Page 144 - JSOM Winter 2018

FIGURE 1 Publications included and excluded at each stage of the to –0.16 (Figure 2 and Table 2). The funnel plot (Figure 3)
literature search. GS, glucosamine sulfate. suggested publication bias because studies with larger stan-
dard errors (smaller sample sizes) were clustered farther to the
left. Publication bias was supported by the significant rank
order correlation (Table 2). Of the 17 studies identified, only 7
reported that GS significantly reduced pain.

In the stratified analysis on pain (Table 2), studies with indus-
try involvement had an SMD that was more than 3 times as
high compared with those without, although studies without
industry involvement still indicated a small pain reduction ef-
fect. Heterogeneity remained in both analyses and there was
evidence of publication bias for the nonindustry studies. Stud-
ies using a per-protocol analysis demonstrated a larger effect
on pain than those using an intention-to-treat analysis. Hetero-
geneity was high and there was evidence of publication bias in
both types of analyses. Effects on the knee appeared larger than
effects on other joints, but only two studies clearly involved
other joints. 60,75 GS provided several times per day had over
twice as large a reduction in pain than a single 1500mg dose
once a day. There was considerable heterogeneity and evidence
of publication bias in the studies involving several daily admin-
istrations, while those administering a single 1500mg dose had
75
day, (3) 400mg three times per day, or (4) a single 1,500mg little evidence of heterogeneity and marginal evidence of pub-
72
dose once daily. 60,64–66,74,77 One study had a variable dosing lication bias. Studies conducted for <6 months demonstrated
76
schedule, with 80% of participants receiving 500mg three larger effects on pain than those conducted for ≥ 6 months
times per day and 20% of participants receiving 1,500mg with high heterogeneity in the shorter studies but no evidence
once daily, although the data from all participants were an- of publication bias in longer or shorter studies. Finally, studies
alyzed together. The glucosamine industry was involved in judged to have higher methodological quality indicated that GS
six 64,65,67–69,77 of the 17 studies (35%). Methodological quality had a smaller effect on pain (but still statistically significant)
scores ranged from 55% to 90% 60,73 of available points, with with low heterogeneity; studies judged with lower methodolog-
67
a mean (SD) of 81% (9%). ical quality indicated that GS had a larger effect on pain than
the higher-quality studies, but with more heterogeneity.
Figure 2 shows the forest plot of the 17 studies examining the
effects of GS on pain; Table 2 shows the summary statistics Figure 4 shows the forest plot of the four studies that exam-
2
and Figure 3 shows the funnel plot. The Q and I statistics ined the effects of GS on joint-space changes, Table 2 lists the
both suggested heterogeneity among the SMDs (Table 2), so summary statistics, and Figure 5 shows the funnel plot. Over-
a random-effects model was used. Overall (all 17 studies), GS all, the summary SMD (Figure 4; Table 2) showed only a very
had a low to moderate effect in lowering pain compared with minor and not significant effect on joint-space changes. The
placebo, with a summary SMD of –0.35 and 95% CI = –0.54 four studies demonstrated little heterogeneity in the SMDs

FIGURE 2 Forest plot of studies examining the influence of glucosamine sulfate on pain.

GS, glucosamine sulfate; Std diff, standardized difference.

142 | JSOM Volume 18, Edition 4 / Winter 2018