Introduction

One-third of the world’s population is infected with Mycobacterium tuberculosis, which causes the infectious respiratory disease tuberculosis (TB) [34]. In 2010, the World Health Organization (WHO) estimated 8.8 million new cases of TB, and 1.1 million deaths from TB among HIV-negative individuals [11, 54].

Active TB disease exerts a substantial toll on quality of life—ranging from somatic symptoms related to disease and treatment to psychological distress from social isolation and stigmatization [18, 25, 47]. The diagnosis of latent TB infection (LTBI) may be misconstrued as active TB disease or even HIV infection, both of which may also lead to further stigmatization; in some communities, the diagnosis of TB is strongly associated with HIV infection [36, 50, 58]. Health-related quality of life (HRQOL), as reported by patients, is therefore highly relevant to understanding and quantifying the true impact of TB.

The purpose of this research is to summarize the impact of TB on quantitative measures of self-reported HRQOL. We have conducted a systematic review of self-reported HRQOL among persons with active TB, LTBI, and those with persistent pulmonary symptoms following treatment of active TB. This updates and supplements a previous systematic review, which addressed HRQOL among TB patients, by expanding the search strategy and searching more databases [22].

Methods

Identification and selection of relevant publications and research

A librarian trained in systematic reviews of medical literature was consulted to construct a comprehensive search strategy. Terms included in the search strategy related to the concepts of the operational definition of HRQOL as described by Wilson and Cleary [55]. We searched 8 databases to retrieve relevant peer-reviewed publications, reporting original research: PubMed, EMBASE, EMBASE classic, PsycINFO, HaPI, BIOSIS, The Cochrane Library, and CINAHL. Databases were searched for articles published between January 1, 1960 and April 1, 2011. The search strategy is provided in “Appendix 1”.

All retrieved abstracts were exported to EndNote X4 software and screened by one author (MB); any abstract that reported self-rated, quantitative measures of HRQOL among TB subjects was eligible for full review. Two authors, fluent in English, read full-text versions of all articles written in English and completed a standardized abstraction form for all studies that evaluated HRQOL among individuals with active TB, LTBI, or post-TB pulmonary sequelae. Abstracted information was reviewed and discrepancies discussed between these two reviewers. If a discrepancy arose, the two authors reviewed the original article together to reach consensus. For those articles published in a language other than English, other research personnel fluent in the language of the publication, as well as English, completed the abstraction form. References cited among the included publications were scanned for additional potential relevant studies, which were also reviewed, if eligible, using the abstraction form.

Articles were excluded from the review if (1) quantitative measures of HRQOL were not available or if there were no subjects with TB included, (2) if data for subjects with and without TB were aggregated together, (3) if the full-text articles were not accessible to reviewers, or (4) if the publication was written in a language that the reviewers were unable to understand. Studies using the standard gamble instrument were included, since it provides an assessment of health utility, a quantitative measure of HRQOL that incorporates uncertainty, which is particularly relevant to health care decision makers.

Data extraction

The standardized abstraction form captured the following information:

  • socio-demographic characteristics of subjects

  • clinical characteristics for subjects with active TB (pulmonary/extra-pulmonary disease, smear status, re-treatment)

  • behavioral risks (smoking, alcohol abuse, and injection drug use),

  • study design features (subject recruitment mechanisms and inclusion/exclusion criteria),

  • accounting for subjects who were (a) eligible from the target population, (b) approached of those eligible, (c) recruited of those approached, (d) completed evaluations of those recruited, and (e) included in the analysis of those who completed evaluations,

  • HRQOL questionnaire administration (subject self-administered/interviewer-administered, language of administration, timing with respect to TB diagnosis and treatment, proxy respondents), and

  • HRQOL results, by subject group

The abstraction form also included a quality rating score using a 3-point scale (2 being well-described, 1 poorly described, and 0 not described in article) for each of the following eight attributes: study population, sampling mechanism, accounting for potential subjects not included in the analyses, quality check of responses, explicit description of HRQOL instruments, data entry check, training of interviewers, and discussion of study strengths and limitations. These eight items were extracted from the STROBE Statementchecklist of items that should be included in observational studies (version 4) based on study characteristics anticipated to vary widely across studies, with particular focus on methods and results [48]. Summary quality ratings were calculated by summing the scores for each of these eight categories; the minimum and maximum possible scores were 0 and 16, respectively.

