Progression from Active to Scarring Trachoma

We identified four prospective cohort studies that reported the relationship between active trachoma and the subsequent development of incident scarring (new conjunctival scarring in those with no prior visible conjunctival scarring), Table 1 [10–13]. Three of these were from Tanzania and one from Tunisia. In addition, we identified a report of a mathematical model of the five year incidence of TS, based on cross-sectional age-stratified data from Tanzania [14]. The study designs were heterogeneous in terms of sample size, clinical grading system used, MDA treatment background and study duration. All four were conducted in hyperendemic communities, which had not previously received mass drug administration (MDA) for trachoma control. After enrolment all the communities received some form of antibiotic treatment, although this was variable in nature and frequency (Table 1). The duration of the follow-up varied from 5 to 14 years. The assessment of clinical signs used either the simplified or detailed WHO Trachoma Grading Systems [2, 3]. In two studies the development of incident scarring was identified by a change in the clinical field grading [10, 11]. In the other two studies the development of incident scarring was identified through the grading of photographs, using a more detailed scar grading system [12, 13].

The earliest study, from Tunisia, found that 17% of individuals (n = 82, children and young adults) without conjunctival scarring at baseline developed severe scarring with distortion of the tarsal plate (C3) over a 14 year period (Table 1) [10]. This study also reported on scarring progression and incident TT, which are discussed below. It did not report the proportion of individuals who developed milder degrees of scarring during the same period, nor did it report chlamydial infection data. A strong association was found between the presence of TI at baseline and the subsequent development of severe scarring (RR = 18), and a weaker association for the presence of TF (RR = 2.8). Unfortunately, the reported details are fairly limited, and the scarring incidence rates are not disaggregated by gender.

The first Tanzanian study was a 7-year cohort of children (baseline age 1–7 years) with and without “constant severe trachoma” (defined by the presence of TI on at least 3 out of 4 examinations during the baseline year) [11]. At 7 years, 29.2% of children with “constant severe trachoma” (n = 96) developed incident TS, compared to only 9.6% of the comparison group (n = 94). Incident scarring was independently associated with “constant severe trachoma”, increasing age and female gender. Incident scarring was associated with the presence of TI at the 7-year time-point. The cohort was tested for C. trachomatis infection at the 7-year time-point only; those that had incident scarring were more likely to be infected at that point (OR 2.48, p = 0.02).

In a separate cohort of Tanzanian children (n = 183) who were aged <10 years at baseline, the 5-year cumulative incidence of TS was 17.4%. In a logistic regression model, the development of TS was strongly associated with increasing age, female gender, “constant severe trachoma” (10+ follicles or TI or both, on at least 3 of 5 visits during the initial 18 months) and/or “constant C. trachomatis infection” (infection on at least 3 of 5 visits during the initial 18 months), compared to those without constant severe trachoma or constant infection, Table 1 [12]. The infection and disease states were not analysed as separate variables.

In the fourth prospective study to investigate incident conjunctival scarring in previously unaffected individuals, a cohort of Tanzanians of all ages (n = 367) was recruited and followed for 5 years (Table 1) [13]. The overall cumulative incidence of TS was 20.4% in 5 years. The cumulative incidence of scarring generally increased with age, although it was noted it was particularly rapid in younger children, for reasons that could not be determined. In this study the relationship between incident TS and the signs of active trachoma or the presence of infection were not reported. The study found that at baseline and at five years, across all age groups, the cross-sectional prevalence of TS was higher in females than males. However, data disaggregated by gender were not presented on the 5-year incidence of new scarring in this cohort.

The data from a mathematical model supports these longitudinal studies. The 5-year cumulative incidence of new scarring in females, based on cross-sectional data, indicated increasing incidence with increasing age: 3.1% in 15–19 year olds and 14.3% in 55–59 year olds [14].

