Highlights
- •
Provides the first objective analysis of confocal assessments of conjunctival goblet cells.
- •
Analyses reveal notable differences between studies that need to be addressed.
- •
Differences between studies appear to be uniquely dependent on the strategies used for image selection.
Abstract
Background
In vivo confocal microscopy (IVCM) has been used for over 10 years to assess the goblet cell density (GCD) within the human conjunctiva, but the reported values have been variable with no obvious indications as to why.
Methods
From publications between 2008 and 2019, representative GCD values were extracted, as well as on the image sampling strategy used.
Results
Average GCD values for any particular group of individuals ranged from 7 to 979 goblet cells / sq. mm, and with one notable outlier removed, an overall group-mean value for GCD (+/− SD) from single site locations was 207 +/− 143 goblet cells / sq. mm from 15 data sets for those usually designated as control subjects, with a value of 190 +/− 161 goblet cells / sq. mm calculated from 20 single site data sets from other (patient) groups. An overall analysis indicated that the reported average values for GCD from different groups of individuals increased according to the number of images assessed / individual (Spearman rho = 0.304), on the number of individuals evaluated to generate an averaged value for each group (rho = 0.367), and the total number of images assessed (rho = 0.346, multivariate analysis partial r = greater or = to 0.522).
Conclusions
In the use of confocal microscopy to assess the number of goblet cells in the human bulbar conjunctiva, the substantial differences reported appear to be linked to the protocols used for image selection, and some type of standardization needs to be developed.
1
Introduction
The human bulbar conjunctiva in healthy eyes is a non-stratified secretory epithelium that includes abundant numbers of goblet cells [ ]. Goblet cells within the bulbar conjunctiva are therefore generally considered to be the major source of the mucus (mucin)-containing layer of the pre-corneal tear film, a layer that contributes to the stability of the tear film [ ]. In histological examination by light microscopy of tissue sections, some of the goblet cells can be seen to be in contact with the conjunctival surface [ ], and the actual openings at their apical ends onto the conjunctival surface can be viewed by scanning electron microscopy-based imaging of excised conjunctival specimens [ ].
Conjunctival impression cytology (CIC) is an alternative method to conventional histology in that it only requires that an absorbent filter is briefly placed in contact with the conjunctival surface so as to sample the goblet cells and their mucin secretion. CIC has been used for over 40 years with one outcome measure being the so-called goblet cell density (or GCD) in goblet cells / mm 2 . Using CIC, a review of the conclusions of various authors is that the GCD is reduced in contact lens wear [ ], although the extent of this presumed compromise of the conjunctiva in relation to the type and / or duration of contact lens wear is yet to be well defined [ , ]. Part of the reason for this uncertainty is because rather different methods have been used to assess the CIC filters, including whether attempts are made to generate GCD values or just make a more subjective (relative) assessment [ ]. Furthermore, in considering the use of CIC on individuals with nominally (normal) healthy eyes, the reported values for GCD have varied widely between different studies with a review of these finding that the averaged values ranged from 24 to 2226 goblet cells /mm 2 for the bulbar conjunctiva [ ]. With such apparent variability, the utility of CIC, as diagnostic tool to assess the goblet cells of the conjunctiva, is currently limited and a case can therefore be made that other methods need to be considered.
in vivo confocal microscopy (IVCM) provides a clinically-applicable method of examining the internal structure of the conjunctiva in the living eye [ ], having first been applied to the cornea with its very distinctive layered organization [ ]. A number of studies have now been published that indicate that conjunctival goblet cells can be identified in confocal images and counts made of these so as to generate estimates of conjunctival goblet cell density. However, an initial scrutiny of the published GCD data obtained with IVCM indicated a wide range of values. It was therefore considered useful to undertake a systematic analysis of the human GCD data obtained from use of IVCM that has been reported in peer-reviewed published articles. The main goals were to be able to obtain an overall estimate of GCD (as derived from IVCM use in humans) and to assess what differences might have been found between individuals with a nominally healthy conjunctiva as compared to various groups designated with different types of ocular surface compromise or disease (i.e. groups of individuals who were not considered to have a nominally healthy conjunctiva). With the general finding that the inter-study variability in the average GCD estimates was notable, an additional goal was to assess and summarise the image analysis selection protocols used and the methods used for calculating averaged GCD values.
