Alteration of taste or smell as a predictor of COVID-19.

We read the studies by Jin et al 1 and Lin et al 2 regarding the GI manifestations in COVID-19 with great interest. Recently, alteration in taste (dysgeusia, hypogeusia, ageusia) or smell (dysosmia, hyposmia, anosmia) has been increasingly reported in Western patients with COVID-19.3–10 However, these symptoms were not reported from these two studies from China.1 2 Therefore, we conducted a systematic review and meta-analysis to assess the prevalence of taste or smell alteration in patients with COVID-19 and to calculate the accuracy of these symptoms in the prediction of COVID-19. We searched articles in the PubMed database from 1 January 2020 through 10 May 2020 using the keywords “COVID-19” and “taste” or “smell” without any limitations in the search fields (registration number: CRD42020187912).

Of the 48 articles identified using the above search strategy, 30 were excluded (9 irrelevant, 9 comments, 5 review articles, 5 case reports, 2 series of cases with taste/smell alteration) and 18 were eligible for meta-analysis (study methods and references can be found in the online supplementary materials).1–5 Six of the 18 studies also included control groups. Overall, the frequency of taste alteration, smell alteration, and taste or smell alteration was reported in 41.3% (95% CI 21.1% to 65%), 45.3% (95% CI 23.3% to 69.4%), and 46.6% (95% CI 27% to 67.3%) in COVID-19 confirmed cases (table 1). The heterogeneity was high, but there was no significant publication bias (online supplementary materials). The frequency of taste or smell alteration was lower in studies from China (0.9%, 95% CI 0.2% to 3.7%) and Korea (12.2%, 95% CI 11.1% to 13.4%), compared with those from USA and Europe (41.3%, 95% CI 21.1% to 65%; online supplementary materials). The frequency of taste or smell alteration was higher in studies that recruited outpatients (63.9%, 95% CI 52.1% to 74.3%) than those that only recruited hospitalised patients (35.8%, 95% CI 29.6% to 42.4%; online supplementary materials).

Table 1

Characteristics of eligible studies and prevalence of taste or smell alteration in COVID-19 cases and controls

Author	Country	Setting	Mean age (years)	Male (%)	Taste alteration		Smell alteration		Taste or smell alteration
					Case	Control	Case	Control	Case	Control
Lechien	Europe	H&O	37	36.9	–	–	86% (357/417)	–	–	–
Roland	USA	O	39	28.0	–	–		–	66% (95/145)	27% (42/157)
Yan-1	USA	O	–	49.2	71% (42/59)	17% (35/203)	68% (40/59)	16% (33/203)	–	–
Beltrán-Corbellini	Spain	H	53	–	35% (28/79)	10% (4/40)	32% (25/79)	10% (4/40)	39% (31/79)	13% (5/40)
Mao	China	H	53	40.7	1% (2/214)	–	1% (2/214)	–	–	–
Giacomelli	Italy	H	60	67.8	29% (17/59)	–	24% (14/59)	–	34% (20/59)	–
Lee	Korea	H&O	44	36.4	12% (389/9191)	–	11% (353/3191)	–	15% (488/3191)	–
Aggarwal	USA	H	66	75.0	19% (3/16)	–	19% (3/16)	–	–	–
Moein	Iran	H	47	67.0	–	–	98% (59/60)	18% (11/60)	35% (21/60)	0% (0/60)
Luers	Germany	O	36/41	56.9	68% (50/73)	–	73% (53/73)	–	–	–
Paderno	Italy	H&O	55	56.0	63% (320/508)	–	56% (284/508)	–	–	–
Vaira	Italy	O	47	33.3	52% (17/33)	–	52% (17/33)	–	–	–
Yan-2	USA	O	53/43	47.7	58% (70/120)	–	59% (75/128)	–	76% (128/169)	–
De Maria	Italy	O	–	–	51% (48/95)	–	–	–	–	–
Merza	Iraq	H	28	60.0	–	–	–	–	27% (4/15)	–
Spinato	Italy	O	56	48.0	–	–	–	–	64% (130/202)	–
Clemency	USA	O	–	–	–	–	–	–	49% (110/225)	15% (108/736)
Menni	UK and USA	H&O	44	25.1	–	–	–	–	65% (4668/7178)	22% (2436/11 223)
Meta-analysis	–	–	–	–	41% (21%–65%)		45% (23%–69%)		47% (27%–67%)

References are shown in the online supplementary materials.