Statistical analysis

Data extracted from original publications were recorded in Microsoft Excel (2010). Demographics, disease characteristics, and HRQOL measures were extracted and summarized by sub-group (active TB, LTBI, TB-free controls, etc.) Variables describing study design and methods were summarized.

We used Microsoft Excel (2010) to calculate standardized mean differences of HRQOL scores between the group of subjects treated for active TB in a given study and a concurrently evaluated comparison group. Meta-analyses were performed for standardized mean differences among studies that compared similar groups and administered the HRQOL questionnaires at similar time points with respect to TB diagnosis and/or treatment. For those publications that were deemed eligible for meta-analysis but did not report the particular measure of interest, investigators were contacted to request these data.

For each group of studies included in a given meta-analysis, we used MIX 2.0 Pro to calculate the random effects pooled estimates, using the DerSimonian-Laird method, of standardized mean differences in HRQOL scores [5, 23]. We used MIX 2.0 Pro to produce forest plots and calculate the I 2 statistics, with 95 % confidence intervals (CI) [5, 23]. These statistics allowed us to assess heterogeneity among studies’ standardized mean differences in HRQOL scores.

We used Microsoft Excel (2010) to calculate effect sizes among studies with longitudinal measures of HRQOL, comparing subsequent with initial measurements [8]. Any effect size of at least 0.50 was considered a meaningful change in HRQOL [43]. We also compared changes in HRQOL scores to previously published estimates of minimum important difference for the relevant instruments, when available.

Results

The search strategy yielded over 15,000 abstracts, 46 of which were eligible for full-text review. We looked for additional publications in the references of each of these 46 articles to retrieve an additional 30 articles for full-text review (Fig. 1).

Fig. 1
figure 1

Sampling and selection of published literature on HRQOL among tuberculosis patients from January 1, 1960–April 1, 2011. SF-36 is short form-36, VAS is visual analog scale; SGRQ is St. George’s respiratory questionnaire; WHOQOL-BREF is the World Health Organization’s Quality of Life BREF; EQ-5D is the EuroQoL 5D; SF-6D is the 6-dimension health utility scores derived from 11 items of the SF-36

Of the 76 full-text articles, 38 (representing quantitative HRQOL evaluations among 28 unique cohorts of TB patients) were relevant to our review [14, 6, 9, 10, 1221, 26, 29, 3133, 35, 37, 38, 39, 40, 45, 46, 49, 5153, 56, 57, 59, 60]. Studies included in the systematic review were published in English, Korean, Chinese, Spanish, and Turkish. Five unique cohorts contributed to meta-analyses; all articles in the meta-analysis were published in English [4, 1517, 33, 38]. Five unique cohorts contributed to analysis on meaningful changes in effect sizes of HRQOL measures over time [14, 15, 29, 32, 33].

The 28 cohorts together included 6,028 subjects from 16 countries, across 5 continents. Twenty-one studies (75 %) collected cross-sectional data, and 7 studies (25 %) studies used a longitudinal design. There were no randomized controlled trials conducted among TB patients that included quantitative, patient-reported HRQOL measures. Primary data collection for these studies occurred between 1992 and 2011.

