Gram Staining: a Comparison of Two Automated Systems and Manual Staining

Various Gram staining automated systems are available to accelerate and standardize the staining process, but a systematic comparison of different systems is largely lacking. The objective of this study was to evaluate two devices in comparison to manual Gram staining. Clinical samples (n = 500; University Hospital Münster, Germany; May to June 2020) were simultaneously Gram stained manually and with two automated Gram stainers (Previ Color Gram, bioMérieux, and ColorAX2, Axonlab).

ABSTRACT

Various Gram staining automated systems are available to accelerate and standardize the staining process, but a systematic comparison of different systems is largely lacking. The objective of this study was to evaluate two devices in comparison to manual Gram staining. Clinical samples (n = 500; University Hospital Münster, Germany; May to June 2020) were simultaneously Gram stained manually and with two automated Gram stainers (Previ Color Gram, bioMérieux, and ColorAX2, Axonlab). The quality was assessed based on four criteria: (i) homogeneous staining of bacteria/fungi, (ii) uniform staining of the background, (iii) absence of staining artifacts, and (iv) congruency between culture and microscopy. Each criterion was rated with 0 (absence) or 1 (presence) point to calculate a quality score (0 to 4 points). The costs for each staining procedure were calculated based on consumables and hands-on time (applying the average wage of a laboratory technician in the public service for Germany and the United States). The mean (± standard deviation [SD]) quality scores were comparable for manual staining (3.06 ± 0.91) and Previ Color Gram (3.04 ± 0.90; P = 0.6), while significantly lower scores were achieved by ColorAX2 (2.57 ± 1.09; P < 0.0001). The total cost per Gram stain was €1.13/$1.34 for Previ Color Gram, €0.80/$0.83 for manual, and €0.60/$0.71 for ColorAX2, respectively. The quality and costs per slide vary significantly between instruments of different manufacturers.

INTRODUCTION

Gram staining is still a key staining protocol in bacteriology, guiding both the diagnostic processes and therapeutic management of bacterial infections. Despite its use over a century, Gram staining is still error prone, as slight variations in the (pre-/post-) analytical process (e.g., prior antibiotic therapy, specimen collection and fixation, decolorization, and subjectivity of stain interpretation) can affect the staining results (1–3).

The quality of manually stained slides depends on the skills of the laboratory technician and can vary over time and between individual persons. In contrast, automated staining systems use standardized protocols, which increase the comparability and consistency of results and facilitate the staining process for those who are less trained or nonmicrobiologists (4). The aim of this study was to compare the staining quality of two automated staining systems (Previ Color Gram, bioMérieux, Marcy l’Étoile, France, and ColorAX2, Axonlab, Baden, Switzerland) and standard manual staining.

MATERIALS AND METHODS

Ethics.

Since we only used anonymized laboratory data that are accessible through activities of routine diagnostics, we did not seek ethical clearance from our institutional review board. In accordance with local authorities, no informed consent from participants is needed (§6 des Gesundheitsdatenschutzgesetzes NRW [GDSG NRW]).

Microbiological samples.

Our laboratory successfully participates in regular proficiency tests addressing (among others) Gram staining, culture, and identification. A number of 500 consecutive convenience samples based on availability in the routine laboratory was deemed appropriate for the study objectives; a specific sample size calculation was not done. All samples (i.e., swabs, blood cultures, cerebrospinal fluid [CSF], and tissue samples) with a request for Gram staining by the clinicians were included in the study; no exclusion criteria were applied. The decision to send samples for microbiological analysis was solely based on the physician’s judgment.

For all liquid samples (e.g., blood cultures, CSF, and respiratory lavages), 15 to 20 μl were evenly spread on microscopy slides, covering an area of 2 × 3 cm². Liquid samples were not concentrated by the use of cytospin. Depending on the specimen, we applied a standard set of solid and liquid culture media (Table S1 in the supplemental material). A volume of 10 μl of a liquid sample was directly inoculated on solid culture media. Inoculated aerobic/anaerobic blood cultures (Bactec; BD, Heidelberg, Germany) were incubated in Bactec 9240 (BD).

Gram staining.

After drying at room temperature, all samples were heat fixated (1 s on an open flame, repeated 3 times) and simultaneously Gram stained by hand and with two automated systems (Previ Color Gram, bioMérieux, and ColorAX2, Axonlab) using standardized protocols of the manufacturers (Table 1). Previ Color Gram sprays the reagents on the slides in a centrifuge system with a capacity of 12 or 30 slides. ColorAX2 uses 10 individual staining chambers for single slides. These chambers are filled with the staining reagent and automatically drained after staining. All consumables were purchased and used according to the manufacturer’s instructions. Staphylococcus aureus (ATCC 29213) and Escherichia coli (ATCC 25922) were used as positive controls, and blank slides were used as negative controls.