Overall, there are surprisingly few prospective studies and relatively limited data investigating the link between active trachoma and/or chlamydial infection and the development of incident TS. Three studies present consistent evidence that severe conjunctival inflammation (TI) is associated with substantially increased risk of new scarring several years later [10–12]. The evidence from these studies for an association between TF, in the absence of TI, and the development of scarring is less clear. In the Tunisian study there was an increase in risk associated with TF (RR = 2.8), however, no test statistics (95%CI) were provided to assess the strength of this evidence [10]. In the Tanzanian populations TF was common and affected most of the children at some point in the observation periods [11, 12]. However, those that were found to have only sporadic disease or infection were not at a significantly higher risk of incident TS than those with no episodes of disease or infection [12].

The evidence from these prospective studies demonstrating a link between repeated or persistent C. trachomatis infection and subsequent incident TS is limited to only one study from Tanzania [12]. In this study, data on disease and infection were combined for the analysis, so it is not possible to determine the independence of the effect of infection from disease. Notably, although a small proportion of individuals with only sporadic disease and/or infection did develop TS, they were not at significantly greater risk of scarring than individuals who had no disease or infection episodes documented. This might suggest that in a programmatic context, reducing the pressure of infection and associated TI through MDA might reduce disease in the majority of individuals below a threshold required for the development of scarring.

As would be expected from the known cross-sectional age-specific prevalence of TS from population based surveys, the three prospective Tanzanian studies and the mathematical model found clear and consistent evidence of increasing risk of incident TS with increasing age [11–14]. However, it is noteworthy that some incident scarring was developing during childhood in these populations. Similarly, the prospective Tanzanian studies found that females were at consistently higher risk of incident TS than males, probably because of a greater lifetime exposure to chlamydial infection [11–13].

West and colleagues suggested that there might be two distinct routes to the development of TS [11]. Firstly, in the “classical” model, which may account for the majority of TS cases, are individuals who are repeatedly exposed to C. trachomatis. In this group the risk of TS is primarily determined by the cumulative number of infection episodes experienced. This is consistent with the observation that TS incidence is higher in females and with increasing age. This model has since been incorporated into the design of mathematical models of the scarring sequelae of trachoma [15, 16]. The second route, which may account for a minority of people who develop TS, is characterised by the development of a severe and persistent inflammatory phenotype, “constant severe trachoma”. Data from several of these prospective studies is consistent with this hypothesis [10–12]. Whether these individuals are responding unfavourably to C. trachomatis, to other bacteria, or are genetically prone to develop a severe persistent inflammatory phenotype is unclear [4].

Progression of Scarring Trachoma

We identified four prospective cohort studies which reported the rates and risk factors for the progression of established trachomatous conjunctival scarring (Table 2) [10, 13, 17]. One was from Tunisia, one from Ethiopia and two from Tanzania. There was some heterogeneity in the study designs, although the two parallel studies we conducted in Ethiopia and Tanzania used similar designs and were reported together [17]. Three studies were conducted in hyperendemic communities [10, 13, 17] and one in a hypoendemic community [17]. The Tunisian and both Tanzanian studies were in areas which had not previously received MDA [10, 13, 17]. After enrolment, three of the study populations received some form of antibiotic treatment for trachoma control, although this was variable in nature and frequency (Table 2). The Ethiopian study was conducted in an area that had previously received MDA on several occasions and continued to do so during the study period [17]. The second Tanzanian study was conducted in an area where the current prevalence of active trachoma is below treatment thresholds [17]. The duration of follow-up varied from 2 to 14 years. The assessment of clinical signs used either the simplified or detailed WHO Trachoma Grading Systems [2, 3].

The identification of progressive scarring used different methodologies. In the Tunisian study the development of scarring was based on a change in the clinical field grading score [10]. In the other three studies the progression of scarring was identified through the grading of photographs. For one study this was done by the comparison of an independent grading of the baseline and five-year photographs [11]. In the other two studies the baseline and two-year photographs were graded using a detailed grading system for scarring [18]. The photographs were directly compared side-by-side for assessment of progression, possibly allowing for more subtle changes in TS to be detected [17].

