Systematic review of the incidence of post-operative trichiasis in Africa

Background Surgery for trichiasis is one of the pillars of the World Health Organization’s strategy for global elimination of trachoma as a public health problem. A high incidence of post-operative trichiasis or other poor surgical outcomes could jeopardize these efforts. In this review, we aimed to summarize the reported incidence of post-operative trichiasis and other poor outcomes of trichiasis surgery in Africa. Methods We conducted a systematic literature search using PubMed, Academic Search Premier, Africa-Wide Information, CINAHL and Health Source Nursing through EBSCOhost, Web of Science, and the Cochrane Central Register of Controlled Trials. Reference lists of included studies were also reviewed to identify further potentially relevant publications. All observational and interventional studies that measured post-operative trichiasis in Africa as an outcome of trichiasis surgery were included. Results Thirty-five papers reporting on 22 studies (9 interventional,13 observational; total 13,737 participants) met the inclusion criteria. The reported incidence of post-operative trichiasis in the included studies ranged from 2% (at 6 weeks after bilamellar tarsal rotation) to 69% (at 3 weeks after anterior lamellar repositioning). The incidence varied by surgical procedure, study design, and length of follow-up. Conclusion Trichiasis surgical outcomes should be improved. National trachoma programmes could benefit from identifying and adopting strategies to improve the performance and quality of their surgical service.

preventing trachoma-related blindness and to encourage other trichiasis patients to consent to an operation, there must be good long-term surgical outcomes with a low incidence of post-operative trichiasis.
WHO's third global scientific meeting on trachoma [8], convened in 2010, recommended that national trachoma programmes should report the incidence of post-operative trichiasis, and target a cumulative postoperative trichiasis incidence of ≤10% at one-year postsurgery. However, several studies report incidence estimates that are considerably higher than this [9]. There are only limited operational data on the incidence of post-operative trichiasis published from settings where trachoma elimination programmes are active [9].
Generating an evidence-based understanding of the magnitude and determinants of post-operative trichiasis would help inform discussions on current outcome targets and strategies to improve surgery outcomes for trichiasis. This review aims to partially fill this gap by consolidating data on incidence of post-operative trichiasis and other poor outcomes of trichiasis surgery from observational and interventional studies conducted in Africa.

Ethical approval
The protocol was approved by the University of Cape Town Faculty of Health Sciences' Human Research Ethics Committee (076/2018) and deemed by the WHO Ethics Review Committee to not require full formal ethics review (0003034). The protocol was registered on PROSPERO: CRD42018085253.

Search strategy
We undertook a systematic search of PubMed, Academic Search Premier, Africa-Wide Information, CINA HL and Health Source Nursing through EBSCOhost, Web of Science [all databases], and the Cochrane Central Register of Controlled Trials, including material published up to and including the month of May 2018. In addition, Google Scholar and the reference lists of relevant reviews and all eligible papers were searched to cross-check for studies not already identified. Websites of organizations identified in published studies were searched for pertinent grey literature. No language limitations were applied in these searches.
The full search strategy is provided as a supplemental file.

Study selection
Two reviewers (GM, PC) independently screened the titles and abstracts of the articles found to determine their potential eligibility for inclusion. The full texts of potentially eligible studies were obtained. Selection for inclusion into the review was conducted by the two reviewers working independently. Any disagreements regarding inclusion of studies were resolved by discussion or by consulting the third reviewer (AWS). Included studies met the following criteria: 1) conducted within Africa; 2) measured post-operative trichiasis (defined as the presence of one or more eyelashes touching the eyeball or evidence of epilation of inturned eyelashes after surgery) as an outcome; 3) included participants aged ≥15 years with previously unoperated trachomatous trichiasis in at least one eye within either [a] an interventional or [b] observational study.
Editorial articles, reviews, expert opinion pieces, conference papers and meeting abstracts were excluded, as were studies conducted outside Africa, studies that did not have a surgical intervention, and those that did not measure or report post-operative trichiasis as an outcome. Any studies published before 1990 were also excluded.
We divided studies into two groups -observational and interventional studies -with the expectation that surgeons operating in the latter were more likely to be highly selected and re-trained before the study than those in the former. It should be noted however, that some interventional studies included in this review did not report criteria-based selection or re-training of surgeons before the intervention.