TABLE 1.

Protocols for manual and automated Gram staining

Characteristic or type of stain	Time for staining (in seconds) by system
	Manual	Previ Color Grama	ColorAX2
			Clinical specimens	Pure cultures (controls)a
Fixation type	Heat	Heat	Heat	Heat
Crystal violet	15	50	45	45
Iodine	30	44	60	60
Ethanol	Until no more color is removed	48b	20	10
Carbol fuchsin	15	48b	60	60

^aPure cultures (i.e., positive and negative controls) and clinical specimens were stained with slight modifications as indicated. For Previ Color Gram, pure cultures and clinical specimens were stained according to an identical protocol (as indicated).

^bDecolorizer (ethanol) und stain (carbol fuchsin) are contained in one dye solution.

Definition of quality criteria.

Due to the lack of standardized criteria, we defined four in-house quality criteria for the assessment of stained slides: (i) homogeneous staining of bacteria/fungi, (ii) uniform staining of the background (5), (iii) no staining artifacts (e.g., precipitates of crystal violet [4]), and (iv) congruency between culture and microscopy (1) (Fig. 1). Each criterion was rated with 0 (absence) or 1 (presence) point to calculate a quality score with an achievable maximum score of 4. For negative samples (no bacteria or fungi detected by microscopy), the maximum achievable score was 3 by default, since the criterion “homogeneous staining of bacteria/fungi” could not be evaluated. Discrepant results between culture and microscopy were further categorized as a major error (smear negative/culture positive) or a minor error (smear positive/culture negative).

FIG 1.

Quality criteria for the evaluation of Gram stains. Pictures show positive and negative examples for uniform staining of the background, artifacts, and homogeneous staining of bacteria/fungi.

Microscopy.

Trained physicians examined ≥10 fields per slide with a light microscope (Axioscope; Zeiss, Cologne, Germany) at ×1,000 magnification under oil immersion. The staining quality was systematically assessed based on the four quality criteria. One physician graded each slide.

Cost calculation.

The total costs for Gram staining per slide were calculated based on consumables, maintenance, and workforce (excluding procurement costs for the automated systems). Consumables and staining reagents differed according to the technique (Table S2). The consumption of reagents was monitored throughout the study to calculate the volume and costs of each reagent per Gram stain. Costs of slides were not included, as they were identical for each method. Workforce costs were calculated based on the sum of hands-on time for each method. The salary scale of nonacademic employees in the public service (Germany, €4.250/month; United States average, $4.360/month [https://money.usnews.com/careers/best-jobs/clinical-laboratory-technician/salary]) was used to determine the workforce.

Statistical analysis.

Categorical variables (i.e., quality criteria) were compared with the chi-square test. Continuous variables (i.e., quality score) were first tested for normal distribution (Shapiro-Wilk test). Since the data were not normally distributed, we used the Wilcoxon rank sum test with continuity correction for the comparison of scores as implemented in R (version 3.6.1).

RESULTS

In total, 500 samples were included in this study, with blood cultures being the most common samples (269/500; 53.8%), followed by respiratory specimens (84/500; 16.8%) and others (e.g., CSF, tissue samples, and swabs [147/500; 29.4]). Cultures grew (multiple detection of pathogens was possible) Gram-positive cocci (218/500; 34.6%), Gram-negative rods (139/500; 27.8%), yeasts (60/500; 12.0%), Gram-positive rods (24/500; 4.8%), and Gram-negative cocci (6/500; 1.2%) (no growth, 146/500; 29.2%).

The quality of Gram staining was comparable between manual staining and Previ Color Gram; both methods achieved similar quality scores, with a slightly better performance of manual staining (3.06 versus 3.04; P = 0.6; Table 2). Samples that were Gram stained by ColorAX2 reached lower-quality scores than Previ Color Gram, mainly due to discordant results between culture and Gram staining and staining artifacts (Table 2). Staining artifacts were not associated with incorrect Gram staining using ColorAX2 (odds ratio [OR], 0.83; 95% confidence interval [CI], 0.6 to 1.2; P = 0.4).

TABLE 2.