The Tunisian study (described above) reported scarring progression in 68.5% of individuals (n = 120) over 14 years (Table 2) [10]. However, this report is limited in the level of detail provided and it appears that the total progression estimate may include some individuals who were incident scarring cases. There is some partial information on progression from mild or moderate scarring (C1 / C2) to the most severe grade (C3), Table 2. The specific risk factors for progression to C3 are not presented. However, as noted above, the development of new C3 in those who had no scarring at baseline was associated with TI at baseline.

In the first Tanzanian study progression of conjunctival scarring occurred in 47.1% of individuals (n = 85, all ages) who had some established TS at baseline, over a five year period (Table 2) [13]. There was no difference by age in the proportion with evidence of progression. No data on the relationship between scarring progression and gender, clinical inflammation or C. trachomatis infection were reported.

In our study in Ethiopia, progression of conjunctival scarring developed in 23.1% of participants (n = 585) over a 2 year period (Table 2) [17]. This was a prospective cohort of adults who had minor trichiasis at baseline that was managed by epilation. They were examined and photographed every six months, with swab samples collected for C. trachomatis PCR and conjunctival gene expression analysis. There was a strong relationship between progressive scarring and increasing number of observed episodes of clinical inflammation, defined as P2 or P3 in the detailed WHO grading system (OR 5.93, 95%CI 3.3–10.6, p<0.0001). There was no evidence of an association between scarring progression and gender, age or body mass index (BMI). The samples from individuals with progressive scarring and a frequency matched sample of individuals who did not show progressive scarring were tested for C. trachomatis by PCR on 4 occasions; all samples were negative for C. trachomatis. Conjunctival inflammation, but not scarring progression, was associated with increased gene expression of a mixture of pro-inflammatory cytokines (IL1B, IL17A, CXCL5), anti-microbial effectors (S100A7) and factors associated with tissue remodelling (CTGF, MMP7, reduced SPARCL1).

In the parallel cohort study we conducted in Tanzania we used the same protocol, including sample collection, as the Ethiopian study; all individuals had conjunctival scarring at baseline, were aged 18 years or more and no individuals had baseline trichiasis [17]. Progression of conjunctival scarring was observed in 30.0% of participants (n = 577) over a 2-year period (Table 2). Again, there was a strong relationship between progressive scarring and an increasing number of episodes of clinical inflammation (P2/P3), of a similar magnitude to that found in the Ethiopian study (OR 5.76, 95% CI 2.6–12.7, p<0.0001). There was no evidence of an association between scarring progression and gender, age or BMI. In this Tanzanian cohort there were 804 people with TS recruited and tested for C. trachomatis by PCR at baseline, of which only 4 samples were positive. At 6, 12 and 18 months after baseline, the samples from the individuals with progressive scarring and a frequency matched sample of individuals in the cohort who did not show progressive scarring were also tested for C. trachomatis and only one sample was positive. In this study progressive scarring was associated with a modest increase in expression of IL1B and S100A7. In addition, increased S100A7, IL17A, CXCL5, MMP7 and CEACAM5 expression were associated with clinical inflammation (P2/P3).

A consistent finding across these studies was that trachomatous conjunctival scarring continues to progress over time. The studies are probably too heterogeneous to provide an overall estimate range for disease progression. However, it is noticeable that in the two studies using the same protocol the rates of scarring progression were similar [17]. Moreover, it might have been anticipated that the Ethiopian cohort, which was conducted in a hyperendemic region, would have experienced a greater proportion of progression than that observed in a region of Tanzania that had a low level of active trachoma, and was therefore not included in the national antibiotic distribution programme.

There was a strong association between conjunctival inflammation and progressive scarring, with individuals experiencing more inflammatory episodes being at greater risk of progression [10, 17]. In two studies the immunofibrogenic correlates of clinically apparent inflammation were investigated using gene expression analysis, and this demonstrated that its presence was associated with increased expression of a range of pro-inflammatory factors [17]. These observations are consistent with the findings for incident scarring outlined above; children who experienced more episodes of active trachomatous inflammation were more likely to develop incident scarring (Table 1) [11, 12].

