Results of the search

The updated searches in 2013 retrieved a large number of references: 151 from the Cochrane Neuromuscular Disease Group Specialized Register, 219 from CENTRAL, 333 from MEDLINE, 525 from EMBASE, 4 from LILACS and 458 from the Chinese Biomedical Retrieval System. There were also 10 references from DARE, 1 from HTA and 9 from NHSEED. In this update, we did not find any newly published studies that appeared to meet the inclusion criteria. As a result, there were still 21 potentially eligible trials identified by scrutinising the titles and abstracts. We excluded 15 trials after we screened their full text. Six trials fulfilled the inclusion criteria (Huff 1988; Wood 1988; Morton 1989; Harding 1991; Tyring 1995; Whitley 1996). Our searches of trials registries did not identify any ongoing trials.

Included studies

We found many studies of antiviral treatment for acute zoster, but only a few studies addressed relevant end points, such as the incidence and duration of PHN or the associated pain after the acute phase. Among the studies, six trials with a total of 1211 participants were suitable for inclusion.

The reports of all included studies explicitly stated Inclusion and exclusion criteria. All the studies recruited adults with the characteristic rash of herpes zoster that had been present for 72 hours or less. One study was limited to people over 50 years of age (Whitley 1996), while another study involved people over 60 years of age (Wood 1988); the other trials only required adults to be older than 16 to 20 years. Sample sizes differed between trials, ranging from 46 to 419 participants (three studies had more than 100 participants (Wood 1988; Tyring 1995; Whitley 1996)). Each trial reported the demographic characteristics of participants, including gender distribution, mean age and mean duration of rash before enrolment; they recorded no significant differences. Altogether, 580 males and 614 females were included in our review (discrepancies in the number of participants were due to occasional missing items of data in some trials). More than 831 (69%) participants were 50 years of age or older. The locations of rashes were reported in several trials (Wood 1988; Morton 1989; Tyring 1995), and one trial enrolled people with acute herpes zoster ophthalmicus only (Harding 1991). Common exclusion criteria across the trials included: immunosuppression; pregnant or nursing women; hepatic or renal dysfunction; systemic glucocorticosteroid therapy or any antiviral therapy; people who required immunosuppressive therapy; cancer; and the presence of crusts on herpetic lesions at enrolment. No trial evaluating the efficacy of antiviral treatment for people who were immunocompromised met the inclusion criteria for our review.

Oral aciclovir was the most commonly used antiviral treatment for herpes zoster, and five of the six studies evaluated its effect, with treatment commencing within 72 hours of the zoster rash. These five studies used oral aciclovir at a dose of 800 mg five times daily: two studies administered the antiviral for seven days (Wood 1988; Morton 1989), two studies administered the antiviral for 10 days (Huff 1988; Harding 1991), and the remaining study administered the antiviral for 21 days (Whitley 1996). Huff 1988 additionally studied a 400 mg dosage five times per day for 10 days. Whitley 1996 randomised participants to aciclovir, prednisone, both aciclovir and prednisone, or neither aciclovir or prednisone, for 21 days.

In the sixth included trial (Tyring 1995), participants were assigned to one of the following regimens: famciclovir 500 mg, famciclovir 750 mg or matched placebos, identical in taste and appearance three times daily for seven days from within 72 hours of onset of the rash. All aciclovir trials followed the participants for six months or longer after enrolment, while in the famciclovir study participants were followed for five months after the lesions healed. Follow‐up included the following approaches: participants returned to see the physician at a prescribed time or as often as clinically necessary (Harding 1991; Tyring 1995); investigators contacted participants by telephone (Huff 1988); or a research nurse or assistant assessed the participants at home (Wood 1988; Morton 1989). The remaining trial did not describe the method of follow‐up (Whitley 1996). The six trials reported loss to follow‐up and withdrawal, but one trial did not refer to the reasons for loss to follow‐up or withdrawal (Huff 1988). Only two trials stated that an intention‐to‐treat analysis was performed (Tyring 1995; Whitley 1996), which was not described in the other trials.

The outcome measures differed between the trials. All the included trials reported the presence or duration of PHN, which was defined with different cut‐off times and intensities of pain. Severity of pain was referred to as a separate outcome in several trials, but measured in different ways: specifically, Harding 1991 used a VAS set vertically at 100 mm, and compared pain scores directly between groups, while Wood 1988 recorded pain by asking participants to mark a linear scale ranging from 'none' to 'very severe', and for analysis, translated the mark on this linear scales to a score of zero to three by quartering. In the other trials, the intensity of pain was only evaluated as 'absent', 'slight', 'moderate', or 'severe', and was often recorded in the acute phase only. Quality of life was measured in Whitley 1996 and included the times to cessation of acute neuritis, return to uninterrupted sleep, and return to 100% usual activity, but it was only evaluated during the first month after disease onset. None of the trials reported quality of life measured with the SF‐36 after six months. All included trials reported adverse events during treatment, or within two weeks of stopping treatment. Adverse events were categorised as 'serious' or 'not serious'.