Comparison of automated versus manual Gram staining

Characteristic	Staining methode			OR (95% CI)c	P valuec
	Manual	Previ Color Gram	Color AX2
Even staining of bacterial/fungal cells (no. of specimens [%])	272 (96.5)a	249 (89.2)b	223 (79.9)b	1.9 (0.8–4.4)	0.14
Even staining of the background (no. of specimens [%])	476 (95.2)	482 (96.4)	411 (82.2)	8.1 (3.1–21.7)	<0.0001
Staining artifacts (no. of specimens [%])	85 (17.0)	78 (15.6)	175 (35)	0.2 (0.1–0.3)	<0.0001
Concordance of Gram stain and culture (no. of specimens [%])	367 (73.4)	368 (73.6)	325 (65.0)	72.5 (35.7–147.1)	<0.0001
Minor errors (no. of specimens [%])	28 (21.1)	35 (26.3)	72 (41.1)	NAd	NAd
Major errors (no. of specimens [%])	120 (90.2)	119 (89.5)	157 (89.7)	NAd	NAd
Quality score (mean [SD])	3.06 (0.91)	3.04 (0.90)	2.57 (1.09)	NA	<0.0001
Hands-on-time of laboratory technician per Gram stain (s)	138	25	7	NA	NA
Cost of consumables and staining reagents per Gram stain	€0.08/$0.09	€0.90/$1.07	€0.56/$0.67	NA	NA
Personnel cost per Gram stain	€0.72/$0.74	€0.13/$0.14	€0.04$/0.04	NA	NA
Hardware maintenance cost per Gram stain	NA	€0.10/$0.13	NA	NA	NA
Total cost per Gram stain	€0.80/$0.83	€1.13/$1.34	€0.60/$0.71	NA	NA

^aBacterial/fungal cells were only seen in 282 samples.

^bBacterial/fungal cells were only seen in 279 samples.

^cUnivariate comparison of Previ Color Gram and ColorAX2.

^dNot applicable, as only a subset of samples (122/500) have data points for both staining automated systems.

^eFive hundred specimens were used for all staining methods.

Blood cultures and respiratory specimens were the two major specimens in our collection. Since both differ in general composition (e.g., presence/absence of cell detritus, tissue, epithelial cells, lymphocytes, secretion, and extracellular matrix), we tested if the two automated systems perform differently for the two sample types. Comparable quality scores for manual staining and Previ Color Gram and lower-quality scores for ColorAX2 were observed if only blood cultures or respiratory specimens were analyzed separately (Table S2 and Table S3).

We compared the overall costs per Gram stain, including material and workforce costs (manual staining, €0.80/$0.83; Previ Color Gram, €1.13/$1.34; ColorAX2, €0.60/$0.71). The costs for consumables were less in the manual staining technique than the automated systems. The two latter required less hands-on time of the laboratory technicians, resulting in lower workforce costs (Table 2). In contrast to Previ Color Gram, ColorAX2 and the manual staining method do not require working time for hardware maintenance. The hands-on time needed to change reagent containers or to rinse tubes of the automated systems is not included in the calculation, as the procedures are equal in all three methods.

DISCUSSION

We evaluated the quality of Gram stains of two automated instruments and compared the results with those of conventional manual staining. The main finding is a significant difference in Gram staining performance between the two automated systems.

Artifacts and discordant results between Gram staining and culture were the main causes of a lower staining quality of ColorAX2 (Table 2). The reason for the high proportion of artifacts is unclear, as they remained after the replacement of the device. Discordant results of ColorAX2 were mainly due to erroneous staining (e.g., Gram-positive rods in the smear and Gram-negative rods in culture) as illustrated by almost identical proportions of major errors (smear negative/culture positive) and minor errors (smear positive/culture negative), particularly in blood cultures (Table 2; Table S2 in the supplemental material). Discordance between Gram stain and culture does not only depend on the staining technique but also on the specimen (38% in respiratory materials, 0.7% in blood cultures) (1, 6). However, as the same set of samples was stained by all methods, we adjusted for this bias.

In the absence of standardized assessment schemes (7) for the evaluation of Gram stains, we developed our own in-house quality criteria. This straightforward approach has some caveats; some genera (e.g., Bacillus and Clostridioides) stain Gram variable and might therefore not always achieve the criteria “homogeneous staining of bacteria/fungi” (Fig. 1) (6, 8). However, this holds true for aged cultures, but we assessed only uncultured, direct smears from clinical specimens (8). In addition, our scheme does not allow for grading each criterion. For instance, a mixed culture of Gram-positive cocci, yeast, and Gram-negative rods might not be fully detected by the two staining approaches (e.g., approach 1 detects only Gram-positive cocci; approach 2 detects Gram-positive cocci and yeast but not Gram-negative rods). In our quality score, both would achieve no points for the criterion “congruency between culture and microscopy,” although approach 2 performed better than approach 1.

Although our study provided important information on the strengths and weaknesses of different automated staining systems, some limitations need to be addressed. First, we were unable to blind the observers of Gram stains since characteristics of the different staining procedures (e.g., impression marks of the staining chambers [ColorAX2] or staining of labels [Previ Gram Color]) were inevitably visible on the microscopy slides. Therefore, we did not control for an analysis bias.

Second, we observed an improvement of the overall quality of manual Gram staining during the study over routine diagnostics, pointing toward a performance bias through laboratory technicians. Therefore, we might overestimate the quality of manual Gram staining in our observation.

In conclusion, the quality and costs per slide vary significantly between instruments of different manufacturers.