Characteristics of all subjects

Twenty-seven of the 28 unique cohorts reported the number of women, which corresponded to 42 % of the subjects in these studies [14, 6, 9, 10, 1215, 1821, 26, 29, 3133, 35, 37, 38, 39, 40, 45, 46, 49, 5153, 56, 57, 59, 60]. Mean age (reported in 13 studies) was the most frequently reported summary measure of age [14, 6, 1315, 20, 21, 26, 32, 33, 38, 45, 46, 52, 53, 59, 60]. Although the mean age of subjects across these 13 studies was 42 years, the mean age ranged from 26 to 62 years within individual studies. Among the 5 studies reporting the proportion of foreign-born subjects in their sample, 75 % of all subjects were foreign-born, ranging from 48 to 89 % across these studies [4, 21, 29, 33, 38]. Eight studies stated they excluded known TB/HIV co-infected patients or included these patients and provided sero-status—696 (29 %) of all subjects with known sero-status were co-infected with TB/HIV, which ranged from 0 to 100 % across these studies [3, 4, 12, 16, 17, 26, 32, 33, 38]. Information on subjects’ health behaviors and socioeconomic profiles was extremely limited (Table 1).

Table 1 Total sample and sub-group characteristics among subjects of reviewed studies at the time of subjects’ initial HRQOL evaluation [14, 6, 9, 10, 1215, 1821, 26, 29, 3133, 35, 37, 38, 39, 40, 45, 46, 49, 5153, 56, 57, 59, 60]
Table 2 Discrimination of HRQOL instruments at initial evaluation by patient groups, among 24 unique studies evaluating persons with active TB [4, 1627, 38]
Table 3 Quality rating scores of articles comprising the 28 unique cohorts evaluating HRQOL among patients with active TB
Table 4 Effect sizes for longitudinal changes in HRQOL measures reported by subjects treated for active TB

Twenty-four of the 28 unique studies (representing 3,541 subjects, 59 % of the total sample) conducted quantitative assessments of HRQOL among persons with active TB [14, 6, 1215, 18, 20, 21, 26, 29, 32, 33, 35, 38, 40, 45, 46, 49, 5153, 59, 60]. Fourteen of these studies concurrently evaluated HRQOL among other patient groups—6 studies evaluated persons with LTBI, while 8 evaluated TB-free, healthy control subjects [13, 1517, 20, 21, 33, 38, 45, 46, 5153, 59] (Table 1).

The remaining 4 of the 28 unique studies evaluated HRQOL among subjects with patients with post-TB sequelae who developed chronic alveolar hypoventilation (CAH) and were using home mechanical ventilation (HMV); see “Appendix 4” and Table 7 in Appendix [9, 10, 19, 37, 39, 56, 57].

Classification of TB patient groups

Fifteen of the 24 studies evaluating HRQOL among patients with active TB specified that disease diagnosis was based on smear and/or culture confirmation, and/or the use of standardized clinical and radiographic criteria (e.g., those of the WHO) [3, 4, 12, 1518, 26, 29, 32, 33, 38, 45, 46, 49, 51, 52, 60]. Nineteen studies (2,586 subjects with active TB, 73 %) specified the disease site of these subjects—2,350 (91 %) had pulmonary TB; 888 (38 %) of these individuals were sputum smear-positive at their initial HRQOL evaluation, which generally indicates more severe and contagious disease [14, 1215, 20, 26, 29, 32, 33, 35, 38, 49, 5153, 59, 60].

A total of 639 subjects treated for LTBI (11 % of the total sample) were concurrently evaluated in 6 studies that assessed HRQOL among subjects with active TB [4, 16, 17, 21, 33, 38, 45, 46]. Five of these 6 studies explicitly stated that subjects with LTBI were diagnosed by a positive tuberculin skin test (TST), representing 533 (88 %) of all LTBI subjects [4, 16, 17, 33, 38, 45, 46].

A total of 1049 healthy control subjects (17 % of the total sample) were concurrently evaluated in 8 studies that assessed HRQOL among subjects with active TB [1, 4, 15, 20, 45, 46, 49, 51, 52, 59]. Confirmation of healthy status was described in only 3 studies; diagnostic criteria included a physical exam plus chest radiographs and electrocardiograms, and/or a negative TST result [4, 20, 45, 46].

