Is Pi MZ a risk factor for the development of chronic obstructive pulmonary disease (COPD)? That is the question many authors have tried to answer during the last decades, but the results of published studies are still conflicting.

A very high percentage of patients with COPD have been smokers, but not all smokers develop COPD. There must be other contributing factors and, with a Pi MZ prevalence of 3–5% in many Western countries, it is relevant to determine whether this genotype is an additional risk factor for COPD.¹ Furthermore, if a dose-response relation exists, it is biologically plausible since plasma levels of α₁-antitrypsin (AAT) are reduced to about 60% in Pi MZ subjects compared with those with the normal Pi MM genotype, and Pi Z individuals with very low levels of AAT have a significantly increased risk for emphysema.

There are a number of reasons for the discrepancy between the studies conducted on this subject. Firstly, smoking is a very significant confounder in the development of emphysema which is almost impossible to control for. Secondly, many studies have been subject to various types of bias, particularly selection bias. Thirdly, only a few studies have been sufficiently large to produce a significant result and very few studies have been population based. Finally, the phenotypic appearance of the Pi MZ genotype may be heterogenic.

In the studies of a causal relationship of a risk factor (Pi MZ genotype) and the development of a disease (COPD) there are two types of designs: case-control studies and cohort studies. Cohort studies can be divided into cross sectional studies and follow up studies.

In case-control studies the researchers identify a group with COPD, find a proper control group without the disease, and compare the prevalence of the Pi MZ genotype between the two groups. The major problems are finding a proper control group from the same population as the cases and selection bias. For example, if a person is known to have the Pi MZ genotype, he may be more likely to have his lungs examined because of Pi Z patients with lung disease in his family. If his lung function is reduced, he suddenly becomes a case instead of a control.

A cohort study compares either the prevalence of COPD in a cross sectional setting or the incidence of COPD in a longitudinal setting of a group of Pi MZ individuals with a matched control group which can be derived from the general population. The longitudinal design is often superior to the case-control design but is much more expensive and time consuming and it may be underpowered. In a longitudinal cohort study it is critical to start counting person years from the date the genotype was established and only to use data collected afterwards. Some authors have used data collected before this date in the calculation of the rate of decline in forced expiratory volume in 1 second (FEV₁), and it can produce biased results.² The best cohort studies are population based with large sample sizes, long follow up times, and no selection bias, but only a few have been conducted on Pi MZ and COPD.³

When published studies are small and inadequate for drawing a firm conclusion, a meta-analysis may be the solution. In this issue of Thorax Hersh and colleagues have combined published case-control studies and cross sectional cohort studies in a meta-analysis and tried to determine whether Pi MZ is a risk factor for COPD.⁴ Their conclusion is the same as that of previous review papers—namely, that case-control studies are in favour of an increased risk and cohort studies are not.⁵ Their final conclusion is that Pi MZ individuals have a small increased risk for COPD or a fraction of Pi MZ individuals carry a larger risk.

Seersholm et al came to the same conclusion in a study of Pi MZ individuals identified through relatives of Pi Z patients either because of lung symptoms (index cases) or through family screening.⁶ The study showed an increased risk for hospitalisation for COPD in Pi MZ individuals, but sub-analysis showed the increased risk to be present only in Pi MZ individuals who were close relatives to Pi Z index patients.

If only a percentage of Pi MZ individuals are at increased risk for COPD, they must carry additional risks—either environmental, genetic, or both. Smoking is the major environmental confounder and only five of the 16 studies included in the meta-analysis corrected for smoking. The pooled odds ratio for these five studies was not significantly increased after adjustment for smoking habits. It is tempting to conclude that the apparent increased risk for COPD found in Pi MZ individuals is caused by differences in smoking habits and not by the Pi MZ genotype. If this is true, Pi MZ individuals should be more susceptible to start smoking and therefore their smoking habits are determined by their genotype. This is not unlikely if the Pi MZ persons were identified through family screening and thus are subject to selection bias. The vast majority of Pi Z index cases have been smokers and smoking habits are influenced by parents and siblings and thus aggregate in families.⁷

The difficulty in discriminating between environment and genes is further demonstrated in a recently published study by Svanes and colleagues who concluded that intrauterine and environmental exposure to parental smoking was related to the development of lung disease in adulthood.⁸ It is, however, also possible that modifier genes (Z allele or others) were responsible for some of the increased risk for lung disease.

Silverman et al investigated quantitative phenotypes in 52 Pi MZ first degree relatives of Pi Z patients with or without significant airflow obstruction. They found a trend towards lower FEV₁ in Pi MZ relatives of Pi Z subjects with airflow obstruction, a trend that could not be explained by differences in smoking habits.⁹ A Danish study of Pi Z patients found that Pi Z index cases have a worse prognosis than Pi Z non-index cases even after controlling for smoking habits, and never smoking non-index cases had no increased mortality compared with the general population.¹⁰ These studies suggest that other genetic factors contribute to the development of lung disease.

Besides protease inhibitor deficiency alleles, no candidate genes for the development of COPD have yet been identified although several have been proposed—including α₁-antichymotrypsin, glutathione S-transferase, and tumour necrosis factor α.¹¹ Linkage analysis of pedigrees ascertained through probands with severe early onset COPD who had normal AAT levels have shown a susceptibility locus on chromosome 2 and possibly several other genomic regions.¹²

The development of COPD is a complex interaction between environmental exposure and possible many genes, and whether Pi MZ is one of the genotypes remains to be shown. Observational studies and meta-analyses based on them should be interpreted very carefully, particularly in view of the impact of smoking and biased case selection. For example, if the study by Seersholm et al had included Pi MZ relatives of Pi Z index cases only, the conclusion would have been that all Pi MZ individuals are at increased risk of hospitalisation for COPD. The variability of quantitative phenotypes of homozygous AAT patients is ideal for the further study of modifier genes but, because of the rarity of the disease, international collaboration will be necessary.

To determine whether Pi MZ is a risk factor for COPD requires a large population based study with long follow up time. The most recently published study found no accelerated decline in FEV₁ in 57 Pi MZ individuals followed for an average of 15 years, but the study may be underpowered.³ In 1991–4 the Copenhagen city heart study genotyped a sample of 9000 individuals from the general population and identified 450 Pi MZ subjects. They have completed questionnaires on smoking habits and have undergone extensive testing including spirometry and, with a follow up time of more than 10 years, the study should soon come up with a sound conclusion. In the meantime, we have to advise individuals with the Pi MZ genotype about their risk for developing lung disease and, based on the current evidence, there is no reason to believe they have an increased risk as long as they do not smoke.