2
Methods
For this literature review, PubMed ( www.ncbi.nlm.nih.gov/pubmed ) was searched to identify original published articles (excluding review articles) providing conjunctival goblet cell density (GCD) values for human individuals on whom in vivo confocal microscopy (IVCM) had been undertaken. This was done, at a first step, using the combined search terms of ‘conjunctiva’, confocal microscopy’ and ‘goblet cells’. Related papers lists were then also checked. All abstracts (regardless of language or year of publication) were then examined to identify suitable studies on humans. Copies of the published articles were then obtained and details extracted on the year of publication, the number of individuals and / or eyes assessed for any particular set of data, the average age of the relevant group of participants and their overall clinical characteristics (e.g. were the individuals designated as control groups with nominally healthy eyes, or otherwise; see later). The main outcome sought after was an average value for conjunctival GCD for any particular group of individuals.
For all the published papers, the methods section was carefully scrutinized so as to obtain the best estimate of the process of image procurement, the basis upon which images were selected for objective analyses of goblet cells (i.e. counting undertaken) and the numbers of such images that were selected / eye. As part of this, attention was given to whether the values reported were from a single cardinal location (i.e. nasal, temporal, inferior or superior sites) or from combining the data from all four sites (which will be referred to as NTIS).
The data obtained from the published articles was entered into Systat (v. 11, Systat, Evanston, IL). The overall values for GCD and the distribution of the reported values were assessed by appropriate statistical tests with the normality assessed using the default Shapiro-Wilk method as incorporated into Systat) and graphical presentations generated. The coefficient of variation (COV = SD / average * 100).was used to represent the variability of the group mean value generated from pooled data sets. The level of statistical significance was set at p < 0.05 for non-normal distribution. Some comparisons between data sets were made with the Kruskal-Wallis ANOVA method (as incorporated in Systat) and some correlation analyses were undertaken using the Spearman rank order non-parametric method. The level of statistical significance for both analyses was set at p < 0.05. Multivariate correlations with MANOVA, as incorporated in Systat, were also undertaken to systematically assess the combined of the number of individuals assessed, the number of images assessed and their combined effects.
3
Results
3.1
Confocal imaging of conjunctival goblet cells
From the published images included in most of the reports, along with accompanying text or figure captions, the general consensus appears to be that conjunctival goblet cells are evident in the black-and-white (grey scale) output images as round-to-slightly oval lighter grey features against a darker grey background. No specific morphometry data on size of these features was included in the papers examined, but several reports appear to re-state a comment made in an early review [ ] that the goblet cells are about 30 μm in diameter. A later report modifies this with a statement that the goblet cells are expected to be 25–30 μm in diameter [ ]. Another way of describing the goblet cells visualized in confocal microscopy is that different investigators have considered them to be uniformly hyper-reflective (i.e. a lighter grey), although this has not always been agreed upon [ ].
3.2
Overview of GCD data sets obtained from published peer-reviewed articles
Within the different 18 published articles selected, the text and / or summary tables provided what were presented as average goblet cell density (GCD) values with these being taken from different groups who were usually specifically identified and designated as ‘control subjects’ and sometimes referred to as ‘normal’ subjects, as compared to other (non-control) groups (with some type of ocular surface compromise or disease. For the first group, it can be implied that these individuals had what many clinicians would presumably consider to be a ‘normal’ healthy appearance to the exposed portions of their bulbar conjunctiva, while in the other group these could have changes in the appearance of the conjunctiva that may or may not have been that evident. For some groups, IVCM images were obtained on a single occasion, while other groups were evaluated more than once either as part of longitudinal assessments of what were usually referred to as ‘control’ eyes or eyes with ocular surface disease with or without some type of intervention (e.g. with eye drops).
In attempting to obtain some type of reasonable estimate of GCD, as well as to assess what differences might have been found between individuals with a nominally healthy conjunctiva as compared to various groups designated with different types of ocular surface changes, the broad strategy adopted was to mainly select what were considered representative average GCD values and minimizing the potential impact of duplicate data sets. Thus was done rather than simply extracting every reported value for GCD. So, for example, in longitudinal studies, initial (baseline) data was selected and then just one set of data from a later (usually the last) time point. Similarly, if the interventions were all very similar (e.g. only slightly different eye drops) and being compared to a single ‘control’ group, then only one data set was selected. Stated another way, not all sub-groups of data were extracted for use. The average GCD values obtained from the published articles and their sources are summarized in Table 1 .