HRQOL and health utility instruments

Thirty-four different HRQOL and health utility instruments were used among the studies described in this review (Table 5 in Appendix). The most commonly used tool was the SF-36; 8 of the 28 unique cohorts reported HRQOL measures from the SF-36 [3, 4, 17, 21, 29, 33, 35, 51, 52]. Most studies used only one instrument, but three studies used as many as four tools [16, 17, 29, 39]. Of those studies reporting the method of assessment, 11 used only interviewer-administered questionnaires, 9 studies used only self-administered questionnaires, and 5 used both [14, 6, 9, 10, 12, 1519, 21, 26, 31, 33, 35, 37, 38, 39, 40, 45, 46, 49, 51, 52, 56, 57, 59, 59]. One study permitted proxy respondents [60]. Twenty-one of the 28 studies that evaluated subjects treated for active TB stated that the administered HRQOL instrument was previously validated or that their research was the validation study for this tool [24, 6, 919, 21, 26, 29, 31, 33, 35, 37, 38, 39, 40, 45, 46, 49, 5153, 56, 57]. Four of these studies explicitly described ceiling effects, and three explicitly addressed floor effects (Table 2) [4, 1627, 38].

Assessment of data quality

The mean quality rating score was 7.3, with scores ranging from 2.0 to 13.0. The mode was 8.0. (The greater the study rating score, the better the perceived quality). Four of the 28 unique cohorts reported a process to check for questionnaire comprehension [4, 16, 17, 35, 45, 46]. However, only one reported the numbers of subjects removed because of poor comprehension [16, 17]. Twelve reported the number of subjects who met the investigators’ inclusion criteria [9, 10, 16, 17, 19, 21, 26, 32, 33, 35, 37, 38, 39, 40, 45, 46, 53]. The proportion of potential subjects who refused participation ranged from 0 to 37 %, while the proportion of subjects who were lost to follow-up ranged from 0 to 8 % (Table 3).

Meta-analyses

Separate meta-analyses were performed for HRQOL measures and health utility measures. Data from three unique cohorts, using two unique instruments, contributed to the meta-analysis of standardized mean differences in HRQOL between subjects treated for active TB and subjects treated for LTBI within 2 weeks of TB diagnosis [4, 16, 17, 33]. The random effect pooled estimate for the standardized mean difference was −0.66 (95 % CI −0.82, −0.50), and the I 2 statistic was 17 % (0, 76 %). We then repeated the meta-analysis, excluding the unpublished data of Bauer et al. The random effect pooled estimate for the standardized mean difference in HRQOL scores was −0.58 (95 % CI −0.75, −0.40), and the I 2 statistic was 0 % (0, 79 %). Data from two cohorts, using two instruments, contributed to the meta-analysis of standardized mean differences in HRQOL between subjects treated for active TB and subjects treated for LTBI after completing 6–8 months of treatment [33, 38]. The random effects estimate for the standardized mean difference was −0.51 (−0.77, −0.26), and the I 2 statistic was 54 % (0, 87 %) (Fig. 2).

Fig. 2
figure 2

Standardized mean differences between groups of subjects treated for active TB compared to groups of subjects treated for latent TB infection, stratified by timing of HRQOL evaluation with respect to TB diagnosis and treatment

Two cohorts (one unpublished), using two instruments, met our meta-analysis criteria and evaluated subjects treated for active TB compared to healthy controls within 2 weeks of TB diagnosis [4, 15]. The random effects pooled estimate of the standardized mean differences in HRQOL of these two patients groups was −1.14 (−1.75, −0.54), and the I 2 statistic was 0 % (0, 85 %). In both cohorts, the standardized mean differences indicated lower mean HRQOL scores reported by subjects with active TB disease than by healthy control participants. Findings were more similar for standardized mean differences in SF-36 PCS scores (−1.22, 95 % CI −1.58, −0.85; Bauer et al. [4]) and WHOQOL-BREF scores (−1.62, 95 % CI −1.96, −1.28; Dhuria et al. [15]) than for SF-36 MCS scores (−0.58, 95 % CI −0.93, −0.24; Bauer et al. [4]).