Qualitative assessment of studies
The methodological quality of the papers included in this review was assessed using a slight adaptation of the Joanna Briggs Institute critical appraisal checklist for prevalence studies [10]. Critical appraisal was conducted by two reviewers independently, with disagreements resolved through discussion or by consulting the third reviewer.

Assessment of heterogeneity
Studies included in this systematic review were checked for heterogeneity by examining their characteristics (study design, population, follow up period, etc.) and risk of bias. Statistical heterogeneity was assessed using I 2 and Chi [2] tests. Heterogeneity was considered as significant if the I 2 was above 50%.

Data synthesis
The studies included in this review were varied in terms of design, interventions, follow up period and outcomes. As a result, there was high statistical heterogeneity which precluded pooling of data for the outcomes (I 2 > =95%). RevMan 5.3 software [11] was used for all quantitative data analyses in this review.
For the intervention studies, treatment effects were measured by calculating odds ratios for the different interventions and 95% confidence interval for these. Dichotomous data from these studies are presented in forest plots without a summary estimate, and as a narrative summary. For observational studies, data are presented as a percentage of those who developed the outcome of interest out of the total number of study participants.

Results of the search
Initial electronic literature searches yielded 5003 articles. After removal of duplicates and a review of titles and abstracts of these and other articles identified by handsearching, 97 publications were selected for detailed review and possible inclusion. Of these, 61 articles were excluded for the following reasons: conducted outside Africa (n = 23); did not measure post-operative trichiasis as an outcome (n = 21); the condition of interest was not trichiasis due to trachoma (n = 3); the article type was a review, editorial or expert opinion piece (n = 14). One prospective study [12] was subsequently also excluded from the review, despite having met the inclusion criteria, due to internal inconsistencies in the description of study methods and findings. Ultimately, 35 articles describing 22 studies (nine interventional and 13 observational studies with a combined total of 13,737 participants) were included. Figure 1 summarizes the flow of the paper identification and selection process. A list of excluded studies is provided as a Supplementary file. Table 1 summarises the characteristics and quality of the 22 studies that met the inclusion criteria. In the  22 studies, the duration of follow-up ranged from 3 weeks to more than 7 years. The incidence of postoperative trichiasis varied widely, from 2.3% [18] to 65% [33] ( Table 1). Bilamellar tarsal rotation (BLTR) and posterior lamellar tarsal rotation (PLTR; sometimes referred to as the Trabut procedure) were the most commonly reported procedures, used in 16 and 13 studies respectively. Other procedures used were anterior lamellar rotation; transverse tarsotomy and lid margin rotation; and lid margin split [13,29,32].