| SITE | REPORTED GOBLET CELL DENSITY (GCD) VALUES |
|---|---|
| FROM CONTROL (NORMAL) SUBJECT GROUPS | |
| NASAL | 262 [ ] 475 [ ] 482 [ ] |
| TEMPORAL | 36 [ ] 49 [ ] 979 [ ]* |
| INFERIOR | 124 [ ] |
| SUPERIOR | 7 [ ] 39 [ ] 205 [ ] 219 [ ] 237 [ ] 239 [ ] 241 [ ] 241 [ ] 252 [ ] |
| NTIS | 86 [ ] 111 [ ] 111 [ ] 406 [ ] 429 [ ] |
| FROM OTHER SUBJECT (PATIENT) GROUPS | |
| NASAL | 60 [ ], 127 [ ], 335 [ ] 408 [ ] |
| TEMPORAL | 25 [ ] 278 [ ]* 326 [ ] 639 [ ] |
| INFERIOR | 36 [ ] 53 [ ] |
| SUPERIOR | 36 [ ] 44 [ ] 57 [ ] 118 [ ] 119 [ ] 135 [ ] 166 [ ] 283 [ ] 332 [ ] |
For the overall analyses, a total of 44 GCD data sets were selected, sometimes with 2 or more data sets from a single published article, with 21 of these being considered to be from nominally healthy eyes from ‘normal’ or ‘control subjects’ [ , ] and 23 from other groups [ , , , ]. These other groups were as follows: contact lens wearers [ , ], glaucoma patients being treated with eye drops [ , , , ], patients with dry eye [ , ], patients with a history of chemical burns to the eye [ ] or patients assessed after trabeculectomy operations [ , ]. There was one notable exception to this broad terminology in that ‘control subjects’ was used to identify a group subjected to medical rather than surgical interventions [ ].
As an overall summary, the data sets in Table 1 were obtained from articles published between 2008 and 2019 of groups of individuals with average ages of between 26–73 y (overall mean value of 47 ± 15 years / data set). Analyses did not indicate any notable effects of age, whether all the data sets were analysed together or just those from nominally healthy conjunctiva were considered, for example (see later).
As is hopefully very evident from Table 1 , there has been diversity in the location of the confocal imaging, with all four cardinal sites being assessed, although none of the reports on nominally healthy conjunctiva provide any notable reasons for the selection of the imaging site. There is a dominance of data sets from superior (S) bulbar conjunctiva (mainly because the same group of researchers have published several different assessments on different groups of individuals) but, just as importantly, others have considered that obtaining images from all four cardinal locations (NTIS) and then providing some type of averaged value from the data was the appropriate method of evaluation of the goblet cells. Eight of the data sets selected were obtained from combining assessments from all four cardinal sites.
From the summary in Table 1 it can be seen that a wide range of average GCD values have been reported for different groups of individuals. From single locations, the reported GCD values could be identical (or almost identical) despite being on different groups of individuals. However, and much more importantly, it should be noted that even from a single locations, some of the average GCD values have differed by a factor of 5 or even 10-fold. When considering all data sets from single or multiple locations, an overall group mean GCD (± SD) of 211 ± 198 goblet cells /mm 2 can be calculated, with the reported average values for any particular group of subjects being between 7 and 979 goblet cells / mm 2 . If only the data sets from single site locations are considered, then the group mean GCD was 219 ± 198 goblet cells / mm 2 . By way of general comparison, five of the published articles provided the GCD data in the form of an overall average value from images taken from all four cardinal sites [ , , ]. Overall, the group mean GCD reported across 8 sets of data for these multiple sites was 179 ± 149 goblet cells / mm 2 , a value that was not substantially different from that which can be calculated from single sites.
The overall distribution of these average GCD values, regardless of site location of the imaging, is shown in Fig. 1 A. This distribution is not normal having a coefficient of skewness value of 1.853, and failed a normal distribution test (p < 0.05). The reason for the high skewness and failed normality test is, in part, due the very high average value of 979 goblet cells/ mm 2 (which was actually calculated from a different sized microscope field of view (see later). That this very high value is clearly an outlier is perhaps better illustrated with a box plot analysis ( Fig. 1 B) which compares the sets of averaged GCD values from those control subject groups considered to have a healthy conjunctiva (CS) with data from all other (O) groups. As shown as an asterisk in the CS group, this data set is clearly an outlier.