Two cohorts (one unpublished), using two instruments, contributed to the meta-analysis of standardized mean differences in health utilities reported by subjects treated for active TB and subjects treated for LTBI, within 2 weeks of their TB diagnosis [4, 16, 17]. The random effects pooled estimate of the standardized mean differences in health utilities of these two patients groups was −0.62 (−0.82, −0.42), and the I 2 statistic was 89 % (69, 96 %). The standardized mean differences in Standard Gamble scores (−0.65, 95 % CI −0.96, −0.33) and in SF-6D scores (−0.76, 95 % CI −1.08, −0.44) in the unpublished cohort of Bauer et al. [4] suggested a greater decrement in health utility than previously reported by Dion et al. In the latter study, the standardized mean differences were −0.19 (95 % CI −0.82, 0.45) for Standard Gamble scores [16] and −0.42 (95 % CI −1.06, 0.21) for EQ-5D scores [17].

Further details about HRQOL and health utility scores in these studies are presented in “Appendix 2” and Table 6 in Appendix.

Meaningful changes in longitudinal HRQOL measures

Five of the 7 unique cohorts using a longitudinal design reported mean HRQOL scores from subjects treated for active TB at at least two time points; one of these five cohorts also reported longitudinal HRQOL scores for subjects treated for LTBI [1315, 29, 32, 33]. Table 4 displays the calculated effect sizes, comparing later with initial values. Additional information is presented in “Appendix 3”.

The greatest improvement in HRQOL occurred during the first 2–3 months of treatment, among studies that evaluated HRQOL among subjects treated for active TB at the beginning of treatment, after the initial phase of treatment, and at/near the end of treatment [1315, 32]. Based on the effect sizes, instruments assessing mental well-being also indicated meaningful improvements in HRQOL after 2 months of treatment [29].

Among subjects treated for LTBI, longitudinal measurements of HRQOL did not suggest meaningful changes between diagnosis and 6 months of treatment [33].

Discussion

Subjects with active TB consistently reported poorer HRQOL than subjects treated for LTBI and untreated controls, across a variety of questionnaires and settings. For example, random effects estimates of pooled standardized mean differences demonstrated that subjects treated for active TB had mean scores 0.66 and 0.51 standard deviations below those treated for LTBI within 2 weeks of diagnosis and after 6–8 months of treatment, respectively. Pooled estimates of standardized mean differences in health utilities among subjects treated for active TB compared to those treated for LTBI within the first 2 weeks of treatment showed similar results. The difference between subjects treated for active TB and healthy controls was even more pronounced. Other studies of HRQOL and health utility measures identified by our systematic review, but which could not be meta-analyzed, also reported a consistently detrimental effect of active TB.

Based on our effect size analysis, we saw a meaningful improvement in HRQOL throughout treatment of active TB, and particularly during the initial, intensive phase.

Health care providers encounter detrimental effects of TB on HRQOL particularly in their patients with active TB disease, but to a lesser degree among those treated for LTBI. Indeed, one study documented poorer HRQOL among subjects treated for latent TB, compared to healthy controls [4]. These decrements are also highly relevant to decision makers, in approaching tradeoffs between providing preventive treatment for a large number of persons with LTBI, versus the smaller number who may ultimately develop active TB [28, 30].

It is worth noting that available studies primarily reflect the experiences of young and middle-aged, predominantly male adults—corresponding to the profile of reported TB cases in most countries [11]. Other socio-demographic characteristics of subjects varied widely, although data were sparse. Over 75 % of subjects with reported immigration status were immigrants to low TB incidence countries. Immigrants also face a unique set of challenges—adapting to new cultural norms and languages, obtaining and sustaining paid work, and accessing health care (for TB diagnosis/treatment and otherwise). These may all independently and synergistically affect HRQOL. Our results may therefore be particularly relevant to low-incidence countries, as immigrants represent the majority of persons screened, diagnosed, and treated for active TB disease and LTBI in such settings.

With a comprehensive search strategy yielding 76 articles, this systematic review builds on a review published in 2009 and is the first to provide pooled estimates from formal meta-analyses [22]. In this earlier review, subjects with active TB were also shown to have substantial deficits in HRQOL, compared to subjects treated for LTBI. Mental well-being was more severely disrupted than physical health among patients in both treatment groups. Our results are also similar with respect to improvement during treatment of active TB [22].