Variables in bold are the results of meta-analysis.

H, hospitalised; H&O, hospitalised and outpatient; O, outpatient.

The presence of taste or smell alteration was associated with increased risk (OR 6.5, 95% CI 5.2 to 8.1) of COVID-19 infection. The overall sensitivity, specificity, positive prediction value, negative predictive value, and accuracy of taste or smell alteration in the prediction of COVID-19 was 54.8% (95% CI 44.3% to 64.9%), 81.7% (95% CI 76.5% to 85.9%), 65.2% (95% CI 55.9% to 73.4%), 73.9% (95% CI 64.3% to 81.6%), and 71.8% (95% CI 67.5% to 75.6%), respectively (table 2).

Table 2

Sensitivity, specificity and accuracy of taste or smell alteration in the prediction of COVID-19

Author	Taste/smell (case)	Taste/smell (control)	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)	Accuracy (%)
Roland	66% (95/145)	27% (42/157)	65 (57.4–72.8)	73.2 (65.8–79.6)	69.3 (61.1–76.5)	69.7 (62.3–76.4)	69.5 (64.1–74.5)
Yan-1	71%* (42/59)	17%* (35/203)	71 (58–81.3)	82.8 (76.9–87.4)	54.5 (43.4–65.3)	90.8 (85.7–94.2)	80.2 (74.9–84.5)
Beltrán-Corbellini	39% (31/79)	13% (5/40)	39 (29.1–50.4)	87.5 (73.3–94.7)	86.1 (70.7–94.1)	42.2 (32–53)	55.5 (46.5–64.1)
Moein	35% (21/60)	0% (0/60)	35 (24.1–47.8)	99.2 (88.2–99.9)	97.7 (72.3–99.9)	60.6 (50.7–69.7)	67.5 (58.6–75.3)
Clemency	49% (110/225)	15% (108/736)	48.9 (42.4–55.4)	85.3 (88.2–87.7)	50.5 (43.9–57.0)	84.5 (81.7–86.9)	76.8 (74–79.4)
Menni	65% (4668/7178)	22% (2436/11223)	65 (63.9–66.1)	78.3 (77.5–79)	65.7 (64.6–66.8)	77.8 (77–78.5)	73.1 (72.5–73.8)
Meta-analysis	–	–	54.8 (44.3–64.9)	81.7 (76.5–85.9)	65.2 (55.9–73.4)	73.9 (64.3–81.6)	71.8 (67.5–75.7)

References are shown in the online supplementary materials.

Variables in bold are the results of meta-analysis.

*Using data on taste alteration.

NPV, negative predictive value; PPV, positive predictive value.

This is the first systematic review and meta-analysis reporting the sensitivity and specificity of taste or smell alteration in the prediction of COVID-19. Our results showed that alteration of taste or smell is a common manifestation of COVID-19 in Western populations. However, these symptoms are rarely reported in studies from East Asia. Thus, the sensitivity of taste or smell alteration in the prediction of COVID-19 varies according to population and geography. Nevertheless, the specificity is high (81.7%) and the presence of taste or smell alteration is predictive of COVID-19 (positive prediction value 65.2%) and a confirmatory test is required in these subjects.

There are some limitations to this study. First, taste or smell alteration was not included in many reports from China or other East Asian countries. The relatively lower frequency of these symptoms in East Asia may be attributed to host genetic factors, variations of coronavirus or differential reporting. Second, only 6 of the 18 studies included COVID-19 negative control group and none from East Asia. Third, relatively few studies reported the clinical courses of taste or smell alteration. Of the studies that reported the clinical courses, complete resolution of taste or smell alteration was reported in 50%–70% of patients with COVID-19 within 2 weeks of symptom onset. More studies from other populations which include a COVID-19 control group and that address on the treatment, clinical course and pathogenesis of taste or smell alteration in COVID-19 are warranted.