Study characteristics and quality
We included all eligible studies regardless of the surgical procedure used, though BLTR and PLTR are the two procedures recommended by WHO for treatment of trachomatous trichiasis [47]. Twenty of the 22 studies included in this review had relatively low risk of bias from the design, conduct and analysis standpoint. Two observational studies [32,38] had relatively small sample sizes which might have affected their study power. Findings from those two studies should therefore be interpreted with caution.
In the latter study, pre-operative corneal staining, corneal opacity and the use of silk sutures were identified as risk factors for early failure of BLTR.
Other risk factors associated with post-operative trichiasis and other poor outcomes include surgeon's experience and technique, short incision lengths, baseline severity of disease and epilation before surgery [18,36]. Two intervention studies reported similar incidence of post-operative trichiasis; 10.4% [13] (12/115 patients) and 11.4% [14] (81/713 eyelids), while an observational study [37] conducted in Ethiopia reported a much higher incidence of post-operative trichiasis of 23.4% (308/1317 eyelids); 95% CI: 19.0-27.8), 6 months after BLTR. In the first study [13], the incidence of post-operative trichiasis was lower in the BLTR group compared to the TTR group (OR 0.84; 95% CI: 0.37-1.86), although this difference was not statistically significant. In the second intervention study [14], the incidence of post-operative trichiasis was higher in the intervention group (those operated on by an ophthalmologist) compared to those operated on by integrated eye care workers (OR 1.32; 95% CI: 0.83-2.11). In all 3 studies, a number of other poor outcomes were noted, including overcorrection, lid notching, pyogenic granuloma, local madarosis and conjunctival and eyelid inflammation [13,14,37].
Follow-up period: over 3 months but less than 1 year A randomized controlled trial conducted in The Gambia to investigate whether perioperative single-dose oral azithromycin could improve surgical outcomes after trichiasis surgery reported a cumulative incidence of 31.5% (129/410 patients) post-operative trichiasis 6 months after PLTR [15]. In this study, the incidence of postoperative trichiasis was higher in the intervention (azithromycin) group compared to the usual care  [15], but in multivariable logistic regression models there was no statistically significant difference between the groups. However, post-operative trichiasis was significantly associated with severity of preoperative disease and inter-surgeon variability [15]. In contrast, the Surgery for Trichiasis, Antibiotics to Prevent Recurrence (STAR) Trial conducted in Ethiopia reported a relatively low incidence of PTT; 4.6% (36/790 operated eyelids); 4.2% in the azithromycin group, 7.9% in the control group; OR 0.48; 95% CI: 0.21-1.10) 6 months after BLTR [6] (Fig. 2). In the latter study, surgeons had been highly selected and retrained before the intervention; while the Gambian study employed the available community ophthalmic nurses.

Follow-up period of 1 year
Among studies with a one-year follow-up period, the highest incidence of post-operative trichiasis at 41.3% (176/426 patients) was reported in the aforementioned Gambian study after PLTR surgery [15]. Unlike at 6 months followup, when the incidence of post-operative trichiasis was higher in the intervention (azithromycin) group than in the control group, at 1 year there were fewer cases of postoperative trichiasis in the intervention than in the control group; this difference was not statistically significant. In contrast, the STAR trial [19,20] reported a relatively low incidence of post-operative trichiasis (7.6%; 107/1414 patients) 1 year after BLTR with significant difference seen between those who received peri-operative azithromycin and those in the control group (OR 0.66; 95% CI: 0.44-0.99). In combining the data from these two trials, those in the intervention (azithromycin) group were less likely to develop post-operative trichiasis (OR 0.81; 95% CI: 0.55-1.20) but the effect was not statistically significant (Fig. 3).
Similarly, at three and 4 years, high incidences of postoperative trichiasis (> 20%) were reported regardless of the surgical procedure used [24,34,35,45]. In the Gambia study [15], the incidence of PTT was high (41%; 110/266 eyelids), 4 years after surgery. This was associated with increasing conjunctival inflammation at 4 years [23]. The STAR trial, however, was exceptional, Fig. 3 Summary of studies comparing peri-operative azithromycin to no azithromycin, with outcomes measured at 1 year reporting a cumulative incidence of post-operative trichiasis of only 12% (161/1322 patients) [19][20][21] in the 3 years after BLTR, with less post-operative trichiasis in the intervention group (OR 0.76; 95% CI: 0.54-1.07] - Fig. 4 Comparison between surgical procedures and study types The incidence of post-operative trichiasis was high (> 20%) across different follow-up periods regardless of the surgical procedure used (Figs. 5 and 6). Only one study provided a direct comparison between the two most common procedures for trichiasis surgery -PLTR and BLTR [25]. In this study conducted in Ethiopia, PLTR was found to yield a lower incidence of post-operative trichiasis compared to BLTR (OR 0.51; 95% CI: 0.36-0.72).