There are three noteworthy findings from these four cohort studies of progressive scarring. Firstly, we found no published evidence that the progression of established scarring was associated with the detection of C. trachomatis infection [17]. Indeed, infection was detected in only very few individuals with TS. This contrasts with the observation, albeit from limited prospective data, of an association between the development of incident scarring and episodes of chlamydial infection [12]. Secondly, we found no published evidence of a difference between males and females in the proportions showing signs of scarring progression. Again, this contrasts with the findings for incident scarring outlined above, which developed more frequently in females (Table 1). Thirdly, amongst people with established scarring the proportion showing scarring progression did not vary with age. This also contrasts with the findings for incident scarring, which developed more frequently with increasing age.

Therefore, once TS is established the risk of scarring progression does not appear to be associated with exactly the same set of risk factors that were associated with the development of incident scarring. This could be interpreted as suggesting that progression of established scarring is less dependent than incident scarring on recurrent C. trachomatis infection, episodes of which are more frequent in females and accrue with increasing age. It may also indicate that other factors such as non-chlamydial bacterial infection, a dry ocular surface or altered tissue inflammatory or scarring responses may also contribute to a pro-inflammatory / pro-fibrotic state. This might be anticipated as scarred ocular surface tissue probably has altered physiology and compromised defence mechanisms. However, prospective data on these factors has not been reported.

Progression to Trachomatous Trichiasis

We identified four prospective cohort studies which report rates and risk factors for the development of trachomatous trichiasis (Table 3) [10, 19–21]. One was from Tunisia, one from Tanzania and two from The Gambia. In addition, we identified a mathematical model based on cross-sectional data from Tanzania [14]. Two studies were conducted in hyperendemic areas and two in hypoendemic areas. None of the study populations had previously received systematic MDA for trachoma control. After enrolment, two of the study populations received some form of antibiotic treatment for trachoma control, although this was variable in nature and frequency (Table 3) [10, 21]. The duration of follow-up varied from 5 to 14 years.

The Tunisian study (described above) reported incident TT in 8.0% over a 14 year period [10]. The risk of developing TT was related to the baseline conjunctival scarring severity, being 1.2% in people with no scarring and rising to 38.1% in those with severe baseline scarring (C3), Table 3. The Tanzanian study was a 7 year cohort of females with and without TS at baseline [19]. In those with TS at baseline 9.2% developed TT, and of those without TS at baseline 0.6% developed TT by 7-years. The development of TT was associated with increasing age and the detection of C. trachomatis infection at the 7-year time-point. The number of individuals with C. trachomatis infection at follow up was 60/513 (11.7%) in the group with TS at baseline and 36/501 (7.5%) in the group with no TS at baseline. The first of the Gambian studies reported the 12-year incidence of TT in people with TS at baseline to be 6.4% [20]. The development of TT was associated with increasing age. The second Gambian study was a five year study of all the residents of 14 villages, most (94%) of whom did not have scarring at baseline. The five year cumulative incidence of TT in those >15 years was 0.4% [21].

The mathematical model estimated the 5 year cumulative incidence of TT in women from a hyperendemic region of Tanzania using age-stratified cross sectional data from different clinical stages [14]. The cumulative incidence of TT was highest in older women (6.4% versus 0.3% in young women) and in women with TS at baseline (15.1% in 55–59 year olds, 3.2% in 15–19 year olds).

Progression of Trachomatous Trichiasis

We identified three prospective cohort studies that reported the rates and risk factors for the progression of trachomatous trichiasis from unilateral to bilateral disease, or from minor trichiasis (<6 lashes touching the eye) to major trichiasis (6+ lashes touching the eye), Table 4 [22–25]. Two of these were from The Gambia and one from Ethiopia. The Gambian studies were both conducted in hypoendemic areas. The Ethiopian study was in a hyperendemic setting with on-going MDA. The duration of follow-up varied from 1 to 4 years.

