often a time delay in obtaining the results, and negative serologic testing results are well known to occur in approximately 10–15% of patients at the time of LDFP onset.^16,22,23 False-positive Borrelia serologies may also occur.²² Other diagnostics including Borrelia culture and nucleic acid amplification tests are not broadly available, have low sensitivity, and also require several days to report results. The FACE DROPS tool's purpose is to assist physicians in the rapid differentiation of LDFP from BP based on clinical signs and symptoms alone to guide point-of-care therapeutic decision making.

The FACE DROPS risk assessment tool accurately predicts BP in patients with low scores and LDFP in patients with high scores; patients with scores of 4 or lower are more likely to have BP (≥93.5% accuracy) while those with scores of 7 or more are highly likely to have LDFP (96.7% accuracy). Despite the overall diagnostic accuracy, caution is warranted because nearly one in 5 patients with LDFP had a FACE DROPS score of 4 or less in this study, indicating that LD testing is always warranted for patients presenting with acute facial paralysis in endemic areas during Lyme season. Empiric antimicrobial treatment of LD is also reasonable while testing is performed. Intermediate scores of 5 or 6 should also be interpreted more cautiously because they are less strongly predictive of LDFP (71.4% accuracy). For these scores, we recommend empiric treatment for both Lyme disease and Bell palsy (i.e., treatment with both antimicrobial therapy and steroids) while testing for LD. Improved discrimination in prediction accuracy between scores might be achieved with larger sample sizes, although the overlap in symptomatology of BP and LDFP suggests that confirmatory laboratory testing may remain necessary in many cases. Although patients with LDFP are more likely than those with BP to report constitutional symptoms, some of these symptoms are strongly predictive of LDFP only when multiple symptoms are present.

The FACE DROPS tool may be used to guide therapeutic decision making. For example, a clinician might decide to treat patients having FACE DROPS scores from −1 to 4 with dual therapy comprising corticosteroids and antivirals pending Lyme serology results. For patients having intermediate FACE DROPS scores of 5 or 6, a clinician might choose to treat patients with triple therapy comprising steroids, antivirals, and anti-Borrelia antibiotics pending confirmatory laboratory testing. We recommend that patients with FACE DROPS scores of 5 or more always be prescribed anti-Borrelia antibiotics pending serologic testing for Lyme disease. The evidence for adjunctive use of corticosteroids and antiviral therapy in LDFP is unclear. If corticosteroids are administered, we recommend that the duration of prescribed antibiotic therapy be significantly longer than the steroid course to ensure eradication of the spirochete in the setting of steroid-induced immunosuppression. Owing to the elevated likelihood of false negatives in LD serologic tests during the initial stages of facial palsy onset in LDFP, we advise repeat serologic testing of patients with FACE DROPS scores in the intermediate (5–6) or high (≥7) range generally predictive of LDFP whose initial LD serology is negative.

Given the large referral catchment area of the MEE Facial Nerve Center, patients in this study represented a large geographic segment of LD-endemic regions in the Northeastern United States. However, LD is caused by different species of spirochetes of the Borrelia genus: in North America, B. burgdorferi dominates while in Eurasia, B. afzelii and B. garinii are more common. Although this tool might prove useful for discrimination of LDFP from BP in Eurasia and in North America, it has long been known that LD symptomatology differs based on Borrelia species and so this tool may have lower accuracy in areas with lower prevalence of B. burgdorferi. For example, radiculitis symptoms are far more prevalent in cases of Lyme disease in Europe than in North America.²⁴ Another limitation of this study is that the patients all presented secondarily to a facial nerve center. A large proportion of secondary presentation to a facial nerve center is the result of slow recovery of facial function or development of PPFP, creating a potential bias in the data toward patients who experienced more severe cases of LDFP and BP. We also note that the FACE DROPS scoring tool currently does not include the season of facial palsy onset, which is relevant because LD is less common in colder months. Although the risk of LDFP is highest in the spring, summer, and fall, infection is nevertheless possible any time of year, even in colder northern states such as Massachusetts.²⁵ Given that most of the patients with LDFP do not recall being bitten by a tick,^15,26 and that the EM rash with facial palsy in an endemic region is considered diagnostic of LDFP, we did not include tick bite or the presence of an EM type rash as predictors in the scoring tool.

Our study included most White non-Hispanic patients, reflecting published demographic trends of LD by race and ethnicity.²⁷ Racial differences have been identified in the natural history of LD that could affect the accuracy of the FACE DROPS system. For example, EM and other rash types may be more frequently diagnosed in lighter skinned than in darker skinned patients, and longer term sequelae of LD such as Lyme arthritis are more common among darker skinned patients, possibly because of delayed diagnosis.^28-30 Future work with the FACE DROPS clinical scoring system will seek to validate the tool in prospective fashion with external validation across multiple point-of-care medical settings and across diverse geography.

The FACE DROPS score is a novel risk assessment tool that helps discriminate BP from LDFP at the point of care. This tool may inform prescribing of antibiotics and corticosteroids in patients presenting with acute-onset unilateral facial palsy in LD-endemic regions.

TAKE-HOME POINTS

• → Lyme disease–associated facial palsy (LDFP) and Bell palsy (BP) often overlap regarding clinical presentation. However, LDFP more often presents with fever, arthralgia, fatigue, headache, dermatomal or radicular signs, and nuchal rigidity. Bell palsy more often presents with postauricular pain.

• → A FACE DROPS score of 4 or lower is predictive of BP. However, 1 in 5 patients with a FACE DROPS score in this range may have LDFP. A score of 5–6 is weakly predictive of LDFP while a score of 7 or higher is strongly predictive of LDFP.

• → LD testing is always warranted for patients presenting with acute facial paralysis in endemic areas during Lyme season regardless of the FACE DROPS score.

• → We recommend treatment of patients with a FACE DROPS score of ≥7 with empiric Lyme antimicrobials only, without use of steroids.

• → It is reasonable to treat patients with a score of ≤6 empirically for Lyme disease until serologic testing has returned negative. This decision should be made based on local epidemiology and seasonality.

Author Contributions

C.R.S. McEntire: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data. S.Y. Chung: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. B. Chang: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. K.J. Barrera: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; analysis or interpretation of data. Y. Zhao: analysis or interpretation of data. J.W. Joseph: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. G.P. Wormser: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. N. Jowett: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data. B.K. Chwalisz: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data.

Study Funding

The authors report no targeted funding.

Disclosure

C.R.S. McEntire, S.Y. Chung, B. Chang, K. Barrera, and Yan Zhao report no relevant disclosures. G. Wormser reports receiving a research grant from Biopeptides, Corp. He has been an expert witness in malpractice cases involving Lyme disease and is an unpaid board member of the non-profit American Lyme Disease Foundation. N. Jowett owns Moderna stock and reports no other relevant disclosures. B.K. Chwalisz reports no relevant disclosures. Full disclosure form information provided by the authors is available with the full text of this article at Neurology.org/cp.