While 28 unique cohorts of patients with TB were included in our systematic review, 3 potentially relevant articles were excluded because they were published in Russian. Hence, our estimates of HRQOL may be affected by some selection bias. However, the main findings of our review were consistent across a variety of settings.

Some information could be gleaned from the English abstracts of two of these articles. Sukhov and Sukhov found that men with chronic pulmonary TB rated their HRQOL worse than men with their first case of pulmonary TB [44]. Shalaeva et al. [42] reported improvements in all sub-scales of the SF-36 among 59 adults who received surgical treatment and chemotherapy for pulmonary TB. This latter finding in particular supports findings of improved HRQOL scores throughout active TB treatment among the studies included in this review.

Only 2 studies retrieved from the search included patients who were co-infected with TB and HIV. Additionally, only 1 study included in this review evaluated subjects with multi-drug resistant TB (MDR-TB), and no studies evaluated children less than 11 years of age. Hence, our systematic review could not adequately address HRQOL among these groups, whose experience may differ substantially from that of other TB patients.

This systematic review synthesized the information from studies of quantitative evaluations of HRQOL; we did not review the body of research using qualitative methods to address HRQOL as it was beyond the scope of our study objectives. However, qualitative studies can offer health care providers valuable insight regarding patients’ experiences and needs in specific settings and should be used to supplement the quantitative information provided in this review.

The conclusions drawn from a systematic review depend on the quality of the individual studies included. One major limitation of the studies is that very few adequately measured and described key social and behavioral determinants of TB [11]. For this reason, our meta-analysis was limited to crude pooled estimates of HRQOL. We did not have sufficient information to permit meta-regression, which could help account for important confounders (e.g., foreign birth, substance use, co-morbidities) when comparing persons with active TB to those treated for LTBI and healthy control subjects.

Most studies were cross-sectional, and none reported measurements from a randomized clinical trial. Longitudinal observational studies can provide valuable insight into changes in HRQOL as patients undergo different phases of treatment, particularly for active TB. A longitudinal design, where subjects with active and LTBI are compared to suitable controls, will be particularly useful in this respect.

Inaccurate measures of HRQOL were possible as only 3 of the 28 cohorts included a process for checking questionnaire comprehension, and only one indicated the number of subjects excluded accordingly. Persons so excluded may have more limited educational attainment or language skills, which may make study samples less representative of the TB patient population, and perhaps lead to overestimation of HRQOL because of this exclusion. Additionally, one study contributing data to meta-analyses was not yet peer-reviewed [4]. Sensitivity analyses with removal of these data were performed accordingly, with generally similar findings.

Refusals ranged from 0 to 37 % in the 12 studies that provided this information. It is possible that subjects who refused to participate had more severe TB disease and/or a higher prevalence of risk behaviors, though this could not be assessed directly. These refusals could also make study samples less representative and potentially bias our results. Similarly, certain eligibility criteria might limit the representativeness of some studies.

Finally, most included studies appeared to be of moderate quality, with the most frequent quality score being 8 out of a possible 16 points. Although this rating system was not formally validated, it may highlight gaps in reporting parameters that are relevant to the assessment of HRQOL in the TB patient population.

Conclusions

In a variety of studies, subjects with active TB consistently reported poorer HRQOL than persons treated for LTBI and untreated controls. This is important for understanding the non-fatal outcomes of active TB and the potential benefits of preventive interventions.

Future research on HRQOL in the TB context should better address social and behavioral health determinants. Further information is also needed for some of the most vulnerable persons with TB, such as those with TB/HIV co-infection, MDR-TB, and younger children.

In the TB context, meaningful cross-sectional differences and longitudinal changes have not been defined for many of the measurement tools we reviewed. A longitudinal study now underway compares HRQOL and health utilities over a 12-month follow-up period, among persons treated for active TB, LTBI, and untreated, healthy controls of similar background. This research will help address this gap in low-incidence settings, allow for better assessment of the benefits and limitations of TB control interventions, and may assist health care providers to better target physical and psychosocial support.