In the first Gambian study, 46% of people with unilateral trichiasis (n = 46) had developed bilateral disease in one year [22]. Of those with minor trichiasis (n = 55), 33% progressed to major trichiasis. In the second Gambian study 29% of people with unilateral trichiasis (n = 42) progressed to bilateral disease in four years and 37% of minor trichiasis (n = 75) progressed to major trichiasis [23]. In this second study several factors were analysed in relation to the progression of trichiasis over four years: age, gender, conjunctival inflammation (at four years), chlamydial infection and pathological bacterial infection (at four years). In univariate analysis only conjunctival inflammation at 4-years was associated with TT progression. However, no factor was significant in a multivariable model. It is noteworthy that at 4-years both conjunctival inflammation and bacterial infection were independently associated with the presence of major trichiasis. Only 2/146 samples tested were positive for C. trachomatis at the 4-year time-point.

The Ethiopian study was of a group of people with minor trichiasis (n = 650) who were in the epilation arm of a randomised controlled trial of epilation vs. surgery, with regular follow up over a two year period [24]. Individuals were equipped with high quality forceps and a relative was trained in how to perform epilation. During the first two years 13% of study eyes progressed from minor to major trichiasis. After two years all participants were offered surgery for their trichiasis. About a third accepted surgery at two years, while the majority (n = 383) chose to continue with epilation and were followed up for an additional two years [25]. Comparing the amount of trichiasis by counting the absolute number of lashes touching the eye at baseline with that at four years, 21.4% of these 383 individuals were found to have had some degree of increase in the number of lashes touching the eye, however this was difficult to assess in the context of regular epilation. In addition, 36.6% had only a modest increase in the degree of entropion, although it should be noted that this is a variable and difficult clinical sign to grade. In a multivariable model, over the four years, progression to major trichiasis was associated with older age and having ≥3 lashes touching the eye at baseline.

Progression to Corneal Opacification, Visual Impairment and Blindness

We identified four prospective cohort studies that reported the rates and risk factors for the development of corneal opacification (CO) and visual impairment (Table 5) [20, 22–25]. Three were from The Gambia, conducted in hypoendemic areas. One study was from a hyperendemic area of Ethiopia, with on-going MDA. The duration of follow-up varied from 1 to 12 years. In addition, the mathematical model based on Tanzanian cross-sectional data provided 10 year estimates for the development of CO [14]. We have not included prospective data on CO change from clinical trials where TT was treated with surgery.

The first Gambian study found the 12-year cumulative progression from TS to CO to be 5.9% and for TT to CO to be 20% [20]. Risk factors for incident CO were increasing age and TT at baseline. Visual impairment developed in 16.5%, but only 2.5% of this was attributed to CO; most was due to cataract (Table 5). The second Gambian study found that in one year incident CO developed in 10% of people with TT and that CO progressed in 34% with major trichiasis [22]. New visual impairment or blindness developed in 9%. The third study from The Gambia found incident CO developed in 7.6% of eyes with TT at baseline over the course of four years, with the risk being greater for those with major TT (Table 5) [23]. Again, the risk of visual impairment was increased with TT, however, the large majority of incident cases of visual impairment were due to cataract.

The four year Ethiopian study of people with minor trichiasis has been outlined above [24, 25]. Incident or progressive CO was determined by the comparison of photographs in this study. At 2 years, 5.5% of the 650 individuals with minor trichiasis had some increase in CO. When modelled at 4-years, incident/progressive CO was associated with being ≥50 years of age and having moderate CO at baseline. Visual acuity deteriorated in 14.5% by >0.3LogMAR by 2 years, however, most of this was not attributable to changes in CO.