Excluded studies

The main reasons for exclusion of the 15 potentially relevant studies were as follows: the follow‐up was too short or irregular to obtain data for evaluation of PHN (Galbraith 1983; Wassilew 1987); the control group was not treated with a placebo but another agent (Benoldi 1991); there was no comparison with a placebo, or no treatment group (Hoang‐Xuan 1992; Wood 1994); the length of time from onset of herpes zoster to treatment was over 72 hours (Esmann 1982; McGill 1983; Cobo 1986; Payne 1989); topical antiviral treatment was used (Mandal 1988); and examination of the methods showed the study to be neither an RCT nor a quasi‐RCT (Mondelli 1996); only an abstract was available, with insufficient data to assess eligibility (Dekonenko 1998); no report was available (in spite of attempts to obtain further details) (Dekonenko 1999); and one study did not specify a limit for the course of the disease in the inclusion criteria (we have as yet received no reply to an email to the author requesting detailed information) (Varotti 2001). See Characteristics of excluded studies.

Allocation

Trial reports of all the included studies stated that they were RCTs. Four of the studies reported the method of randomisation (Huff 1988; Wood 1988; Morton 1989; Whitley 1996). The Huff, Whitley and Wood studies used a computer‐generated randomisation code to randomly assign participants to the treatment groups (Huff 1988; Wood 1988; Whitley 1996). In Morton 1989, the laboratories of the centre conducting the trial provided each course of study medication in a numbered and randomised bottle. Harding 1991 and Tyring 1995 did not describe the method of randomisation. In two studies, allocation concealment was probably done: in Wood 1988, participants and investigators enrolling participants could not foresee assignment because the allocation was done by the study centre, and in Morton 1989, the centre conducting the trial randomised sequentially‐numbered bottles containing trial treatment courses. However, it was unclear from the reports if there was adequate allocation concealment in the other included studies. Therefore, risk of selection bias due to generation of a randomised sequence or its concealment prior to assignment was unclear in those four trials (Huff 1988; Harding 1991; Tyring 1995; Whitley 1996).

Blinding

All of the six included trials were double blind, using placebo in the control group. Most of the trials used matching placebo identical to the active drug, but only described blinding in broad terms, such as 'double blind,' which made it impossible to know exactly who was blinded (Higgins 2011a). Only two trials reported blinding in detail: Whitley 1996 stated, "all research personnel remained blinded to drug assignment until the study was completed and the database was locked", and Morton 1989 specified that "the investigators, research nurse, general practitioners, and patients were all fully blinded to treatment status". Further all the outcomes, including persistence and severity of pain and adverse events, were participant‐reported, so the risk of bias from knowledge of the allocated interventions by participants and personnel during the study or by outcome assessors was low in those two trials, but there was insufficient information for the other four trials (Huff 1988; Harding 1991; Tyring 1995; Wood 1988) to make an assessment.

Incomplete outcome data

The included studies all reported the time of follow‐up: one study used five months (Tyring 1995), and five studies used six months (Huff 1988; Wood 1988; Morton 1989; Harding 1991; Whitley 1996). Five of the studies reported information about follow‐up and dropouts, and clearly described the reasons for dropout; Huff 1988 did not mention the reason for dropout. Harding 1991 considered the presence of bias due to incomplete outcome data since four participants withdrew because of side effects, and the study analysed data on the remaining participants (Harding 1991). One study claimed to have used intention‐to‐treat analysis (Whitley 1996); however, the other five included studies did not state whether or not the analysis was by intention to treat. Nevertheless, we were able to obtain sufficient information from the trial report to restore dropouts to the correct group and perform an intention‐to‐treat analysis in our review.

Selective reporting

Since most of the included trials were designed primarily to assess efficacy of antivirals during the acute phase of herpes zoster and all reported negative results or marginal positive results for antiviral treatment for preventing PHN, we considered reporting bias introduced by the investigators to be small.

Other potential sources of bias

Differences in the definition of PHN and end outcome measurements in the included trials made our review of the relevant studies complicated. Although we stated the definition of PHN in our review as 'persistent or recurrent pain at the site of shingles at least 120 days after the onset of the acute rash', we still evaluated all relevant studies which adopted different definitions of PHN to avoid missing potential data. Instead, we have attempted to perform a subcategory analysis of different study data using alternative cut‐off times.

