Katsaus «Ernest PJ, Viechtbauer W, Schouten JS ym. The infl...»1
A systematic computerized search was performed in PubMed, EMBASE, and all databases
and registers of The Cochrane Library, in April 2009 The search was limited to articles
in English, Dutch, French, or German. A total of 2450 articles were identified. All
titles and abstracts were screened, and articles were excluded based on predefined
exclusion criteria. Of the remaining 782 articles that were studied completely, 48
articles fulfilled the selection criteria. The selected studies had to follow patients
with glaucoma for minimally 1 year with the use of conventional visual field examinations.
Twelve articles that studied 30 methods in ten studies were included in the meta-analysis.
All methods were named and classified in six groups according to their main characteristics.
1. Glaucoma progression analysis (GPA)
- similar to the visual field endpoint in the EMGT study
- an event analysis based on pattern deviation values and is included in the HFA software.
- When significant deterioration (p < 0.05) is seen on the pattern deviation probability
maps of the GPA printouts in the same three or more points on three consecutive follow-up
tests, the software interprets this as likely progression.
2. Group (AGIS & CIGTS)
- Two methods that were based on the AGIS method, use a scoring system to grade each
visual field in the follow-up period.
- The AGIS score is based on the actual decibel deviations at the total deviation plot,
while the CIGTS algorithm is based on the p-values obtained from the total deviation
- Both scoring systems range from 0 to 20, with 0 representing no field loss and 20
- Visual field series are considered to be progressive if the score has a minimal increase
of four (with the AGIS method) or three points (with the CIGTS method) and is confirmed
by two additional tests.
3. Point-wise linear regression (PLR)
- a linear regression analysis is performed in different individual locations at the
4. Linear regression analysis with visual field indices
- Visual field index (VFI) is calculated by the software of the HFA. Each location on
the visual field contributes to the VFI, although it is more heavily weighted to central
areas of the visual field. A location that is not significantly (p < 0.05) depressed
on the pattern deviation probability map is considered to have a 100 % sensitivity.
The VFI is expressed as one percentage, where 100 % represents a normal visual field
and 0 % represents a perimetrically blind eye. The HFA software performs a linear
regression analysis of the VFI against time.
5. Combined a PLR and a linear regression analysis of the MD value
- The methods in this group were variants of the threshold noiseless trend (TNT) program
- The TNT program filters perimetric results and takes into account dependency relations
in the visual field.
- Moreover, it combines linear regression analyses of the MD, the cumulative defect
curve, and different locations at the visual field.
- Suspected progression is seen for the first time that one of these parameters indicates
progression. If this result is repeated by two consecutive examinations or if two
or more parameters indicate progression, TNT indicates definite progression
6. Clinical group - methods based on clinical judgement
- Classified in this group were methods based on entirely subjective assessments of
visual fields by multiple observers, who had to agree on progression while they were
blinded for other clinical data.
- Other methods in this group used certain algorithms for the assessment of visual
fields, for example based on the clinical judgement of scotoma’s.
- One clinical method was based on nonparametric ranking of MD values. This method
objectifies the commonly practised method of monitoring glaucoma patients with the
use of MD values. A visual field series is considered progressive as the MD value
of a follow- up visual field is worse than the MD of the worse of two baseline fields.
This has to be confirmed on at least two visual fields.
- In total, 1040 eyes of 948 patients with glaucoma were studied in the ten studies
- All patients were derived from western countries, with mean baseline MD values ranging
from -3.3 to -10.4 dB, and mean age ranging from 58 to 73 years.
- An average of 1.7 visual fields per year were analysed in the studies.
- Patients received various treatments during the follow-up period.
The mean estimated incidence proportion of progression
- 0.21 (95 % confidence interval (CI) 0.15, 0.26) in 6 years, indicating that on average
21 % of the study eyes progressed in 6 years (range from 2 % to 62 % in 6 years, depending on the method)
- The incidence proportions of progression according to 30 methods ranged from 0.02
(CI -0.02, 0.05) to 0.62 (CI 0.47, 0.78).
- GPA was the most frequently studied method, with six studies in this meta-analysis.
With an incidence proportion of 0.16 (CI 0.14, 0.19), the GPA is in the middle of
the ranking of all 30 methods.
- The AGIS based methods and most of the methods based on linear regression analysis
with indices showed lower incidence proportions than the GPA method.
- Methods that are based on clinical judgement or the TNT program showed higher incidences
than the GPA.
Factors associated with incidence of progression
- Follow-up time was significantly associated with the incidence of progression, with an increase in the incidence proportion of approximately 2.1 per cent points
per extra follow-up year (p < 0.001).
- Baseline MD was also significantly associated with the incidence of progression, increasing the incidence proportion by 0.9 per cent points per extra dB of MD loss
(p = 0.025). No accelerating (quadratic) relationships between these two predictors
and the incidence proportions were found (p = 0.93 and p = 0.77, respectively).
- Approximately 82 % of the heterogeneity in this analysis can be accounted for by the
variety of methods used in studies. The rest of the heterogeneity was explained by the mean baseline MD value and the
- Tutkimuksen laatu: tasokas
- Sovellettavuus suomalaiseen väestöön: hyvä
The results of the model can only be generalized for clinically treated glaucoma patients
with a mean baseline MD value around -7 dB and a mean follow-up time of 6 years. The
estimates of progression should be corrected by adding 0.9 % to the incidence or by
subtracting 0.9 % from the incidence, for each dB decrease and increase in baseline
MD value, respectively. In the same way, the incidences should be corrected by adding
2.1 % for each year extra follow-up. The chosen method accounted for nearly all differences
in the incidence of progression that we found in the included studies, with the exception
of the part that can be explained by the baseline MD value and the follow-up time.