The mathematical modelling study, described above, found that the overall incidence of CO was higher in older women: 2.8% in 45–54 years versus 0.16% in 15–24 years [14]. The cumulative incidence rates were estimated to be very much higher in those with TT at baseline: 53.5% in 45–54 year olds and 27.2% in 15–24 year olds. This model estimated around half of all CO was due to causes other than trachoma. In women <35 years these other causes dominated, whereas in older women trachoma was the main cause.

Conclusions

Conducting long-term prospective studies of the rates and risk factors of progressive trachoma from active disease through to blindness is complex and expensive. Here we have reviewed the longitudinal studies that contribute to our understanding of this disease process. In addition to these longitudinal studies, there are many cross-sectional studies that provide indirect evidence about risk factors for progression. The prospective studies are relatively few in number, of variable design and frequently small in size. The determination of the presence of progression in conjunctival scarring can be challenging, particularly where there is also extensive inflammation masking scar tissue. Despite these limitations, collectively they provide evidence that supports much of the widely accepted description of the natural history of trachoma, illustrated in Fig 1.

The studies of incident and progressive conjunctival scarring were consistent in showing a strong association with repeated or persistent conjunctival inflammation. In contrast, the evidence linking the development of incident conjunctival scarring with C. trachomatis infection is limited to only one study. Moreover, in two large cohorts with scarring progression rates of 23.1% and 30% over 2 years only very few episodes of chlamydial infection were detected.

The model proposed by West and colleagues, and outlined above, suggests that the majority of individuals who develop TS do so after repeated rounds of infection by C. trachomatis. In addition, there appears to be a subgroup of individuals who experience repeated or protracted intense inflammation, who are at increased risk of developing scarring. It is anticipated that the introduction of control programmes involving MDA and F and E interventions to limit transmission, will reduce the number of episodes of infection experienced and therefore the risk of developing incident scarring. However, the potential impact of control measures on halting progression of previously established scarring is less certain. A treatment intervention that could reduce chronic conjunctival inflammation in adults in trachoma-endemic areas might be desirable to prevent scarring progression.

The lack of a clear association between chlamydial infection and scarring progression in adults could have a number of explanations [17]. It is possible that the number and frequency of clinical observations and tests for chlamydial infection in these studies were too few to capture infection events. It is also possible that the detection of infection in adults is more difficult than in children, because of lower infection loads or shorter infection duration, due to the acquisition of some protective immunity [26–28]. This potential explanation might imply that only brief bursts of infection with C. trachomatis is needed to trigger chronic inflammation and an associated scarring response.

Alternatively, it is possible that conjunctival inflammation observed in adults with established conjunctival scarring may have additional causes besides C. trachomatis. A number of studies have found cross-sectional associations between non-chlamydial bacteria cultured from the conjunctival surface and the presence of inflammation at all stages of the natural history of trachoma [23, 29–33]. Currently however there are no longitudinal data to determine whether this inflammation associated with other infections represents a pro-fibrotic state that contributes additional scarring, or whether it is a by-product of the altered physiological environment of a scarred conjunctiva. Longitudinal studies that include tests for C. trachomatis and non-chlamydial bacteria are needed to determine their relative contributions to scarring progression.

These observations raise important issues of programmatic significance for trachoma control, particularly over whether conjunctival scarring can continue to develop and progress in individuals with intermittent or chronic inflammation, but in the absence of on-going regular chlamydial infection. Several long-term population based studies of the impact of MDA have demonstrated that even after C. trachomatis has been cleared or brought down to very low prevalence levels, the prevalence of signs of active conjunctival inflammation persist for some years [21, 34]. While it is possible that the cross-sectional nature of these studies may have missed recent infection episodes, it is striking that the inflammatory phenotype was still present after the prevalence of C. trachomatis infection had been reduced to a low level, and this may have significance for disease progression. A question worth consideration is whether the clinical sign TI should be included in the monitoring of trachoma control in addition to TF, because of the consistently strong association between TI and incident and progressive scarring. Finally, these data support the recommendation that trachoma control programmes maintain mechanisms to detect and treat trichiasis for many years to come.