We also considered performance bias due to use of other treatment measures. Exclusion criteria usually ruled out people on other relevant treatments from the studies but in some trials, other conventional medications, such as corticosteroids, analgesics, anti‐inflammatory drugs, topical preparations and antidepressants, were permitted during treatment and follow‐up (Morton 1989; Harding 1991; Whitley 1996). Although study reports stated that there was no significant difference in the use of any additional medications between the treatment and placebo groups, we thought this might have had some degree of impact on the results.

In several trials, the trialists asked participants to contact investigators if PHN developed or to recall their condition over the entire period at one time point of follow‐up. Recall bias in reporting of the condition or inaccurate recall of adverse events could have produced error in these trials (Huff 1988; Morton 1989; Wood 1988).

Primary outcome measure

The presence of PHN six or four months after the onset of acute herpetic rash

Aciclovir versus placebo

Five trials compared oral aciclovir with placebo, and two of the trials (Wood 1988; Whitley 1996), with a total of 476 participants, provided relevant data on the presence of PHN six months after the onset of the acute rash. Whitley 1996 recruited a total of 100 participants in the oral aciclovir and placebo group and only reported the results with RRs and 95% CIs, without detailed information on the number of participants with PHN during the follow‐up period (we contacted the trial author for the primary data but did not receive any reply). Therefore, we conducted the meta‐analysis in RevMan 5 with the inverse variance method for the two trials (Wood 1988; Whitley 1996). In the meta‐analysis, the incidence of PHN events for the aciclovir group was not significantly different from that for placebo (RR 1.05, 95% CI 0.87 to 1.27, P = 0.62; Analysis 1.1; Figure 2), which indicated that oral aciclovir did not significantly prevent PHN six months after the onset of rash.

Figure 2

---

---

Forest plot of comparison 1: Oral aciclovir versus placebo or no treatment, outcome 1.1 The presence of PHN 6 months after the onset of the acute herpetic rash.

We also conducted a meta‐analysis for the presence of PHN using a cut‐off time of four months after rash onset, since half of the included trials reported relevant data at this cut‐off time point and so met the definition of PHN in the present review (at least 120 days from rash onset) (Desmond 2002). The three trials included a total of 609 participants (Huff 1988; Wood 1988; Harding 1991). The incidence of PHN was slightly, but not significantly, less in the treated than in the placebo group (RR 0.75, 95% CI 0.51 to 1.11, P = 0.15; Analysis 1.2; Figure 3).

Figure 3

---

---

Forest plot of comparison 1 Oral aciclovir versus placebo or no treatment, outcome 1.2 The presence of PHN 4 months after the onset of the acute herpetic rash.

Famciclovir versus placebo

The only trial comparing two dosages of famciclovir with placebo (Tyring 1995), defined PHN as pain after rash healing, and followed participants up five months after rash healing. No detailed data were obtained about the presence of PHN six months after the rash onset. The proportion of participants who had PHN was similar in each of the three treatment groups: 44.2% (61/138 participants) for the 500 mg dose, 50.4% (68/135 participants) for the 750 mg dose, and 38.4% (56/146 participants) for placebo. For the 500 mg and 750 mg doses respectively, RRs of PHN versus placebo were 1.15 (95% CI 0.87 to 1.52) and 1.31 (95% CI 1.01 to 1.71) (Analysis 2.1). The trial indicated that oral famciclovir did not significantly prevent PHN after rash healing, although the article authors reported that famciclovir could promote the resolution of PHN (approximately two‐fold faster than placebo).

Secondary outcome measures

Adverse events during treatment, or within two weeks of stopping treatment

Adverse events included serious and non‐serious events, and we analysed them as a whole. Table 1 provides details for individual studies.

Open in table viewer

Table 1. Adverse events in included studies

Study name	Antiviral agent	Adverse events
Harding 1991	Aciclovir	No serious adverse events developed. 12 of the aciclovir recipients (50%) and 11 of the placebo recipients (50%) reported non‐serious adverse events. Nausea and vomiting were most frequent. Other infrequent adverse events included anorexia, malaise, dyspepsia, heartburn and depression, with an even spread between the two groups
Huff 1988	Aciclovir	No serious adverse events developed. Non‐serious adverse events such as nausea, diarrhoea and headache developed in both groups (aciclovir and placebo), without significant differences
Morton 1989	Aciclovir	In the week of treatment, there was no significant difference between groups in the numbers of events reported, with a mean of 2.7 events per participant on aciclovir and 3.1 on placebo. In the second week there were 1.3 events per participant on aciclovir, just significantly fewer than the 1.7 events per participant on placebo (P = 0.044). There were 2 deaths during the study, 1 in each group. None was proven to be drug related
Tyring 1995	Famciclovir	No serious adverse events developed. The most frequently reported non‐serious adverse events were headache (23.2% of participants receiving 500 mg of famciclovir, 22.2% of participants receiving 750 mg of famciclovir, and 17.8% of participants receiving placebo, respectively) and nausea (12.3%, 12.6% and 11.6% of participants, respectively)
Whitley 1996	Aciclovir	The most frequently reported clinical adverse events were gastrointestinal symptoms, especially nausea and vomiting. The rate of adverse events was lower in the placebo group (12%) than in the aciclovir plus prednisone placebo group (27%), without significant difference
Wood 1988	Aciclovir	A large number of adverse events were reported, including depression, headache, anorexia, nausea, tiredness, and vomiting. During treatment, 124 aciclovir recipients reported a total of 302 adverse events, and 129 placebo recipients reported a total of 359 adverse events. The incidence was similar in the two treatment groups. None of the adverse events was considered serious