Discussion
Our systematic review suggests that there is a huge disparity between policy and practice with respect to the incidence of post-operative trichiasis in Africa. WHO recommends that national trachoma programmes strive to achieve ≤10% post-operative trichiasis by 1 year after surgery [8], but in the eight studies we identified that had one-year data, only one had an incidence of postoperative trichiasis ≤ 10% [19]. This suggests that there is a need to put in place or strengthen existing measures to improve the proportion of patients achieving satisfactory outcomes. A logical approach would be for national trachoma programmes to establish and implement policies and systems to follow up patients, assess surgical outcomes and maximize the performance of individual surgeons through post-surgical audits with provision of additional support, where needed.
BLTR and PLTR are currently the most commonly used procedures, with evidence presented elsewhere suggesting that they produce better outcomes than other techniques [13,23,24]. Only one recent study [25] has directly compared the two. Additional research is warranted to confirm the apparent superiority of PLTR [49].
Our study has clear weaknesses for estimating the rate and determinants of success in trichiasis surgery programs; these should be acknowledged before further conclusions are drawn. We did not collect program-level data, but rather collated reported post-operative trichiasis incidence estimates from the literature, with referenced papers derived from the experience of academically supported settings over three decades. The contemporary incidence of bad outcomes when surgery was being routinely performed in unselected program delivery environments could easily have been higher (or lower) than noted, and there are too few data points and too many potential explanatory variables to try to infer possible changes in incidence over time.  Data on the phenotype of post-operative trichiasis were generally not published; the potential implications for vision of having one or two peripheral trichiatic eyelashes are likely to be considerably less than those of large numbers of central trichiatic eyelashes. It is also important to note that most studies included in this review were conducted before publication of the International Coalition for Trachoma Control's training and supervision guidelines [50,51], and before mannequinbased surgical simulation training [52] became available. These measures are likely (though formally unproven) to improve the outcomes of trichiasis surgery.
How else could success rates be optimized? A large randomized controlled trial in Ethiopia suggested that a single dose of peri-operative azithromycin was associated with a 33% reduction in post-operative trichiasis by 3 years after surgery, compared with topical tetracycline prescribed for 6 weeks [19][20][21]. These data differ from a smaller clinical trial comparing peri-operative oral azithromycin with two post-surgical weeks of topical tetracycline conducted in The Gambia [15], where no significant difference in outcomes was observed. One important difference between those two trials was the absence of trial-specific training and standardization of surgeons in the Gambian study, which might have resulted in a higher incidence of post-operative trichiasis; significant variability between surgeons was noted, with incidence of post-operative trichiasis ranging by surgeon from 0 to 83%. The Ethiopian study [19] used surgeons who had been trained and certified by the study team prior to the start of the trial. One interpretation of this combination of findings would be that, though azithromycin might have a protective effect against postoperative trichiasis, it cannot overcome shortfalls in training, standardization and supervision of surgeons. Other differences between the two environments might also, or alternatively, be responsible. The mechanism for azithromycin having an impact, if it does, is not necessarily clear: it could potentially relate to either the antiinflammatory or anti-infective action of the macrolide.
In another randomized controlled trial conducted in Ethiopia [17], there was no evidence that use of absorbable sutures was associated with a lower incidence of post-operative trichiasis at 1 year after surgery compared to the use of silk (although absorbable sutures were found to reduce the incidence of granulomata) [16,Fig. 5 Incidence of PTT among patients who had BLTR, by type of study and duration of follow-up 17]. Despite careful standardisation, there was significant variability in outcomes between different surgeons in this trial. To protect patients, operator-dependent factors, including dexterity, handedness, visual acuity, rigor of training and certification, supervision and motivation, must always be considered possible contributors to the occurrence of post-operative trichiasis. The tendency noted in the current review for post-operative trichiasis incidence to be lower in interventional than in observational studies was most marked for interventional studies where surgeons were re-trained prior to initiation of the study, reinforcing the (perhaps expected) point that greater skill levels drive better results.

Conclusion
Since 1990, published estimates suggest that anatomical outcomes of the surgical management of trachomatous trichiasis are worse than expected by WHO. More effort needs to be made to better understand the routine incidence, phenotype and determinants of post-operative trichiasis. This knowledge is critical to help design strategies to minimize its occurrence. In the meantime, stringent adherence to recommended follow-up schedules [53] is recommended to ensure that patients experiencing sub-optimal outcomes are detected early and offered appropriate care.