Oral aciclovir versus placebo

None of the included trials reported any serious adverse effects attributable to the experimental treatment during treatment or within two weeks of stopping treatment. The most commonly reported non‐serious adverse events were nausea, vomiting, diarrhoea and headache, which were recorded in most of the included trials. Morton 1989 reported untoward medical events as numbers of mean events per participant, which was just significantly different between the two groups in the week after treatment (1.3 and 1.7 events per participant on aciclovir and placebo respectively, P = 0.044). These data were not suitable for inclusion in the meta‐analysis for the incidence of adverse events.

The meta‐analysis of four aciclovir trials indicated that the incidence of adverse events for oral aciclovir (178/355 participants (50.1%)) was not significantly different from placebo (174/354 participants (49.2%); RR 1.01, 95% CI 0.88 to 1.15, P = 0.91; Analysis 1.4; Figure 4) (Huff 1988; Wood 1988; Harding 1991; Whitley 1996).

Figure 4

---

---

Forest plot of comparison 1 Oral aciclovir versus placebo or no treatment, outcome 1.4 Adverse events.

Oral famciclovir versus placebo

The trial (419 participants) comparing famciclovir with placebo concluded that famciclovir was well tolerated, with a safety profile similar to that of placebo (Tyring 1995). The most frequently reported adverse events were headaches and nausea. The frequency of adverse events in the famciclovir and placebo groups was similar (see Table 1). Investigators classified adverse events as "related", "possibly related to study medication", "unknown", or as cases in which assessment was missing. The most common events related to study medication were headache (8.0%, 8.1%, and 6.8% of participants receiving 500 mg famciclovir, 750 mg famciclovir and placebo, respectively) and nausea (5.1%, 3.0% and 8.2% of participants, respectively), but without significant differences. There were no reports of any serious adverse events in this trial.

Additional outcome measure not specified in the protocol

The presence of herpetic neuralgia one month after the onset of acute herpetic rash

Four trials recorded relevant data on the presence of herpetic neuralgia one month after the onset of the acute herpetic rash (Huff 1988; Wood 1988; Morton 1989; Harding 1991). A meta‐analysis of these trials, with a total of 692 participants, showed a significantly lower incidence of pain in the aciclovir group (153/347 participants (44.1%)) than in the placebo group (184/345 participants (53.3%)) (RR 0.83, 95% CI 0.71 to 0.96, P = 0.01; Analysis 1.3; Figure 5).

Figure 5

---

---

Forest plot of comparison 1 Oral aciclovir versus placebo or no treatment, outcome 1.3 The presence of herpetic neuralgia 1 month after the onset of the acute herpetic rash.

No data on this outcome were reported in the famciclovir trial (Tyring 1995).

Subgroup analyses

Sensitivity analysis

No significant heterogeneity was present when we conducted a meta‐analysis to investigate oral aciclovir for preventing PHN six months after the onset of the acute herpetic rash. To perform a sensitivity analysis, we excluded trials with a high or uncertain risk of bias (Huff 1988; Morton 1989; Harding 1991), which left one trial with a low risk of bias (Wood 1988). Wood 1988 reported the presence of herpetic neuralgia comparing aciclovir with placebo treatment by monthly intervals after the onset of the acute herpetic rash (111/190 versus 110/186 at one month, P = 0.89; 26/190 versus 26/186 at four months, P = 0.93; and 22/190 versus 20/186 at six months, P = 0.80). As a result of the sensitivity analysis, the results of the primary outcome did not change, still without a significant difference between groups (RR 1.08, 95% CI 0.61 to 1.91, P = 0.80, Analysis 4.1; Figure 6). The results for the presence of herpetic neuralgia one month after the onset of the acute herpetic rash changed from a significant difference (RR 0.83, 95% CI 0.71 to 0.96, P = 0.01) to no significant difference (RR 0.99, 95% CI 0.83 to 1.17, P = 0.89).

Figure 6

---

---

Forest plot of comparison 4 Sensitivity analysis, outcome 4.1 The presence of postherpetic neuralgia.